Ten stages. Four jobs.
Most teams automate their GTM pipeline one stage at a time. That is the expensive way to do it. Here is the framework that collapses ten stages into four things worth building.
Open any go-to-market playbook and you get the same picture: a pipeline of ten stages, from ICP definition to a generated, accelerated revenue number at the end. It is a good map. The problem starts when people try to automate it - because the instinct is to build an automation for each stage. Ten stages, ten workflows, ten subscriptions, ten things to maintain.
That is a lot of surface area for something that breaks the week your ICP shifts or a model changes underneath you. And most GTM automation does not die at the build. It dies in maintenance.
There is a cleaner way to see the same pipeline. Those ten stages are not ten distinct problems. They are four jobs that keep repeating.
The four jobs
- 1Research - find the single strongest buying signal for every account, at scale. Clay · Claygent
- 2Generate - turn one point of view into a full content set, and later into the brief that arms the rep. Claude / GPT
- 3Score - rate every lead on fit, intent and timing, with a one-line reason you can audit. Claude / GPT
- 4Route - on a threshold, assign the right rep, draft the brief, and write it back to the CRM. Make + Claude
Read the pipeline through that lens and the repetition jumps out. Research does not just build your ICP list - it runs again inside qualification, when you re-research a lead to score it. Generation writes your demand content and later writes the account brief for the rep. Scoring decides both your nurture branches and your qualified handoff. Routing carries a lead from qualification through enablement into a clean CRM record.
| The 10 stages | ResearchClay · Claygent | GenerateClaude / GPT | ScoreClaude / GPT | RouteMake + Claude |
|---|---|---|---|---|
1ICP & Market SegmentationDefine and prioritise best-fit accounts | ||||
2Positioning & MessagingDifferentiated angle per account | ||||
3Demand CreationContent, campaigns, SEO, ads | ||||
4Lead CaptureLanding pages, forms, instant enrichment | ||||
5Nurture & EngageBranch content by score | ||||
6Score & QualifyreuseRe-research the lead, then score it | ||||
7Route to SalesAssign the right rep, alert, SLA | ||||
8Sales EnablementreuseGenerate the brief, route it to the rep | ||||
9CRM & AutomationWrite everything back, keep data clean | ||||
10Pipeline GeneratedThe output - not a job, the result | ↑ Measure · Analyse · Optimise → feed learnings back into the four jobs | |||
Why this is cheaper - and why it compounds
The saving is not just fewer builds. It shows up in three places.
Build once, not ten times. One scoring engine called in nurture and in qualification is one thing to build, not two.
Maintain one thing, not ten. This is the bigger, quieter win. When your ICP shifts or a model changes, you fix one prompt and every stage that calls it updates at once. Ten bespoke automations means ten places to hunt down and patch.
It compounds. Every time you tune a job from the measure-and-optimise loop, every stage using it gets better simultaneously. Ten separate automations do not compound - you improve one and the other nine sit stale.
Where it breaks - and how to hold it honestly
This holds when a job is genuinely the same across stages. Sometimes it is not. Scoring an inbound form-fill and scoring a cold outbound account use different signals, so you reuse the engine but swap the rubric. The rule is: reuse the structure, tune the inputs. If you find yourself rebuilding a job from scratch for one stage, that stage probably needs its own job - and that is fine. Four is the pattern, not a hard ceiling.
Get the build kit
The four prompts, ready to paste, plus this mapping as a one-page reference.
FAQ
What are the four GTM jobs?
Research, generate, score and route. Research finds the buying signal for each account. Generate turns one point of view into content and rep briefs. Score rates every lead on fit, intent and timing. Route assigns the rep, drafts the brief and writes it back to the CRM. Each one repeats across several of the ten stages.
Why not automate each pipeline stage separately?
Stage-by-stage means ten builds and ten things to maintain. When your ICP or model changes, you patch ten workflows. Automating the four repeating jobs means four builds you reuse everywhere - and one prompt to fix when things change.
What tools run each job?
Research runs in Clay with its Claygent research agent. Generate and score run on a model such as Claude or GPT. Route runs in an orchestrator such as Make, which calls the model to draft the rep brief and writes the result back to your CRM.
Does this replace my CRM or existing tools?
No. The four jobs are the reusable logic that runs across your stack - your CRM stays the system of record. The point is to stop building bespoke automation per stage and start building reusable jobs that any stage can call.