# AI Agents in GTM: What They Should Truly Do > [!question] Lingering Questions > - Are the 72% (non-selling time) and 25-30% (CRM decay) statistics independently verifiable, or just single-source vendor citations doing a lot of argumentative work? > - Does "adoption is a design problem, not a training problem" hold once reps have already been burned by first-wave AI outreach and actively distrust the output? > [!tip] Article Ideas > - "Define the job before you buy the tool" — a vendor-neutral AI-agent buying checklist derived from this piece's framework, stripped of the Common Room product pitch, extended past GTM into WordPress/DevRel tooling decisions. > [!link] Cross Reference > - [[3. Resources/MOCs/AI & the Future of Knowledge Work MOC]] > - The verification-cost theme here ("bad foundation → bad output faster") rhymes with the reasoning-effort/verification-cost claims in this same batch's Fable 5 notes — both argue you can spend more fixing an ungrounded agent than the agent saved you. > [!quote] Vocabulary > - **Information vs. context** — the article's central evaluative heuristic: information is a raw fact (a funding round), context is that fact plus what makes it actionable (three new leaders, two pricing-page visits, matched to past closed-won deals). ## Metadata - Source: [[3. Resources/Readwise/Articles/AI Agents in GTM What They Should Truly Do]] - Author: Common Room (commonroom.io) - URL: https://www.commonroom.io/blog/ai-agents-gtm/ ## Author's thesis AI agents create real GTM leverage only when pointed at a well-defined, data-intensive job sitting on a clean data foundation — the value comes from job-fit and data quality, not from raw model autonomy. The piece wants RevOps readers to adopt a "define the job, fix the foundation" discipline before buying any agent tool — while positioning Common Room's own unified-data layer as the foundation that discipline requires. ## Key arguments (≤3) 1. **Agent value equals data quality times job fit**, not agent intelligence — deployments disappoint because teams point agents at the wrong job (their biggest workflow) rather than the right one (repetitive, data-intensive, clear definition of done, inconsistently done today). 2. **Research → personalization → data hygiene compound only in sequence** on a shared clean foundation; personalization run standalone (without research context) just repeats the mail-merge failure that already trained buyers to distrust AI outreach. 3. **The tool didn't fail, the deployment did** — reframes AI disappointment as a job-selection and data-readiness failure, not a model-capability failure. ## Claim atoms ### Agent value equals data quality times job fit > [!quote] ID:rw_01kx13dy4h1psaeyj10s7rv1s0 > An AI agent is only as useful as the data it operates on and the job it's been given to do. Where/why: "Why most AI agent deployments disappoint" section — the article's load-bearing thesis sentence. ### Deployment failure, not tool failure, explains disappointment > [!quote] ID:rw_01kx13dy4h1psaeyj10s7rv1s0b > The tool didn't fail. The deployment did. Where/why: Same section, the reframe that follows the thesis — relocates blame from vendor tech to job/data choices. ### Bad foundations make agents fail faster, not slower > [!quote] ID:rw_01kx13dy4h1psaeyj10s7rv1s0c > Nobody's AI stack can outrun a bad foundation. They just generate bad output faster. Where/why: Data hygiene section — the aphorism that pivots into the vendor's foundation pitch; strong on its own, worth separating from the pitch that follows it. ### Research must produce context, not just information > [!quote] ID:rw_01kx13dy4h1psaeyj10s7rv1s0d > Context is what reps need before they pick up the phone. Where/why: Research-agent section — the information-vs-context distinction is the piece's most transferable heuristic for judging any AI output, not just sales research. ## Permanent candidates - **Define the job before selecting the automation** (promoted — see linked Permanent note) — the cleanest, most portable claim in the piece; holds independent of the vendor pitch. - Note the piece's structural bias: every diagnosed problem resolves to a Common Room product (Context360, the personalization agent, DataAgent). Take the job-first/data-first framework; treat "you need *our* unified foundation" as an unverified premise, not a conclusion. ## Processing checklist - [ ] Sanity-check the 72% (non-selling time) and 25-30% (CRM decay) figures against a non-vendor source before citing externally - [ ] Read Common Room's linked "AI GTM strategy" and "signal-based outbound" posts if this thread gets pulled on further — they're cited as self-referential support - [x] Connect into [[3. Resources/MOCs/AI & the Future of Knowledge Work MOC]] ## Links - [[3. Resources/MOCs/AI & the Future of Knowledge Work MOC]] — because applies-in: "define the job before the tool" is a reusable AI-adoption heuristic for any domain, not just GTM. - [[3. Resources/MOCs/Product & Startup Instincts MOC]] — because applies-in: recognizing the vendor-blog pattern (genuinely useful framework, self-serving "foundation" conclusion) is a product-evaluation skill. - [[3. Resources/MOCs/DevRel MOC]] — applies-in: the MOC lists this note under Synthesis additions (2026-08-11, GTM backlog).