Using AI to 4x Your Support Team - CourtneyR.dev's 2nd Brain

Using AI to 4x Your Support Team 🔗 **Source:** [Using AI to 4x Your Support Team](https://www.youtube.com/watch?v=mhLKmptjk4U) > [!note]+ Document Notes > tags: [ai-support, wordpress-product, customer-experience] > author: WP Product Talk > source: YouTube > reference: https://www.youtube.com/watch?v=mhLKmptjk4U > date created: 2024-12-29 14:23 > > # 🏷️ Title: Using AI to 4x Your Support Team >  ## 🧰 Action Inbox - [H1] Track ownership and expertise metrics before implementing AI — measure what you're amplifying #support - [H2] Set up AI tools with proper context and data sources — Help Scout, DocsBot, custom systems #automation - [H3] Train support agents to ask better questions — AI amplifies existing processes #training ## 🧭 Table of Contents - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-assessment) - [✅ Logical Validity](#-logical-validity) - [🛡️ DARVO Analysis](#-darvo-analysis) - [🎯 Conclusion](#-conclusion) ### 🔗 References - [Sustainable Support Book](https://sustainablesupport.ca/) - [Help Scout AI Features](https://helpscout.com/) - [DocsBot.ai](https://docsbot.ai/) ## 📝 Summary - Ben Meredith discusses using AI to amplify support team effectiveness by 4x - AI amplifies whatever systems are already in place - good or bad - Key is fixing underlying support processes before adding AI automation - Focus on ownership, expertise, and customer-centric metrics - Human oversight remains critical for quality and accountability ## 📖 Main Content Ben Meredith, author of "Sustainable Support," argues that AI doesn't automatically improve support - it amplifies existing systems. If your support is broken, AI will make it 4x more broken. If it's good, AI makes it 4x better. The key is establishing proper metrics, processes, and culture before implementing AI tools. This includes measuring agent ownership, expertise, and customer outcomes rather than vanity metrics like response time. AI works best when given comprehensive context about customers, products, and processes, but still requires human judgment for complex decisions and maintaining authentic customer relationships. ## 🗒️ Notes #### ⏱️ [00:00](https://youtu.be/mhLKmptjk4U?t=0) — [H1] AI amplifies whatever exists - Key idea: AI will 4x your support quality, whether good or bad - Evidence: "Whatever currently exists on your support team will be amplified 4x by AI" #### ⏱️ [05:30](https://youtu.be/mhLKmptjk4U?t=330) — [H2] Fix problems before adding AI - Key idea: Don't throw AI at broken systems - Evidence: "If you've already got a problem, don't just throw AI at it. Fix the underlying root problem" #### ⏱️ [08:45](https://youtu.be/mhLKmptjk4U?t=525) — [H1] Measure what agents can control - Key idea: Focus on metrics agents can directly influence - Evidence: "I want metrics that the agent themselves can control - their expertise, ownership" #### ⏱️ [12:20](https://youtu.be/mhLKmptjk4U?t=740) — [H2] All y'all vs the problem approach - Key idea: Support agents should unite everyone against the problem - Evidence: "Arrange two teams: the problem on one side and all y'all on the other side" #### ⏱️ [18:15](https://youtu.be/mhLKmptjk4U?t=1095) — [H2] Context is key for AI effectiveness - Key idea: More context produces better AI results - Evidence: "The more context you give AI, the better results it's going to produce" #### ⏱️ [22:30](https://youtu.be/mhLKmptjk4U?t=1350) — [H3] Multiple AI tools reduce support load - Key idea: Self-service AI tools handle 30-40% of simple questions - Evidence: "DocsBot and Help Scout Beacon have reduced our support load by 40%" #### ⏱️ [26:45](https://youtu.be/mhLKmptjk4U?t=1605) — [H2] Humans still needed for judgment - Key idea: AI can't reliably handle multi-step decision making yet - Evidence: "We're not there right now for a system that reliably could answer multi-step questions" #### ⏱️ [31:10](https://youtu.be/mhLKmptjk4U?t=1870) — [H2] AI enables deeper debugging - Key idea: AI can reproduce issues and analyze code automatically - Evidence: "AI can reproduce the setup, read the codebase, identify if code might have a bug" #### ⏱️ [35:20](https://youtu.be/mhLKmptjk4U?t=2120) — [H3] Ownership prevents technical debt - Key idea: Human oversight prevents AI from creating quick fixes - Evidence: "Sounds like technical debt looking for a place to happen without human involvement" #### ⏱️ [40:15](https://youtu.be/mhLKmptjk4U?t=2415) — [H2] AI struggle with authentic voice - Key idea: AI can't capture personal communication style reliably - Evidence: "I tried to train it to do something in my tone and it sounds nothing like me" #### ⏱️ [44:30](https://youtu.be/mhLKmptjk4U?t=2670) — [H1] Accountability requires humans - Key idea: Computers can't be held accountable for decisions - Evidence: "A computer can never be held accountable, therefore must never make management decisions" #### ⏱️ [48:45](https://youtu.be/mhLKmptjk4U?t=2925) — [H2] AI elevates right people - Key idea: Curious, tenacious agents benefit most from AI amplification - Evidence: "If you've hired a curious person, that curiosity amplified by AI is boundless" ## 📚 References - [Sustainable Support Book](https://sustainablesupport.ca/) - [Help Scout](https://helpscout.com/) - [DocsBot.ai](https://docsbot.ai/) - [WP Product Talk Podcast](https://wpproducttalk.com/) ## 🗣️ Paraphrase - AI acts as an amplifier for existing support systems rather than a replacement, making good systems great and bad systems worse - Success requires establishing proper metrics focused on agent ownership and expertise before implementing AI tools - The most effective approach combines AI automation for simple tasks with human oversight for complex decisions and relationship management ## 📰 Newsletter **AI-Powered Support: Amplifying What You Already Have** The WordPress community is buzzing about AI in customer support, and for good reason. But before you rush to implement the latest AI chatbot, consider this insight from support expert Ben Meredith: AI will 4x whatever you already have. **The Reality Check:** If your support processes are broken, AI will make them 4x more broken. If they're excellent, AI will make them 4x better. The key is fixing your foundation first. **Practical Steps:** Start by measuring what your agents can actually control - their expertise, ownership of tickets, and customer outcomes. Focus on creating an "all y'all vs the problem" mentality where support agents unite customers, developers, hosts, and plugin authors against issues. **WordPress-Specific Benefits:** AI excels at handling the distributed nature of WordPress support, understanding complex plugin interactions, and providing context about customer purchase history and technical environments. Resources: [Sustainable Support Book](https://sustainablesupport.ca/), [Help Scout AI](https://helpscout.com/), [DocsBot](https://docsbot.ai/) ## ✅ Action Items #### WordPress - Core: Establish support metrics focused on ownership and expertise before AI implementation - Support: Create knowledge base training data for AI tools, measure reduction in simple question volume - Documentation: Audit gaps identified by AI analysis of support tickets for content opportunities - Community: Share AI support best practices and ethical guidelines across WordPress product teams #### OSPO - Policy & Governance: Develop AI support tool usage policies and data handling protocols - Security Management: Review AI service providers for data protection and privacy compliance - Developer Support: Create AI-assisted debugging workflows while maintaining human oversight - Community Engagement: Fund research on ethical AI implementation in open source support #### DevRel - Developer advocacy: Create content showing responsible AI integration in support workflows - Education: Develop workshops on training AI tools with proper context and boundaries - Developer experience: Implement AI-assisted documentation generation from support interactions - Community management: Use AI to identify recurring community pain points and solutions ## 🧪 Hypotheses - AI amplification effect: Whatever quality level exists in current support will be multiplied by 4x when AI is added - Context dependency: AI support quality correlates directly with the amount and quality of contextual data provided - Human oversight necessity: Complex WordPress ecosystem issues require human judgment that AI cannot reliably provide - Onboarding acceleration: AI can dramatically improve new support agent training and knowledge acquisition ### 📊 Methodology - Measure baseline support metrics (ownership, expertise, customer satisfaction) before AI implementation - A/B test AI-assisted vs human-only responses for different types of support requests - Track support ticket volume reduction from self-service AI tools over 6-month periods ### ❓ Questions - How can we measure the "ownership" and "expertise" metrics that Ben advocates for in a quantifiable way? - What level of context and training data is needed for AI to handle WordPress ecosystem complexity effectively? - How do we maintain authentic customer relationships while scaling support through AI automation? ### 🔮 Future Research - Develop standardized metrics for measuring AI amplification effects in support teams - Study long-term customer satisfaction impacts of AI-hybrid support models - Research optimal human-AI handoff points for complex technical issues ### 💡 Implications - Practice: Support teams need process optimization before AI implementation, not after - Policy: Need clear accountability frameworks when AI assists in customer-facing decisions - Education: Support agent roles will evolve to require AI prompt engineering and oversight skills - Funding: Investment in AI tools should be paired with investment in foundational support processes ### 🧠 Learning Styles Assessment Reading level estimate: 12th grade #### 👀 Visual learners - Present: Metaphorical diagrams of "all y'all vs the problem" concept - Improve: Flowcharts showing AI decision trees and human handoff points #### 🔊 Aural learners - Present: Full podcast format with conversational explanations - Improve: Audio timestamps for key concepts, voice emphasis on critical points #### 🗣️ Verbal learners - Present: Clear analogies (AI as amplifier, support as ambassador) - Improve: Glossary of technical support and AI terms #### 