fivos-account-enrichment

Fivos Account Enrichment

Company-level enrichment. Runs after sourcing, before contact pulls. All enrichment research runs through Clay (mcp__claude_ai_Clay__*), which is already wired in this environment. Clay orchestrates the underlying providers (BuiltWith, Crustdata, Firecrawl, PDL, etc.) under one roof.

Hard rules

• Clay is the orchestrator. Do not call per-provider APIs directly from this skill. Clay handles provider routing, credit accounting, and dedup inside a table.
• One Clay table per (client, campaign). Never mix clients in the same table - makes credit accounting and list reuse impossible.
• Always pilot before the full run. A 1-3 row pilot catches column-spec mistakes that would otherwise spend across hundreds of rows.

Inputs

• Company CSV from fivos-account-sourcing (qualified).
• Column spec - the custom fields to add. Typical sets:
  • Tech stack (tech_stack, primary_cloud, crm, warehouse).
  • Funding (last_round_stage, last_round_amount_usd, last_round_date, total_raised, investors).
  • Hiring signals (open_roles_count, open_roles_target_titles, hiring_velocity_90d).
  • Headcount history (headcount_now, headcount_90d_ago, headcount_growth_pct_90d).
  • Web / content signals (blog_last_post_date, homepage_last_change_date, has_press_page).
  • Compliance / security (soc2, iso_27001, hipaa, gdpr_dpo_listed).
• Budget ceiling (optional).
• Client tag + campaign tag (required).

Outputs

1. Enriched CSV exported from the Clay table with all requested columns.
2. Running Clay credit usage ledger.
3. Coverage report per column.

Provider-to-column cheat sheet (what Clay routes to under the hood)

Column type	Primary provider	Fallback
Tech stack	BuiltWith	Crustdata (coarser), Firecrawl (homepage scrape)
Funding graph	Crustdata	AI Ark, Apollo
Hiring signals	PeopleDataLabs (PDL)	Crustdata, Firecrawl careers-page
Headcount history	PDL	Ocean.io, AI Ark
Revenue estimates	Crustdata	Apollo, AI Ark
Web signals (blog, homepage, press)	Firecrawl	Apify
Compliance flags (SOC2, HIPAA, etc.)	Firecrawl (footer / trust page)	Manual validation
Industry taxonomy	Ocean.io (NAICS) + AI Ark (plain-english)	keep both
Custom intent signals	Exa	Firecrawl

You don't call these providers directly from this skill - you tell Clay which column to add and Clay picks the provider.

Workflow

Step 1 - Column plan + cost estimate

For each requested column, map to a Clay data point (using the cheat sheet above). Get current credit availability:

mcp__claude_ai_Clay__get-credits-available({})

Print a condensed plan:

Enrichment plan - <client> / <campaign>

Clay credits available: N
Columns requested:
  - tech_stack (BuiltWith)
  - last_round_stage (Crustdata)
  - open_roles_count (PDL)
  ...

Estimated credits used: ~X (Y rows x Z credits/row average)
Est. cost: $<credits_x_credit_price>

Proceed? Manager approval required if over budget.

Gate on approval if over budget; auto-proceed otherwise.

Step 2 - Create the Clay table and ingest the CSV

Single Clay call:

mcp__claude_ai_Clay__add-company-data-points({
  table_name: "<client>_<campaign>_accounts",
  source_csv: "<path to enriched account CSV>",
  data_points: [
    "tech_stack",
    "last_round_stage",
    "last_round_amount_usd",
    "open_roles_count",
    "open_roles_target_titles",
    ...
  ]
})

For companies not already in the table, first use:

mcp__claude_ai_Clay__find-and-enrich-company({...})

Step 3 - Pilot (1-3 rows)

Before running the full set, run the column enrichment against 1-3 rows and inspect:

• Are the returned values sane?
• Is coverage acceptable?
• Any columns that came back null for all pilot rows? → revisit column spec or provider.

Step 4 - Full run

Once the pilot looks right, extend the enrichment to all rows. Monitor the Clay run:

mcp__claude_ai_Clay__get-task({ task_id: <id> })

Step 5 - Ask questions / custom research

For ICP-specific intel not in a standard column (e.g. "do they use OpenAI or Anthropic", "are they hiring for AI engineers"), use Clay's natural-language research:

mcp__claude_ai_Clay__ask-question-about-accounts({
  table_name: "<client>_<campaign>_accounts",
  question: "Does this company publicly use OpenAI or Anthropic in their product?"
})

This is the replacement for "custom scrapes" - Clay spins up the research per row and writes a column with reasoning.

Step 6 - Coverage audit

Compute coverage per column. Flag any high-priority column below 70% coverage - usually a Clay fallback can close the gap (e.g. swap BuiltWith for Firecrawl homepage scrape).

Step 7 - Export + handoff

Export the enriched table from Clay as a CSV. Hand off to fivos-contact-sourcing - the contact step uses company-level signals (hiring, funding) to drive persona spec and contact-count-per-account.

Cost controls

• Never enrich columns without a downstream consumer. If the copywriter doesn't need it and scoring doesn't weight it, drop it.
• Always pilot on 1-3 rows.
• Use mcp__claude_ai_Clay__get-credits-available before every large run to avoid mid-run credit exhaustion.

References

• ~/.agents/skills/gtm-meta-skill/enriching-and-researching.md - enrichment patterns, coalescing, multi-pass pipelines.
• ~/.agents/skills/gtm-meta-skill/provider-playbooks/builtwith.md
• ~/.agents/skills/gtm-meta-skill/provider-playbooks/crustdata.md
• ~/.agents/skills/gtm-meta-skill/provider-playbooks/firecrawl.md
• ~/.agents/skills/gtm-meta-skill/provider-playbooks/peopledatalabs.md
• ~/.agents/skills/gtm-meta-skill/provider-playbooks/exa.md