Enrichment
Domain → person → role → buying group in one pass. Stack-of-record stays current.
In DomainEmailAccount
Out Enriched recordBuying group
ClayApolloLinkedIn
Agentic GTM,
in the wild.
We've shipped agentic GTM into live engines. Not a packaged product, a working pattern. Below is the shape it takes when it lands: agents that own motions, flows that move signal, surfaces wired to one source of truth.
Agent patterns
Six agents, each shipped into real work.
These are the agent shapes we keep reaching for. Patterns we've built into live engines, not a product menu. Pick one. The canvas shows how it reads, gates, routes, runs and hands off.
Inputs
T ICP signal
⫶ Target account
↗ Brief
⌖
Lookup account clay.accounts.find
◊
Account exists? totalFound > 0
◉
Research news + hiring + intent
ƒ
Buying group? persona ∈ ICP
◉
Build sequence personalise(brief)
ƒ
Reply listener classify(reply)
⏚
Handoff warmth > threshold
Outputs
Sequence live
CRM update
AE handoff
Flow patterns
Four reusable flows. The plumbing under every agent.
Enrichment, research, signal capture and orchestration. Versioned, observable, shared across every agent. Built once per engagement, reused everywhere after.
Inputs
Domain Email Account Trigger Schedule Webhook
Research
Account-level synthesis (news, hiring, tech stack, intent) turned into briefing docs.
In AccountTrigger event
Out Briefing docIntent score
ClaudeApolloCrunchbase
Signal capture
Site, CRM, support, marketing, third-party, captured into a single signal record.
In Site visitIntent feedCRM event
Out Signal recordRouting trigger
Common RoomDealfrontClay
Workflow automation
Reusable Make / n8n / Workers flows. No copy-paste glue, no scripts on a rep's laptop.
In TriggerScheduleWebhook
Out Action chainAudit log
Make.comn8nCloudflare Workers
Outputs
Enriched record Briefing doc Signal record Routing rule Action chain
Surfaces
Four surfaces, one signal layer.
Website, CRM, strategy and stack, wired into the same signal layer. Not bolt-ons. Not duplicated systems. One source of truth, four places to act on it.
WEB Active
Website
Live chat, lead grading, eligibility tools and content that adapts to who's looking.
- Lead grading
- Eligibility tool
- Adaptive content
- Site chat
CRM Active
CRM
Auto-enrichment, scoring and next-best-action wired into the source of truth.
- Auto-enrichment
- AI scoring
- NBA prompts
- Field hygiene
STR Active
Strategy
AI partner that synthesises signals into quarterly bets and motion changes.
- Signal synthesis
- Quarterly bets
- Motion brief
- Budget rebalance
STK Active
Tech stack
Provisioning, integration and observability as a service. The stack stays current.
- Provisioning
- Integration
- Observability
- Migrations
Signal to action
Signal in, action out.
Captured, categorised, scored, prescribed. The signal layer turns multi-source noise into the action queue agents actually run on.
01
Capture
IngestionMulti-source ingestion at scale (site, CRM, support, intent and third-party) normalised into one signal record.
First-party signal Third-party signal CRM event
02
Categorise
ClassificationClustering and tagging. Raw signal turned into the action queue the agents actually use.
Signal cluster Owner tag Routing path
03
Score
PrioritisationPriority = f(signal, ICP fit, recency, deal context). Calibrated to your motion, not the vendor demo.
Priority score Tier band Decay rate
04
Predict + prescribe
ActionPipeline health, churn risk, conversion modelling, next-best-action, surfaced where the work happens.
Pipeline forecast Churn risk NBA prompt Motion shift
Start with one
Want a piece of this running for you?
Most agentic work starts as one agent, one flow, one dashboard, wired into your real stack. Add the next piece once it's earning its keep.