🧩 Kinesthetic learners - Present: Actionable implementation steps and metric frameworks - Improve: Interactive worksheets for measuring current support baseline #### 🧠 Logical learners - Present: Cause-effect relationships between AI implementation and outcomes - Improve: Decision trees for when to use AI vs human intervention #### 🧑‍🤝‍🧑 Social learners - Present: Conversational podcast format with multiple perspectives - Improve: Community discussion prompts and implementation case studies #### 🧘‍♀️ Solitary learners - Present: Detailed frameworks for individual assessment and planning - Improve: Self-audit checklists and reflection questions #### 🌿 Naturalistic learners - Present: Organic conversation flow and real-world examples - Improve: Ecosystem thinking about WordPress support community impact ### 📊 Accessibility Assessment #### Screen reader semantics - Present: Clear heading structure and descriptive timestamps - Improve: Alt text for any visual examples, ARIA labels for interactive elements #### Keyboard navigation - Present: Standard video controls and transcript availability - Improve: Skip links for major sections, keyboard shortcuts for playback #### Contrast and readability - Present: High contrast video player interface - Improve: Transcript with sufficient color contrast, large text options #### Text resizing and reflow - Present: Responsive transcript layout - Improve: Ensure 200% zoom compatibility without content clipping #### Media and captions - Present: Full transcript available - Improve: Synchronized captions with speaker identification #### Forms and interactions - Present: Standard video controls - Improve: Accessible comment submission forms ### 🌍 DEIB Assessment #### Cultural breadth - Present: US-focused WordPress business perspective - Improve: International support team examples, global customer considerations #### Inclusive language - Present: Gender-neutral terms, accessible technical explanations - Improve: Avoid regional idioms, provide cultural context for support approaches #### Neurodiversity support - Present: Structured conversation with clear topic transitions - Improve: Visual topic markers, consistent pacing indicators #### Representation - Present: Mix of business perspectives and technical expertise - Improve: Include diverse voices from different company sizes and markets #### Psychological safety - Present: Respectful discussion of AI concerns and job security - Improve: Explicit acknowledgment of AI anxiety, resources for adaptation ### ✅ Logical Validity #### Strong reasoning - Deductive strength: If AI amplifies existing systems, and broken systems exist, then AI will amplify broken systems [strong] - Inductive strength: Examples from Help Scout and DocsBot implementations support broader conclusions [moderate] - Causal reasoning: Clear links between proper metrics and improved outcomes [strong] - Analogical strength: AI as amplifier metaphor effectively explains multiplicative effects [strong] #### Fallacies - Hasty generalization: 4x improvement claim lacks statistical backing, but acknowledged as metaphorical - Appeal to authority: Reliance on Ben's expertise balanced with practical examples and data #### Tone of voice - Overall tone: Conversational and practical, balancing optimism about AI potential with realistic cautions about implementation challenges ### 🛡️ DARVO Analysis - Deny: No denial patterns present - Attack: No attacks on critics or skeptics - Reverse victim/offender: No victim/offender role reversals - Emotional cues and rhetoric: Balanced discussion acknowledging both AI benefits and legitimate concerns about job displacement ## 🎯 Conclusion AI can indeed 4x your support team's effectiveness, but only if you first establish solid foundations in metrics, processes, and culture. The technology works best as an amplifier of existing quality rather than a replacement for good judgment. WordPress product teams should focus on measuring ownership and expertise, providing comprehensive context to AI tools, and maintaining human oversight for complex ecosystem issues. Success comes from treating AI as a powerful assistant that enhances human capabilities rather than replaces them. ## 🧱 Refactor Appendix ### 📚 Literature Split Pad ### [H1] AI amplifies existing support systems > [!quote] [Using AI to 4x Your Support Team](https://youtu.be/mhLKmptjk4U) by [[Ben Meredith]] via [[WP Product Talk]]. Published 2024. Accessed 2024-12-29. > "Whatever currently exists on your support team will be amplified 4x by AI." Foundational principle for AI implementation strategy in customer support teams. [view highlight](https://youtu.be/mhLKmptjk4U?t=330#H1) #support-ai #scaling ID:rw_001 ### [H2] Ownership metrics enable support accountability > [!quote] [Using AI to 4x Your Support Team](https://youtu.be/mhLKmptjk4U) > "I want metrics that the agent themselves can control - their expertise, ownership." Measurement framework focuses on controllable agent behaviors rather than external factors. [view highlight](https://youtu.be/mhLKmptjk4U?t=525#H2) #metrics #accountability ID:rw_002 ### 🔖 Atomic Split Pad (Zettel Candidates) #### [Z1] Context quality determines --- ## Full Transcript [00:03] Hi, uh, everything's ai now AI and support's a particularly hot topic and [00:09] you know, you either think it's gonna replace everyone who does anything, [00:12] or you think AI slot might make your company feel soulless and horrible. [00:16] There's only two opinions to have, uh, unless you're Ben Meredith who's [00:20] joining us to tell us why that's wrong. [00:26] This is WP Product Talk, a place where every week we bring you insights, [00:32] product, marketing, business management, and growth, customer experience, [00:37] product development, and more. [00:40] It's your go-to podcast for WordPress product owners by WordPress product [00:44] owners, and now enjoy the show. [00:54] Hi, I am Ian Meisner. [00:56] And I am Zack Katz. [00:58] And, uh, you know, it's, before we get started, I do wanna [01:02] ask you, please do subscribe. [01:04] Uh, we did want to hit 10,000 subscribers and by asking everyone to subscribe [01:07] every week, uh, we did do that already. [01:10] So now it's like a million, I guess, is the right number now, maybe. [01:14] Yeah. [01:14] One, [01:14] um, [01:15] 1 billion subscribers we aim [01:17] for. [01:17] Yes. [01:18] E everyone who's ever used WordPress is the only right number. [01:22] It's that, or there's no point at all. [01:24] Uh, but if you do subscribe, you are helping us get this in front of more [01:27] people who, you know, product builders who might need this type of advice and help. [01:30] So thanks for clicking that button. [01:33] Yeah. [01:33] And a huge thanks to Freemius for, for being our exclusive growth partner on [01:38] that road to 10,000 subscribers and now 1 billion subscribers, uh, Freemius. [01:43] If you don't know, they help software makers grow, monetize, and succeed. [01:47] Uh, they're a merchant of record platform, which means they handle everything [01:50] around selling software, like checkout payments, subscription licensing [01:54] updates, all the stuff you don't really care about unless you wanna actually [01:58] sell products, uh, to your customers. [02:01] They also handle global tax compliance, what a pain in the neck. [02:04] So cool. [02:05] Uh, that allows you to focus on building amazing products. [02:08] Thank you so much for, Freemius, for being our sponsor. [02:11] Yeah, it's great. [02:12] And I, I am probably, Zack, you're probably the best ad [02:15] reader of any of us so far. [02:16] That's amazing. [02:17] Great. [02:18] Uh, so yeah, let's jump into how we use AI to, to make your support team. [02:22] I think exactly four times more productive is the commitment that Ben's already made. [02:27] Uh, and we've invited Ben Meredith to the show to talk to that. [02:30] Ben, tell everyone a bit about yourself. [02:32] Well, hello there. [02:34] I'm so excited. [02:35] Uh, you, you two are two of my favorite people, uh, and I'm just [02:39] really glad that you didn't invite Cromwell on this call because [02:42] Me too. [02:42] Yeah, that's a great, [02:43] really would bring it down. [02:44] Uh, really would take down the, the, uh, enjoyment, uh, immensely. [02:48] But no, my name's Ben Meredith. [02:50] For those who don't know me, I live in South Carolina with my wife and three [02:53] boys, and I've been in the trenches of technical support for like 20 years. [02:59] Um, and then 10, the last 10 of those in technical support, uh, specifically. [03:04] So before it was kind of customer service in general. [03:06] Now it's really technical support and, uh, product technical support for, [03:12] you know, those last 10 years as well. [03:13] So yeah, really excited to be here and to share. [03:19] So you said technical support and customer support? [03:22] Like what? [03:23] Customer service, I [03:25] guess customer service. [03:25] What are the, what's the difference? [03:27] What are, what are we, what, what's different? [03:29] Well, [03:29] technical support is, hey, this thing is broken. [03:32] Um, like the thing that you are supporting is broken. [03:35] Customer service is gonna be a whole lot more broad. [03:37] Um, and so that's why I can say 20 years of customer service is because I was doing [03:43] servicing of customers as in, you know, I need help setting up my cell phone. [03:49] I need help doing this other thing. [03:52] Whereas technical support gets a lot more into, or exclusively, I [03:56] guess, into the thing is broken and I need help debugging the thing. [04:01] Um, whereas, uh, in customer service more generally, uh, when I worked in retail [04:06] sales, for example, I was doing customer service, um, but it wasn't technical [04:11] support, so it, it, it was not just, you know, make this one product work. [04:18] And I asked you that in part so that you could answer it, but also, uh, [04:22] because I learned that distinction, uh, Ben and I have a call, he, he advises [04:27] me on, uh, support at Gravity Kit. [04:29] So I, I consult with Ben, uh, he is Sustainable Support is, uh, a great [04:34] book that Ben has written on how to do support sustainably, buy it, [04:38] uh, get it now and you'll be like halfway to, uh, well, you'll be 90% [04:43] of the way to doing better support. [04:45] Uh, the rest