GTM SPRINT · Field guide · Nº 01

Don't automate stages. Automate the jobs that repeat.

Field guide Nº 01·2 min readPromptsResearchGenerateScoreRoute

The 10-stage pipeline runs on four jobs that show up again and again: research, generate, score, route. Build each one once, call it everywhere it repeats - cheaper to build, far cheaper to maintain, and it gets smarter every cycle.

12345678910 → ResearchGenerateScoreRoute Ten stages collapse into four reusable jobs.

01Clay · Claygent

Research

Find the single strongest buying signal for every account, at scale - so targeting and positioning run on real triggers, not guesses.

Reused in 12345678910

Visit each account's site, LinkedIn and recent news. Return the strongest buying signal from the last 90 days (funding, senior hire, launch, hiring spree) in one sentence, with the source URL. No signal → return "none".

02Claude / GPT

Generate

Turn one point of view into a full distribution set - and later, into the brief that arms the rep. One idea, many surfaces.

Reused in 12345678910

You're my demand-gen writer. Take this POV asset: {asset}. Produce 3 LinkedIn posts (hook + 120 words), 1 nurture email (subject + 90 words), 3 ad hooks. Match this voice: {samples}. One idea each, no emoji, no hashtags.

03Claude / GPT

Score

Rate every lead on fit, intent and timing - with a one-line reason you can audit. The same brain decides nurture and qualify.

Reused in 12345678910

Score this lead 0-10 on fit, intent and timing against my ICP: {icp}. Lead + enrichment: {data}. Return JSON {score, tier, reason}. The reason must name the signal that moved the score - don't restate the fields.

04Make + Claude

Route

On a threshold, assign the right rep, alert them, draft their brief, and write it all back to the CRM. No lead sits cold, no manual entry.

Reused in 12345678910

Lead scored {score}, routed to {rep}. Write a 3-bullet brief: (1) who they are + why now, (2) the trigger signal, (3) the opening line to use. Under 60 words. Plain, no fluff.

Build each job once. When your ICP shifts or a model changes, you fix one prompt and every stage that calls it updates at once. That's why this is cheaper to run - and why it compounds every time you tune it from the measure-and-optimize loop.

Test. Learn. Refine. Repeat.

Stages 6 & 8 light up under two jobs - that overlap is the reuse paying off. gtmsprint.run · @FrankiChamaki

Common questions

What is the four jobs field guide?

A one-page reference with the exact prompts for research, generate, score and route, ready to paste into your stack.

How do the four jobs map to the 10 pipeline stages?

Each job is reused across several stages. Stages 6 and 8 call two jobs each, and that overlap is where the reuse pays off.

Do I need all four to start?

No. Research and generate cover the first stages on their own. Scoring and routing can be added once you have lead volume to sort.