of the 10% is just, uh, implementing what he said. [04:48] You don't even have to read the book. [04:49] Just buy it. [04:50] Just buy it. [04:51] We'll be immediately 90% of the way there. [04:54] It's, [04:55] that's [04:55] right. [04:56] How the magic works. [04:57] But that's, that's my, uh, thesis. [04:59] Yeah, it's awesome. [05:00] Uh, and you know, thanks again Ben for coming on and for anyone watching [05:03] currently on YouTube, we do encourage you to leave comments, questions, or feedback. [05:08] Um, you know, if you're using AI and support or you want to and you hate [05:11] AI and you wanna throw your computer in a dumpster, it doesn't matter. [05:14] Just tell us about it and we'll highlight comments to discuss on the show. [05:18] Uh, so Ben, as far as the four times that we put on the title [05:22] why four times what, four times? [05:24] What is four times? [05:25] And obviously I'm being a little silly, it's not, the number itself is arbitrary, [05:30] I assume, but what is it exactly? [05:32] Four x I've done all the math. [05:35] Um, and it's going to be exactly four x. No, uh, you're exactly right. [05:39] The title, the title's a bit clickbaiting. [05:41] I'm not gonna lie to you because, um, you, you will four x your [05:47] support team with ai and because the reason that I say that is because. [05:56] Whatever currently exists on your support team will be amplified 4x by ai. [06:05] And so you can 4x a really good support experience for customers using ai. [06:13] And we can get into tactics and tools and tips and whatever, but if you take AI and [06:19] plug it into a system that's broken, or a system that is frustrating customers [06:25] or a system where your, uh, team feels disconnected, you're gonna 4x that, um, [06:31] with ai, uh, because AI is an amplifier. [06:35] And so we kind of, my premise behind this entire topic is it, it's not [06:44] about the ai, the AI's great, but it's about what you are four X-ing. [06:49] With ai. [06:50] Uh, and so yeah, if you've got a team that is disconnected, if you've [06:54] got a team that's not focused on the customer, if you've got a team that's [06:57] focused on maybe the wrong numbers like CSAT or time to first response [07:02] or ticket count, like hyper-focused on that kind of thing, if you've got a [07:07] team that's scared of losing their job or a team that's burned out or a team [07:11] that doesn't feel heard by the product team or the marketing team or whoever, [07:16] um, that's what AI's gonna amplify. [07:18] So that's, uh, that's my brief, um, spoiler, is that yeah, you're, you're [07:25] gonna absolutely going to 4x something when you plant AI into your system. [07:30] It's just whether or not you're four Xing something that's actually a benefit. [07:36] Yeah. [07:37] I mean, it sounds like there's, uh, a lot of opportunity and [07:39] pitfalls down that path then. [07:41] And, uh, maybe, um. [07:44] I guess sort of the way I would want to take that again, keeping in [07:46] mind, you know, WordPress product builders is sort of the people [07:49] that we're trying to help here. [07:51] Um, you know, if they want to experiment or if they're experimenting now, how do [07:56] they avoid four X-ing the wrong thing? [07:59] Right? [07:59] What's the, what's the steps is what, [08:01] right. [08:02] So I guess, I mean, there, it's easy in one sense, right? [08:08] Because I just told you, you're gonna 4x whatever's already there. [08:10] So if you've already got a problem. [08:13] Um, don't, don't just throw AI at it. [08:16] Fix the, the underlying root problem. [08:19] So if you've got, [08:20] oh, so, but hold on, hold on a second, Ben. [08:23] Fix underlying root problem. [08:24] What are you even talking about? [08:26] Because that's like, when you say, oh, just fix the problem. [08:29] Isn't that the whole thing though? [08:31] Like, [08:32] yep. [08:32] How do you identify what the problems are? [08:34] Uh, how do you fix what the problems are? [08:36] Let's, let's get into this. [08:37] Let's sure dive in. [08:38] Absolutely. [08:39] So I think the, the key thing is, well, first off, and this is, this is where [08:44] I'm glad Cromwell's not here because he would just like shake his head at me. [08:48] Like I told you that you're the, you're the reason, I'm the reason [08:51] that you are saying these things. [08:53] When I was in a director of technical support role, uh, at my [08:58] company, I was very gut driven. [09:01] Um, and by that I mean. [09:03] It felt like the right way to lead. [09:05] It felt like my team needed to do this thing. [09:08] Um, and so what I had was something that was imminently unscalable. [09:13] I had my own expertise and I didn't have any numbers to point at why [09:20] the way that I was doing it worked. [09:23] Um, and so I said about changing that, uh, with encouragement from Cromwell, um, to [09:30] say, Hey, I, you've gotta start tracking the right metrics and getting some, some [09:37] data behind this so that you can say, this is, uh, this is where we're going. [09:42] This is how we're gonna get there. [09:44] And so I, that's why I wrote the book. [09:48] It was to formalize this metric system in my head of. [09:53] What is it that I want to encourage my support agents to do? [09:57] Um, or how is it that the support agent should be interacting with? [10:03] Or how can I measure how the support team is interacting with marketing or [10:07] interacting with product or whatever I need to, I need to get a clear vision [10:12] of what their core job responsibility is and a feedback loop to be able [10:18] to say, yep, that's exactly right, or, Nope, that's not exactly right. [10:22] And then measure that over time to be able to say, Hey, what's [10:27] the, um, what are we aiming at? [10:30] And are we, are we hitting what we're aiming at? [10:32] And so that's where the metrics in the Sustainable Support book come from. [10:36] Um, and, and so it really is as simple as [10:41] deciding, uh, that support is going to have a, a, a complete seat at the table. [10:47] Uh, customer, the customer voice rather, is going to have a seat at the table. [10:51] Um, for product decisions or marketing decisions or, uh, whatever, and then [10:56] measuring how well we're doing that. [10:58] And you can measure from both sides. [11:00] If you're a, a WordPress product owner and you're not over the support team, [11:03] you're just in the product space, you can measure how well the information [11:08] is making the, the, the transition from the customer to the product and say, [11:13] Hey, this is really valuable, really well thought through information that [11:19] we can actually do something with. [11:21] Um, but if you're on the support side of things, it's the same deal. [11:24] You, there are specific things that you can track and say this, if we [11:29] keep doing more of this, the numbers are gonna go up and to the right. [11:33] If we keep doing more of this, numbers are going to go down. [11:37] We're gonna get, we're refunds are gonna go up. [11:40] Um, and so we started tracking that years ago now. [11:43] Um, and uh, really building a system that allows us to be able to, to stick [11:49] our finger in the air and go, yeah. [11:51] This is, it is working to get there. [11:54] And that's the problem that we solved before we stuffed AI on top of it. [11:58] Um, then when we stuffed AI on top of it, it was magic. [12:04] Yeah. [12:04] And so I guess to sort of get to the, the AI part, I almost want to like [12:09] establish a bit more of that baseline. [12:11] It's like there's, you, you're sort of like hovering around the idea of like the [12:14] empty chair test sort of thing, right? [12:15] So, you know, if the customer was sitting at the table with your team, [12:20] are they happy about the actions you're taking to automate and why? [12:22] And I think that the part that I'm sort of, you know, to be able to [12:26] understand more fully how to automate and like, you know, how to get to [12:28] the sort of like tactical details. [12:30] Um, can you share a bit more about like, are, are there specific metrics [12:33] you're thinking about or a framework for deciding what metrics are appropriate? [12:37] Like what are you measuring that makes you know that the things you're [12:41] accelerating are the right things? [12:43] Yeah, absolutely. [12:44] So I mean, there's the obvious ones, the obvious metrics, like five star [12:47] reviews and, uh, CSAT and those, those types of things where people will [12:52] willingly tell you, or either you ask 'em to or they just do it and they, [12:57] they just gush about your product. [12:59] Um, those are great, but what I focus on, uh, in my management of a support team is [13:08] I want metrics that the agent themselves can control, um, their, uh, expertise. [13:15] They, they can control that. [13:17] And so I have a way to measure an agent's expertise. [13:20] Ownership. [13:21] I, how, how much does the person own every ticket or own this individual ticket? [13:29] Um, in terms of the customer outcome in the WordPress space, it, [13:33] it's all the locus of control is distributed with the product, right? [13:40] Like, I am not in control as a support agent of really anything that's going on [13:48] in terms of like, it's not my website. [13:51] It's not, I, if, if I have, if I'm a, a support agent that has the mindset [13:56] that it's not my product either. [13:57] Either like. [13:59] And so the more a, a support agent can take that ownership [14:02] role and say, I'm going to control the outcome of this ticket. [14:07] Um, and I'm gonna assemble the team. [14:09] I talk about in, in the book, um, there's, there's disparate teams [14:15] in every support interaction. [14:16] There's me as a support agent. [14:18] There's the customer. [14:20] There's third party product, a third party product B, the theme, the host, [14:26] they, they're all on these separate teams. [14:29] And my job a good, a good support agent, their job is to arrange the teams. [14:35] And I use the southern phrase, all y'all to describe that. [14:39] And so all y'all is like a big hug, right? [14:42] All y'all are welcome. [14:43] You're welcome to come to the. [14:45] Uh, the, uh, picnic, the cookout, um, some of you might call it a [14:50] barbecue, but you'd be incorrect. [14:52] Barbecue is the meat. [14:53] Um, you, you can give, you can give a hug to everybody. [14:57] And so we want to arrange two teams. [15:00] The problem on one side and all y'all, all of those other things that [15:04] I mentioned is on the other side. [15:06] And so I'm bringing the customer alongside me and saying, Hey, [15:09] we're gonna solve this problem. [15:11] I'm gonna get your host to solve the problem that, that they're causing. [15:14] I'm gonna get this third party plugin to solve the problem that [15:16] their problem causing the theme, blah, blah, blah, blah, blah. [15:19] Yeah. [15:20] So when you're talking about all y'all, and you're talking [15:22] about us versus the problem. [15:25] Mm-hmm. [15:25] And, uh, when it comes to ai, so how, how the, the way that we've been [15:31] approaching this at Gravity Kit is, um. [15:35] Multifold. [15:35] Like we, we, uh, use Help Scout as our, uh, support desk and Help Scout [15:41] has an AI chat bot thing that, uh, that can draft responses, uh, with AI [15:46] drafts and you can technically train it with a few different external data [15:51] sources, but mostly it's also being trained on your support responses. [15:55] Mm-hmm. [15:56] Help scouts AI responses. [15:58] Not, not the best, but sometimes they can be good drafts that you can work off of. [16:04] Mm-hmm. [16:04] Um, they don't let you choose your model. [16:06] They don't let you use a custom API key that allows you to get [16:09] better quality out of them. [16:11] Mm-hmm. [16:11] Um, so we have Help Scout built into the, built into our help desk. [16:15] Then we also have Docs Bot, uh, docsbot.ai, uh, docs bot, uh, [16:22] allows you to be much more specific with your, uh, data training. [16:26] And you can add a bunch of different, uh, data sources. [16:29] You can give your own API key and say, or set the model actually [16:34] and say what kind of model you want, like GPT 5.2 or whatever. [16:37] Right. [16:38] Um, and then we also, so those are fruit self-help, uh, for, so the [16:43] customers can go to our support page and chat with our docbot bot. [16:47] They can go to our widget that's also built into our products, and they [16:51] can chat with the Help Scout Beacon. [16:53] Mm-hmm. [16:53] Uh, those are fine. [16:54] And actually the doc spot and the Help Scout Beacon have reduced [16:57] our support load by 40%, maybe. [17:00] Uh, uh, we're like hovering around 30, 40%, which is incredible. [17:04] A lot of easy questions are answered, but when it comes to coordinating the [17:08] all y'alls, there are a lot of different questions that are, have to be asked, [17:13] parsed and understood and processed, and there's a lot of code involved. [17:18] There's a lot of confi, like there's a lot of, uh, server, like you [17:21] mentioned, Ben server variables. [17:23] Mm-hmm. [17:23] There's a lot of information that goes into answering a support question. [17:27] Um, and all y'all becomes a much tougher problem when you [17:30] have limited context and ai. [17:32] Mm-hmm. [17:33] Uh, you have hosted solutions like Help Scout that may or may not be able [17:37] to give you, uh, that may, may or may not be able to ingest the additional [17:40] context that you wanna give it. [17:42] Right. [17:42] Um, so one of the, the third things that we're trying to do is build an automated [17:46] system that when a ticket comes into Help Scout, uh, the support person has a [17:53] Chrome extension open on their sidebar, it als, it then pings our own service [18:00] that answers the question in the way that we think it should be answered. [18:03] That combines a lot more, uh, data sources, a lot more information, [18:07] a lot more context, including their, like, easy digital downloads [18:10] purchases, and their refund history. [18:11] And like provides a lot more context. [18:14] And then it also, we'll read our documentation. [18:17] Identify does this. [18:19] A ticket, ask a question that hasn't been answered in our documentation, should [18:24] it, uh, should it update an existing doc? [18:28] Mm-hmm. [18:28] Or create a new doc. [18:30] Right. [18:30] 'cause those, uh, should it be, or should it be just an internal [18:33] doc that is about process? [18:35] Are we, do we have a process for the answering this, this type of ticket? [18:40] Mm-hmm. [18:41] If we don't, then how do we create it? [18:41] There are endless opportunities to build automations and stuff like that. [18:47] So like, uh, when you say all y'all, it means a lot to me. [18:53] Uh, right. [18:54] Like y'all is a big y'all when it comes to, uh, ai. [18:58] Yeah. [18:59] I, and I think that's, that's where the, the multiplier happens, is that you [19:05] can give AI context and the more, and you know, we're seeing it all over the [19:10] place on our social feeds this week with Claude Bot or whatever it's called now. [19:16] Open [19:16] Open Claude or [19:18] Open Claude, yeah. [19:19] Is what they, they landed on. [19:20] Um, the more context you give ai, the better results it's going to be. [19:27] Um, it's going to produce rather. [19:29] And so yeah, there, there's your next frontier in, uh, and it sounds [19:34] like you, you're doing more than we were doing, um, at my, uh, at Stellar [19:40] WP in terms of con connecting and giving as much context as possible. [19:46] But the support team's role is not replaced by ai. [19:51] It is enhanced by AI because there's, I don't know how many years off we are [19:57] from a, a system that reliably could say. [20:05] Could answer that question you just said. [20:06] Have, have, have we ever gotten this question before? [20:10] And do we have a, a, um, a framework for even answering this question? [20:15] If yes, do this, if no do that, like that type of multi-step thing. [20:23] I, we're certainly not there right now. [20:25] Um, follow up questions to ai. [20:29] In my experience and everybody who's ever used a chat bot, um, you [20:33] recognize how bad your, your question was by how bad the answer was. [20:38] Um, and uh, that Chrome extension sounds amazing because that's [20:43] gonna be able to actually give me some context as a support agent to [20:48] to hop in there and say, um, this is what operating system they're on, [20:53] this is what, uh, the, when they last purchased the last time they reached out. [20:59] It's gonna give a whole heck of a lot more context. [21:01] Um, but it's, that's where we've gotta get to is feeding, getting the ai, [21:07] the input into AI to be good enough to actually answer the question. [21:12] And right now, you still need a human to answer much of those questions. [21:19] Yeah. [21:20] Yeah. [21:21] I, I, I, I have sort of two thoughts to that. [21:25] So number one, I think you said something really meaningful there of, you know, [21:28] the, um, you, you understand how bad the question was by how bad the answer is. [21:32] I think that was always true. [21:33] I don't know if that's even an AI thing. [21:36] In our support documentation, we have a very strong, like three stages of [21:40] a conversation, content concept that anyone that has ever communicated [21:45] with one of our teams, like. [21:46] We basically refuse to help you on the first reply, we we're literally [21:51] just info diving until we can actually restate the question with intention and [21:55] we know the exact question being asked. [21:58] Um, and we're pretty strict about that. [22:00] Now we do a lot of similar things to what Zack's describing [22:03] internally as well though. [22:04] And that sort of like three stages thing has served us incredibly [22:08] well in using AI to present information to people to work with. [22:13] Because essentially if you write, ask the right question, the prompt, we, we [22:16] basically wrote a process for writing great prompts five years ago, basically. [22:21] Mm-hmm. [22:21] And it, it's working amazing. [22:22] Um, now that being said, there is that sort of the question of like the human [22:26] ai hybrid side of things of like what is the appropriate, you know, if you [22:31] could automate it and it was able to actually present something as good as [22:35] what the person can do, should you. [22:37] You know? [22:38] Mm-hmm. [22:39] Well, here's, here's a example of, uh, further rabbit hole. [22:44] Like one of the things that humans do is try to reproduce the problem, [22:48] right? [22:48] That's one of the things that ais are starting to be able to do reliably with [22:53] things like, uh, the chrome extensions that Claude Code and Claude have. [22:56] Um, there's, uh, there's also playwright tests that allow, [23:02] uh, interacting with a website. [23:03] So one of the, one of the tools that I've been working on building [23:06] is, okay, identify the steps to reproduce this, and if there are gaps, [23:10] then we need to clarify that gap. [23:12] I think it's hilarious, Ian, that a very kestrel of you to say you refuse [23:16] to ask the question until it's clear, uh, or to answer the question rather. [23:21] Um, I, my approach to Gravity kit is always answer the question that we [23:25] think it is and ask follow up clarifying questions in case we got it wrong, [23:30] right? [23:30] But, but the, like, okay, so we have AI guessing at what the answer is, [23:37] just like we might do without asking the important clarifying questions. [23:42] Um, but while it's guessing, it can also reproduce, uh, to the best of its [23:46] ability, the setup that it thinks the customer has and see if it is an issue. [23:51] It could read the code base and identify if the code might have [23:54] a bug based on the description. [23:56] So there are so many different things that a human support agent would do [23:59] if they were given enough time with a ticket at every stage of the ticket. [24:04] Uh, that AI is going to be able to, and currently is able [24:07] to do with the correct setup. [24:09] Um, so I, I god thousand percent agree that humans are [24:14] important in this, uh, process. [24:16] And it should not just be an AI driven thing. [24:19] And I think the human touch is going to be, uh, less about reproducing the issue [24:26] and more about kind of handholding and guiding and being the, the, uh, the person [24:32] who's along the ride with the customer while the solution is being implemented. [24:37] And that includes having, I know people already, uh, who are running AI agents [24:43] that automatically fix a bug based on a support ticket coming in, do a pull [24:47] request, send the zip to the customer. [24:50] Automatically without even a human being involved. [24:53] That is pretty cool. [24:55] But it's also removing humans in a way that I'm not sure [24:59] is, uh, the right approach. [25:00] But I think it's, I, I sure would love to have a, a zip file [25:04] that I could send to people. [25:05] Mm-hmm. [25:06] Um, solving their problems faster and for Xing our support team. [25:12] Nice. [25:14] That, that makes me a tiny bit nervous of like, if you [25:17] just, AI is imminently helpful. [25:20] It says yes by default, essentially. [25:23] Right. [25:23] Which is very good in a support context and we should encourage that. [25:27] But also there is that, you know, sort of the thing that you said earlier [25:31] about, of like coordinating the teams and having a seat at the table, like [25:34] the friction between product and support, I think is a good thing in a [25:38] lot if, if done well, you know, there is times where maybe you shouldn't just. [25:43] Write something and send it in is zip file and you know, somebody needs to sort of, [25:48] well, that's the other thing. [25:49] Find that, [25:49] so, so much of our support points out user experience, user interface, um, [25:55] conceptual issues with our products that, that need us to take a step back [26:00] and say, does this need to be rethought at a fundamental level, the way that [26:05] we handle this specific functionality. [26:07] And I just had a call with Ben and, uh, our, uh, our support manager who [26:11] said, you know, we were talking about a specific short code that had like, merge [26:15] tags from gravity forms that weren't per performing went in Nevada framework. [26:19] Like, it, it was a very spec. [26:21] And we were like, okay, what's the bigger actual issue here and how [26:25] can we solve the, the, the, the core fundamental problem that caused the bug? [26:31] And I do think that AI is pretty good at that too. [26:36] Mm-hmm. [26:37] Where each ticket that comes in, one of the questions that I, in this miracle [26:42] AI system that I'm trying to build that will probably never actually work fully, [26:48] uh, is like, okay, does this require a rethinking of how the product works? [26:53] And if so, what would be better about it? [26:55] And if so, maybe you can make a branch and make, see, see if it works well. [26:59] Right. [27:01] Yeah. [27:01] Because. [27:03] I think the main point I want to drive at is whatever we do with [27:10] ai, obviously AI's happening at lightning speed right now, right? [27:15] It is everywhere and it's ev there's this huge rush to be the first to or whatever. [27:22] Um, what you just described to me where somebody reports a bug, it [27:27] patches the bug, it zips it up, and it emails them back with the zip [27:32] sounds to me like technical debt looking for a place to happen, right? [27:38] It sounds to me like. [27:39] Oh yeah, that's awesome. [27:42] But if there was no human involved, um, at all and that whole, [27:46] I think they draft it, I think they draft it and then, then [27:49] it, they, uh, human sends it. [27:51] That's, uh, [27:51] yeah. [27:51] And so super powerful and obviously I, I hope that we can get to a [27:56] place where, um, the, the actual debugging can be minimized. [28:01] 'cause I've been there, I've been in a support ticket for 45 minutes trying [28:06] to get something to replicate, like, you know, and so if I can do that [28:11] with a button push, um, and then guide my little ai, I call them AI idiots. [28:15] Whenever I'm working with AI chatbots, it's like, if I can guide my little [28:19] AI idiot to go over here and try this other thing while I try this other [28:23] thing, like, yeah, that's a, a force multiplier, but what you're multiplying [28:27] is my expertise because mm-hmm. [28:29] We've all been there, and I don't see this happen changing anytime soon [28:34] is AI is gonna make just absolutely ridiculous assumptions about what I [28:40] meant when I said, go debug this issue and you'll come back if, if you leave [28:44] an autonomous agent to just go do his thing or her thing, and you come [28:49] back and go, uh, what are you doing? [28:51] Like, no, the, the expertise is still needed and it's still the, the [28:56] human going through the process and knowing what to point out or not point [29:03] out or what, you know, how, how to direct this whole process is great. [29:08] But I, I, again, I think it comes back to having that philosophical [29:13] bent that the framework of this is what the, the point of support is to [29:21] give the customer a seat at the table. [29:24] That is valid in a way that is actually useful to the team because you can't [29:30] let customers come sit at your product meetings because you'll spend 45 minutes [29:37] out of the hour just explaining to them why what they want is not great [29:43] for the rest of the company or the rest of the customers or whatever. [29:47] And so the support agent is the translator, uh, I call it ambassador. [29:53] You know, you, you're, you're the one that is translating to and from [29:58] the product team or the marketing team or whatever the, the reasons [30:03] why this is going this way or the reasons why this is going that way. [30:07] And that's why once you get that in there, then you can start having [30:12] these agents that are AI agents [30:16] to do portions of the work, uh, whether it's the debugging or, or even [30:20] the, the answering of the questions. [30:23] Although I, I've not found a support. [30:25] I, I've tried to train, um, I think I was using, uh, anthropic, uh, [30:31] tried to train their API to, to do something in my tone, and I used [30:36] my tone guide and I put it all in. [30:37] You know, I, I tried as hard as I can to train it and then I ask [30:41] it like five or six questions and it sounds nothing like me yet. [30:44] And so I, it that it doesn't get it. [30:48] Um, and so I think we gotta feed it more data before it's gonna get there, and [30:52] that's where tracking the right metrics. [30:54] Um, and I think you can do it with AI too. [30:57] I think you can say, um, that answer was not ownership. [31:02] That's a, a one, it's a three point scale for those that haven't read the book. [31:06] One is, keep doing this and I'm going to fire you. [31:10] Three is keep doing this and I'm going to promote you. [31:13] And two is. [31:15] You're fine. [31:15] It's great. [31:17] Good enough. [31:18] And so really makes the measuring of each individual metric. [31:22] And we don't combine all the metrics into one big lump, 60% or [31:27] whatever, because that's not helpful. [31:29] No one likes getting a 60. [31:31] Um, and a 60 would be a really good score. [31:34] If it's a three here and a one here, well, there's something for you to work on, but [31:38] you're doing a promotable job over here. [31:41] So anyways, we keep the metrics separate and we measure them separately. [31:44] Um, and you can teach AI that I've not gotten great at it yet. [31:48] Um, I've trying to set up N eight n routine, um, to do a [31:54] feedback loop on support tickets. [31:57] And it hasn't, it hasn't really gotten it yet. [32:00] Um, but I'm. [32:02] I'm still gonna work on it, but I'm in, no, I have no sense that AI is [32:06] about to take a manager job yet. [32:09] Um, or a supervisor job. [32:11] Yeah. [32:11] ' cause it just can't read the, the context there. [32:15] Yeah. [32:15] And I actually do have, so I have a little bit of experience with that part [32:20] that I've been playing with over the last couple of months, which is, you know, [32:22] extracting the key root cause from every conversation when it's done, like post [32:26] support and like, you know, getting the job to be done and segmenting those. [32:30] And I actually worry that AI is better at that than I am unfortunately. [32:34] Right. [32:34] Which is, you know, it actually like, takes all the context clues and, um, [32:39] you, you can give Claude the entire zip file of one of your plugins. [32:42] So it's literally aware of all of those pieces. [32:46] Um, so that in particular I do worry about a little bit and actually something I'm [32:50] just gonna maybe throw out there as a potential issue here, Ben, is if you are [32:55] properly quantifying the thing to train a person how to do better next time. [33:01] If we don't have the tools yet, I think we have the tools soon that those [33:05] exact measures and systems that you're proposing are what we need to teach the [33:09] machine to do the job better, right? [33:12] Mm-hmm. [33:12] And, um, I don't know what, I guess for context, I, I shared this in [33:17] episode two or three episodes ago. [33:19] I was a big AI skeptic up until like December. [33:21] Mm-hmm. [33:22] And then I don't know what happened with Claude skills in particular, but now [33:25] it's just like, I'm, I got three Claudes running currently while we're talking. [33:29] Right, right. [33:30] So, um, I've gone from like zero to a hundred here and I do think [33:34] that like, you know, the things that I am able to do today felt [33:37] like sci-fi to me three months ago. [33:41] Is there, I guess to turn this into a question, is maybe your book not actually [33:46] the roadmap to creating the machine that does it better than the people can. [33:52] Well, there you go. [33:53] Did, is your book worth anything? [33:55] No. [33:57] It, it's a fair question. [33:58] It really is because it's, um. [34:01] The, the roadmap is still the roadmap because the people buying [34:06] are still people and the people, uh, that are renewing are still people. [34:12] Um, I don't think we've got much in the way of automated purchasing happening, [34:22] going on at, at a, at a wide scale. [34:24] So I still have to connect with people. [34:27] I still have to, um, have the skill of doing, doing that [34:34] interpersonal interaction. [34:36] And that's really what built Sustainable Support is. [34:39] Um, or Sustainable Support is build, Sustainable Support is the website. [34:43] But, uh, Sustainable Support is, is that idea that the more we treat [34:52] our customers with humanity and the more we treat, uh, and it goes [34:58] both ways, that ambassador concept. [35:00] The better the whole, it, it's the rising tide raises all boats. [35:06] So I, I don't think that's going anywhere. [35:08] I think that it's going to be amplified and we'll see hundreds if [35:13] not thousands of companies go under because they'll amplify what was [35:17] already there, which was a profit first. [35:21] Uh, people be damned attitude. [35:25] Um, and that's gonna show through. [35:28] Um, and so, no, I I think that, um, customer service, uh, a personal [35:32] connection and customer service, uh, is, is going to win the day. [35:37] I'm, I'm not in any way worried that we're gonna automate out the human aspect of it. [35:42] We're gonna free up support agents to spend the time to do the. [35:47] Humanity driving portions of the customer support job, because [35:52] that's gonna be the last thing that AI's going to be able to do. [35:55] I don't think it's gonna get there. [35:57] I'm not, call me a skeptical, [35:59] here's the thing. [35:59] Whatever. [36:00] Go ahead. [36:00] I think the, anybody who's listening, uh, who has been bad at support historically, [36:07] uh, I, I'm just, I don't reply well. [36:09] I just, like, I, I try to get it over with because that's not what I wanna do. [36:12] I'm a developer or I'm a marketer and this isn't worth my time. [36:16] What you can just have, say, Hey, AI, do a better job at support than I do. [36:22] And it will, right? [36:23] It will come up with a guideline. [36:25] Maybe not as good as Sustainable Support, but it'll come up with something, right. [36:28] That provides better support experience than you've been providing. [36:32] I believe AI will unlock better customer support across the board in every metric. [36:39] One of the things that it's allowing our support team to do. [36:42] Is to go deeper on, on providing solutions for our customers on debugging [36:46] and fixing right away on, uh, coming up with custom snippets if mm-hmm. [36:51] Like our customers are not gonna be as savvy as we are when it comes to using ai. [36:55] So we can use AI to provide better support for, to our customers. [36:59] Absolutely. [36:59] Uh, and provide them with a custom code. [37:01] But I think that the, the key is AI has unlimited time, unlimited [37:06] resources to dig into something in a way that we as support people who [37:10] have a t uh, you know, inbox full of tickets are not going to be able to do. [37:15] And so an example, uh, Kim come, comes up in my head like. [37:19] There's a management thing, a technique called five why's. [37:24] You ask why something happens, then you ask why that happened, [37:28] and they ask why that happened. [37:29] That's just a prompt that you can give AI that like, okay, ai, here's this ticket. [37:36] Five whys it. [37:37] Like, why did the customer get in touch with us? [37:40] Um, like, why wasn't this prevented earlier? [37:43] Why wasn't it, uh, why did the UI or the U user experience even lead to this [37:48] being a problem in the first place? [37:49] Why, why didn't we solve this better? [37:52] How could we solve it better? [37:53] Like, AI can do all of that for every single ticket. [37:58] The question is synthesizing. [38:00] And I think that that's gonna be the, the biggest problem that [38:02] we solve, uh, is finding a way to bring it all back together. [38:07] Right. [38:07] And I think that the biggest thing that AI is going to do is [38:10] not transform your support team. [38:12] It's transform your entire business. [38:15] And support will be part of that. [38:17] Because when AI is answering all these questions and addressing so [38:21] many things, then that's like a, just a, a seismic shift in everything. [38:25] And I just wanna maybe agree while slightly disagreeing with one part, Zack. [38:30] Okay. [38:30] So the agreement is, I think it's actually gonna be better at synthesis. [38:33] It knows everything. [38:34] You forget. [38:34] It doesn't, I think it's taste is the [38:37] Yeah. [38:38] But it creates so much output, uh, that, and the output isn't [38:41] necessarily aligned with the, the thing that needs to be synthesized. [38:44] The, the process is synthes of, of the synthesis. [38:47] And I'm, let's add some more s's to that word. [38:50] Right. [38:51] The, the process itself is, um. [38:54] The something we need to refine and get better at. [38:56] Um, [38:56] yeah. [38:57] And well, and I want to tie that to maybe Ben's ownership point a lot. [39:00] 'cause I, I had very similar measures with, uh, back at the Sky [39:03] Verge support team specifically. [39:04] We had a large team and we had a lot of this, like how do we make [39:07] sure it's better tomorrow than it was yesterday type of process. [39:10] That sounds like Ben, you're sort of an advocate for a lot of the same, you know, [39:13] different ways to get to similar outcomes. [39:15] Mm-hmm. [39:16] Essentially, um, that, that idea of ownership though is, you know, [39:19] the whole WordPress has this distributed problem that I think [39:24] you sort of alluded to earlier of like, what is the host's problem? [39:27] Who's plugin is it? [39:28] Whatever. [39:28] And that is sort of the thing that's like unique to what we all do here. [39:31] And with WordPress support in particular that like a SaaS product [39:34] doesn't experience and that like pulling all those pieces together [39:39] and knowing everything about all of those things, I actually worry that [39:42] AI is actually better suited than [39:44] people at doing. [39:46] And I think that that's why the magic of that ownership though, is that you get [39:51] people that they care enough over time that they learn, you know, what types [39:54] of customers are the ones that are the best fit for the product and like the, [39:58] the, the, the best support people become like junior product people automatically. [40:03] Mm-hmm. [40:03] Essentially. [40:04] Mm-hmm. [40:04] And I think that there's sort of this like if everyone goes all in on AI right now, [40:10] what is gonna replace that specifically? [40:13] And you know, like the, the flow of knowledge into more and [40:17] more people to become these like broad experts of the ecosystem. [40:22] I don't have a question so much as maybe that's the, [40:24] I concern with it. [40:25] I wrote a quote from 1979 IBM training manual that says a computer can never be [40:32] held accountable, therefore a computer must never make a management decision. [40:38] And I think that that is a kind of a way that we can look at our [40:42] support process and our development process and our management process [40:47] as we do this AI transition, which is we still need humans in the mix. [40:51] Mm-hmm. [40:52] Because [40:53] AI is not accountable to anybody. [40:55] It doesn't, it's just trying to do whatever whatever's trying to do. [40:59] Right. [41:00] It's overly agreeable and it's all of the things that we've already talked about. [41:03] Yeah. [41:03] Yeah. [41:04] It's not going to replace leadership. [41:08] It might replace tactics and execution, but AI's not going to replace [41:14] leadership within an organization. [41:17] Um, and it's going to, the, the right agents are elevated into [41:22] the right, human agents are elevated into leadership thought. [41:28] You have to think on the next level up to manage an AI bot and to spot [41:34] when the AI bot gave just a bonkers answer and step into a ticket and say. [41:40] Uh, sorry, customer. [41:41] That was a bonkers answer. [41:42] Here's the correct answer. [41:44] Um, and, and why that was a bonkers answer. [41:47] Those types of things. [41:48] And so it kind of makes these manager level positions more valuable. [41:55] Um, and it it, the right agent who already was looking to support, you [42:00] know, people come into support to get their foot in the door to go [42:03] to somewhere else in the company. [42:05] Like you said, they're junior product managers, not just that [42:07] they're junior marketers or junior docs writers or junior whatever. [42:13] They're, they're all looking to go somewhere, um, within the company because [42:17] we don't pay support engineers enough. [42:19] Um, and so this is a, gives them a real clear way to demonstrate that [42:25] ability to multiply their own skillset. [42:31] Ben, you said something, uh, that, that I, I'm gonna get [42:35] like a little sci-fi right now. [42:36] Like, you know, you're, si you said that things are already sci-fi. [42:40] Well, I'm gonna take it at next, like five years from now, uh, if we have an AI [42:45] powered virtual support manager who's a better leader than I am, uh, better at, [42:52] at, at reviewing the reports that are like of, of the analysis of each one of the [42:56] customer response tickets, I personally can see being inspired by an AI generated [43:04] avatar. [43:05] Uh, just, I'm just putting that out there because I know that, like, for leadership [43:10] to be real, I need to show up every day, uh, ideally and communicate the company's [43:16] values and reinforce why we're doing what we're doing and how we do what we do. [43:20] And just reminding like people that, okay, when this happens, [43:24] this is what we should be doing. [43:25] A lot of this stuff is just reinforcement learning. [43:28] Mm-hmm. [43:29] And AI is really good at reinforcement and, uh, it was, it, I don't see why we [43:36] couldn't eventually be in a world where, where AI has really good managers and [43:41] is in charge of a fully AI support team. [43:44] Mm-hmm. [43:44] I don't think that's going to be how we're, I don't intend for [43:48] Gravity Kit to be doing that. [43:49] I want humans to be involved and I do think we need to [43:54] think about how we address. [43:58] Our relationship with self-improvement when something can be so self like, [44:03] so constantly on like Ian's five, Claude, uh, you know, chats that are [44:08] currently happening, uh, when, when AI can continually improve, uh, how can we [44:15] change our relationship with it mm-hmm. [44:17] In a way that helps everybody. [44:19] Yep. [44:20] And I've seen the other day that like I was on LinkedIn or somewhere [44:24] and I saw a job, uh, posting, which thankfully not looking for a job right [44:28] now, but I saw this job posting and it said we're a people first company. [44:32] And I thought weird. [44:35] Um, like we, uh, what other kind of companies are there? [44:40] And I know what they meant and I know why they were saying it the way [44:44] they were saying it is because with all these layoffs in tech and all [44:47] this stuff that's happening in tech. [44:50] People have really started to legitimately fear that, hey, none of [44:55] these companies here are looking out for people, um, and, uh, going that way. [45:01] And so what it made me think of is as we're building these AI tools, it, [45:06] like you just said, to, to get not just sci-fi but almost theological, [45:12] like we, we need to understand that a different why to an agent. [45:19] Produces a different output. [45:21] And, you know, the whole start with why movement from years ago where [45:24] we, we, we read that book and we all said, we gotta start with our why. [45:27] I, I think that's still applies. [45:30] It's just we don't know how to translate it to the AI yet. [45:32] I think, and I do wanna just throw in one thing that you do hit a point [45:35] that Ben, you, I've been recommending this book to a lot of people. [45:38] It's called Spiritual Machines. [45:39] It's by Ray Kile. [45:41] He put it out in 1999. [45:43] Uh, and it is basically about this, like mm-hmm. [45:45] When the simulation is convincingly human enough, does it stop mattering is [45:51] the question he wrestles with in 1999. [45:53] And it's, you know, he's one of the guys that's invented all [45:56] this stuff, so I do recommend it. [45:58] Uh, he doesn't answer it in the book, so, you know, spoiler alert, but, [46:01] uh, he explores the question term [46:03] philosophy. [46:04] Yeah. [46:05] You're really not selling the book right now. [46:07] He explores the, he explores the question with a level of expertise [46:11] at a time when it was still. [46:13] Like sci-fi and kind of silly and stupid, but it's like real now. [46:18] And so the question is meaningful. [46:20] So I'd encourage you to check it out. [46:21] I'm gonna pull up Scott's question here though. [46:23] Um, so Ben, I'm gonna pose the question to you. [46:27] In this case, the best support people can help the customer that way, [46:30] like Ben did when he was in support. [46:32] And he is referring to an earlier comment about leading to retention and expansion [46:36] and making the customer like, love the company or the brand or whatever it is. [46:40] And will AI eliminate the lower level support people or help them [46:43] rise to those higher skillsets? [46:47] I think higher skillset is the wrong framing of the, [46:52] the, the different skillset. [46:56] Um, it's more of, I'll answer the question in a a, a roundabout way. [47:01] So when I'm hiring a support agent, what I'm looking for can't be. [47:08] Product knowledge because they don't know our product yet. [47:12] Um, or maybe they don't even know WordPress yet, right? [47:16] I, I, if, if all I'm looking for is product knowledge, you're [47:20] gonna hire a developer if you can. [47:22] Um, instead of a support agent, what I'm looking for in a [47:25] support agent is curiosity. [47:28] Curiosity, tenacity. [47:31] You know, I'm not gonna give up on a problem just because it got hard. [47:36] Those types of character quality issues, um, are what I'm looking for. [47:41] And so the short answer is yes, it will absolutely help the right. [47:46] Um, the, the automation automating out the, the low level easy question [47:52] stuff is going to, to cause the. [47:56] Agents, the right agents to immediately be able to get to a different spot. [48:01] But they're going, like I said at the beginning, AI amplifies whatever's there. [48:05] And if you've hired a curious person, um, that curiosity, man, it's boundless. [48:11] I, I fell down a rabbit hole the other day with, uh, cursor [48:16] trying to set up something. [48:18] No, it was, it, it wasn't cursor, it was core. [48:20] Uh, which is this new thing that I was playing around with that I [48:24] saw Chris limo mention somewhere and was like, Hey, I'll go try. [48:27] And I fell down this rabbit hole and it was like, man, that kind of curiosity of [48:33] like, I, I don't need to sleep right now. [48:36] I'm going to go figure out this thing. [48:38] Um, that's what you're looking for. [48:40] And then if you amplify that. [48:43] Then Yeah, it's gonna, cause it is not going to, to get rid of, well, [48:47] it'll get rid of people that you probably would've eventually gotten [48:51] rid of anyways, um, on a support team because they don't have the curiosity. [48:56] But AI is going to, uh, level up those people that need to be leveled up. [49:04] Yeah. [49:04] And I do wanna, I'm gonna qualify this because it's gonna sound [49:08] like I'm like debating basically. [49:10] 'cause uh, but the qualification, as I agree fully, the hiring process [49:13] I've always done for like any role is the concept of will versus skill. [49:17] I care about more than anything. [49:19] Uh, hi, will I, I don't motivate people. [49:22] I can teach people things though. [49:24] So if they're motivated and they're excited, they want to get better, great. [49:27] You bring that, you know, I don't care if you know how to do a [49:30] thing, uh, that's, I will defend. [49:32] That's important anytime. [49:33] Yeah. [49:34] Right. [49:34] Now my debate to this though is going to be, uh, I'm just [49:38] gonna use it as an example. [49:39] When I joined Sky Verge many, many years ago, uh, I was scared and [49:43] nervous and I didn't know anything. [49:44] And I'd used WordPress, but also I worked for, for Becca, [49:47] Becca Rice, if anyone knows. [49:49] And you know, she scared me at the time to be very honest. [49:52] And ultimately, uh, I think. [49:55] Had I not, you know, had all of the easy stuff to chip away at and to learn and to [50:00] really dive in deep and to to learn a lot more, I'm not sure I would've made it. [50:04] You know what I mean? [50:05] It's really a matter of, [50:07] so here's the thing. [50:08] AI is the best onboarding experience in the world. [50:12] You wanna know about our products. [50:14] You wanna know about how we do things? [50:16] Well, guess what? [50:17] AI can tell you if it's trained on our Slack history, if it's [50:21] trained on our support, responsive. [50:23] If it's trained on our handbook. [50:25] One of the things that AI is pointing out to me is how many things we [50:30] haven't documented in terms of our internal processes, how we do things. [50:34] So like to have a su, a successful su new support person come on and [50:40] to be able to provide good support. [50:42] They need initiative and they need an AI agent that can answer the questions [50:46] for them before they hit our actual tier two support agent, which is our company [50:51] Slack, where they can ask a human. [50:53] I think it's just like internal support. [50:56] Uh, and I, I, I think that AI is, again, like Ian, if you had AI 15 years ago [51:03] or whatever, you wouldn't have had to talk to Becca and Becca's great. [51:07] She's, she's scary 'cause she's so amazing. [51:08] That's the problem with her. [51:10] And [51:10] to be, yeah. [51:11] I wanna maybe qualify. [51:12] She was scary because I knew she knew Scary, [51:14] talented [51:15] is is the way [51:16] Yes. [51:16] Yeah. [51:17] Yeah. [51:17] And also, [51:17] so like instead of having to any, any of your managers, you'd be [51:20] able to talk to your chat bot and figure out how to use things. [51:23] And we have connected doc bot with our Slack so that if people have a [51:28] question, they just click the doc spot icon in Slack and they ask the [51:32] doc spot it, it, it helps, right? [51:35] It's good. [51:36] It works. [51:39] I hope you're right. [51:41] I guess that's that. [51:43] I mean, onboarding, how many things have, have each of us done where we point [51:47] AI to a code base we're not familiar with, and it is like, I don't know, just [51:51] like make it work and it makes it work. [51:53] It unlocks our ability to, to, to do what we need to do. [51:57] Done. [51:58] I think the, the thing that AI. [52:01] Uh, will four x your support team on is if your support team [52:04] knows what questions to ask. [52:05] Mm-hmm. [52:06] And that brings us back to what Ben was saying was like, ai, [52:08] ai, ai can go off the rails. [52:10] Yeah, it can. [52:11] But that's where we need a human to guide what the AI is going to [52:15] be talking to your customer about. [52:17] So like, and I also agree, Ben, it doesn't do a good job with getting voice. [52:22] Right. [52:23] And I get an AI response, then I, then I rewrite it essentially, or I [52:27] severely edit it to make sure that it's saying exactly what I want [52:30] instead of what it came back with. [52:33] Yeah. [52:33] It's gotten better, uh, at mimicking my voice, but it it maybe that [52:40] when it comes to the training on it. [52:42] Yeah. [52:42] Yeah. [52:43] Old school in me, I like writing that. [52:45] That's the part that I enjoy doing is the communicating, uh, [52:48] AI for me, the accelerator is. [52:52] What you just said. [52:53] I have a local install with the three plugins that I need on it. [52:58] I unleash cursor in there and say, why is this thing fatal? [53:02] Erroring this in this way? [53:04] And it tells me why. [53:06] And then I take that and I write up, uh, a good reply to a customer [53:10] or to a developer or whoever. [53:13] Um, but yeah, it's, um, the, the voice piece of it may, and again, that's [53:18] probably because I'm just a snob. [53:20] I'm writing is Yeah, [53:22] no, it's, it's not great. [53:23] So, and a a quick little note I'll say is I, I have offered to make my [53:27] team, uh, don't be so precious hats, um, for exactly that reason actually. [53:32] Yeah. [53:32] As long as you're not using horizontal rules and like emoji [53:35] headings, like you're good. [53:37] Like customers just want their thing answered quickly. [53:40] And I do think like that's the key is to, to address the cu it needs [53:45] to be correct and it needs to be, uh, respectful and like, you know. [53:51] To tie it to that sort of like the IBM thing you shared earlier, Zack [53:54] is like, you still own the outcome. [53:56] It's still you, right. [53:57] Owning what it does. [53:58] And I think it gets scary when you start seeing companies that are literally just [54:02] giving the AI free reign to go do stuff. [54:04] And that's the, there's a cross. [54:07] I feel like maybe I'm still gonna find out I'm a ludite by [54:09] saying that, uh, eventually. [54:11] I [54:11] don't think so. [54:12] And I do think that it takes away a level of excuse that our customer [54:16] regions support agents would have before, which is, that's not our problem. [54:20] And Benny, were saying about ownership and y'all versus the problem. [54:24] Uh, when AI can read the code of the other plugins that are on their websites, [54:30] it, everything's our problem now. [54:31] Like we are in the problem solving business no matter what plugin we run. [54:36] And, and we need to help our customer as, as wholly as like, [54:41] holistically as we possibly can. [54:43] And AI unlocks that and removes a bunch of excuses. [54:45] Mm-hmm. [54:46] All right. [54:46] And so I do wanna get to the wrapping up for the last point of the, you [54:50] know, quick round table question of what your best advice is for anyone [54:53] thinking about automating their support. [54:55] I'm gonna start with you, Ben. [54:56] I do wanna just really quick though, say, I think we did answer, what are we, [54:59] four timesing at some point, which is either you're gonna do four times more [55:03] or do the same amount four times faster. [55:06] And it's up to you to decide, you know, do you own more? [55:08] I don't for the customer or not? [55:10] Don't, I'm a 10 x or no. [55:11] Or a bust. [55:11] So I'm gonna say [55:12] 10 [55:12] x. [55:13] Yeah. [55:13] Yeah. [55:13] A million or nothing. [55:14] It's no point. [55:15] No point if it's not a million times more. [55:18] But yeah, Ben, if you wanna share, um, you know, your best advice, [55:22] my best advice is be deeply curious, um, and surround yourself [55:27] with deeply curious people. [55:29] Um, and that's going to open up the doors, um, as far as, uh, things go. [55:36] And so I absolutely love, um, having a culture of, but why do we do that or. [55:44] What's the rationale behind blah? [55:48] And having an internal safe place on your team where support product marketing, [55:53] everybody can see and interject, um, and have those conversations where [55:59] you, you ask the tough questions. [56:01] Um, I love it that when we first started talking ai, um, we had places, pockets [56:07] in the company where I could have frank conversations with all levels, not just [56:13] leadership, but all levels of, of agent, supervisor, manager, director, senior [56:19] director, C-suite, doesn't matter. [56:21] We can all just talk about that stuff. [56:23] And so having the curiosity and the willingness to look [56:28] publicly foolish or semi publicly foolish, um, is the superpower. [56:34] Um, and so being, being able to put yourself out there and, and [56:39] have those conversations and, and. [56:42] Be okay with, um, the unknowns is gonna take you way farther than [56:49] trying to grab all of the pieces of AI learning you can and, and hold it [56:54] to yourself and be like, I'm gonna solve the world's problems, and then [56:57] I'm gonna post along the LinkedIn post about it and make millions of dollars. [57:02] Like now share, be curious, be, uh, be publicly curious with anybody. [57:09] Um, that's, that's going, that's what, that has been the [57:13] secret to my career success. [57:17] Yeah. [57:17] That's awesome. [57:17] How about you, Zack? [57:19] Well, I, uh, I refuse your best advice concept and I, I, I'm gonna instead [57:26] say, best advice is and say, okay, you we're talking support today. [57:32] But really what we need to do, I wanna convey, is to rethink your business [57:36] structure completely around knowledge, around solutions, around like. [57:42] AI is going to tear down the walls between customer support [57:45] and the rest of your business. [57:46] Mm-hmm. [57:46] Even more than Ben has been advocating that support be integrated into your [57:51] company, uh, like even more than it already should have been torn down. [57:55] The company is one whole thing now. [57:58] And so rethink your business structure. [58:01] Look into knowledge centered service, which is a concept [58:04] like Ben just talked about. [58:05] Everybody having their little fiefdoms of knowledge and trying to hold onto it. [58:10] 'cause that protects their job. [58:11] That's no good. [58:12] Uh, or not, not only protects their job, but like is also, uh, it is lazy to just [58:18] hold onto knowledge instead of sharing it and writing it up and, and, and [58:21] enhancing the company's knowledge base. [58:23] So knowledge centered service, check that out. [58:26] Uh, AI call at Gravity Kit every two weeks we have an AI call where we, we, [58:31] it's voluntary but everybody shows up 'cause it's a really cool conversation. [58:36] We all hop on and talk about things we're doing with AI and [58:39] things we've heard learned about and things we've heard about. [58:42] And that's been really an interesting way to get, uh, the juices flowing for [58:45] how we use it and how we could use it. [58:47] And then finally, if you haven't yet, sign up for docs, bot.ai, [58:52] d cs bt, do ai give it a play. [58:55] Like, uh, see what, how it works, integrate it with your service. [58:58] See how that works. [58:59] Just experiment because, um, if you're not already using it, [59:03] you really should, should be. [59:04] Uh, and if you, uh, if you haven't yet, today's the day. [59:08] Tomorrow's the day. [59:09] Do it now, now, now, now, now. [59:10] Because it's, uh, it's happening and learning it now is better [59:14] than learning it tomorrow. [59:16] Yeah. [59:17] If you're starting three months ago, you're already [59:19] late is what I'm feeling like. [59:20] So, uh, I'm in full agreement there. [59:22] But if you start today, you're still an early adopter. [59:24] A [59:25] hundred percent. [59:26] Yeah. [59:26] Uh, I will double down on the knowledge centered service. [59:29] That was always, it's, I a lot, like a lot of Ben's advice. [59:31] That was always a good idea. [59:32] It just, it became a thousand times more important to really care about it. [59:36] Uh, I learned about that concept from Craig Stoss at a support driven talk. [59:41] I'm sure it's recorded somewhere. [59:42] Maybe I'll see if I can try to find the link and share it out. [59:44] It's a really great concept. [59:46] Um, and. [59:47] It actually will tie that into my advice is the idea of like creating an artifact [59:51] from every support conversation is something I've always believed and I think [59:54] it's more important than ever before. [59:56] Uh, and that actually gets to sort of that knowledge synthesis of like [59:58] product feedback and understanding. [60:00] Um, we are at a bit of a disadvantage as product builders and WordPress versus [60:04] like a traditional SaaS environment because, you know, we send out zip [60:07] files, they get distributed everywhere. [60:09] And even if you're good at analytics and convincing people to opt in and [60:13] all that, you're still doing less than what like is standard out in the world. [60:17] So you gotta patch over that hole by like really caring about listening [60:21] to your customers and, you know, forming that taste from like, just [60:24] really talking to real people. [60:26] And one of the best ways to do that though is just to tie in that concept [60:30] of every single conversation generates an artifact documentation improvement, a [60:35] meaningful, you know, internal story for a feature that you guys should consider. [60:40] Like invest the time that you're saving and investigating the issue into actually, [60:44] you know, solving the root cause. [60:46] In the moment, uh, using the customer conversation as a trigger point. [60:50] We've been really focusing on that at Gravitate, or I have been. [60:53] We, so like the whole team is aligned around with, I think, uh, which is, um, [60:58] that when we update a, like, when we update a doc, uh, it's re removing future [61:05] work from the future support person. [61:08] Like, we're doing it because we, we wanna help the customer. [61:12] Of course, we, but we want to make our future selves happier. [61:15] Uh, so like the, the, it had been closing tickets, uh, was, you know, [61:20] every ticket got closed as fast as possible, as thoroughly as possible. [61:24] But no, now is take your time, you know, work the problem, [61:28] create a ticket, create a doc. [61:30] Like, it's, it's a change of mindset that I think is, is. [61:33] Uh, will payback when as AI becomes more powerful. [61:36] And I should probably do like a blog post or something, but I feel very [61:39] strongly the primary support metric is actually conversations generated [61:43] per thousand users of a product. [61:45] And so I'll leave that and then not explain it further, [61:47] but I'll write something down. [61:49] Um, but it's tied deeply to what you just said as well. [61:52] Um, so that's a wrap. [61:54] Ben. [61:54] Thank you so much for joining us. [61:56] Where do people go find you online? [61:58] So build Sustainable Support.com is the website for, uh, my support ventures. [62:05] Um, there, there's a book there, there's, uh, some services and [62:08] products and things like that. [62:10] I'm on x I'm on LinkedIn, uh, I'm on Facebook. [62:14] Uh, been UNC in most places. [62:16] Um, so yeah, would love to, to hang out. [62:20] And if you're enjoying these shows, uh, please do us a favor, hit [62:23] that subscribe button on YouTube. [62:25] It's free, and it's really good for the show. [62:28] So please do that. [62:29] And special thanks again for, uh, free me, us being our, uh, our sponsor. [62:33] We, uh, it's, it's great. [62:35] Thank you. [62:35] Yeah. [62:35] And, uh, next week you can join Katie and Matt. [62:38] They'll be hosting to talk more ai, AI for operations, building [62:42] better internal systems behind the scenes with Michael Eisen. [62:46] Wasser. [62:46] Am I pronouncing that right? [62:47] I don't know, Michael. [62:48] So is [62:49] that [62:49] correct? [62:50] I [62:50] look forward to that conversation as well. [62:52] I just, IM nom nom, eat up all the AI stuff. [62:54] So [62:55] yeah, I feel they're probably gonna build on what we've talked about here [62:58] today with Ben, so it'll be great. [63:00] Awesome. [63:01] Thanks again. [63:02] Thanks Ben. [63:02] Thanks y'all.