ICP matching
Reads your ideal-customer profile and finds companies and people that fit: by industry, size, geography, tech stack, and signals like hiring or funding.
Operational tier · Sales
An AI agent that fills the top of your funnel.
Give it your ideal-customer profile. It finds matching companies and people, enriches them with firmographic and contact data, scores them for fit, and hands a clean, ready-to-work list to your sales team, or straight to first-touch outreach.
What does an AI lead generation agent do?
A lead generation agent turns an ideal-customer profile (ICP) into a working pipeline. It searches public sources and data providers for companies and contacts that match your criteria, enriches each record (size, industry, tech stack, role, verified email), scores them for fit, deduplicates against your CRM, and writes the result to your CRM or a sheet. Optionally it hands the best-fit leads to a sales outreach agent for personalised first-touch. Operational tier, built on Claude. Typical output: 200–1,000 qualified, enriched leads per week with verified contact data.
theagency47 · Updated June 2026 The spec sheet
How the Lead Generation Agent actually works.
| Job-to-be-done | Turn an ICP definition into a deduplicated, enriched, scored list of target accounts and contacts |
| Tier | Operational (departmental flow) |
| Underlying model | Anthropic Claude Sonnet |
| Trigger | Scheduled weekly run + on-demand for a new campaign or segment |
| Inputs | ICP definition, exclusion list, existing CRM (for dedup), data-provider credentials |
| Tools available | Web search, data providers (Apollo / Clearbit / ZoomInfo), email verification, CRM read+write |
| Autonomous decisions | Fit scoring, dedup matching, enrichment source selection, list segmentation |
| Escalation rules | Ambiguous fit → flag for human review · Low data confidence → marked, not sent · Provider rate-limit → pause & notify |
| KPIs measured | Leads sourced, enrichment coverage %, email-verification rate, fit-score distribution, cost per qualified lead |
| Eval suite | 20 test cases (fit scoring, dedup accuracy, enrichment correctness). Pass rate target 90%+. |
What it does
The work it takes off your team.
Enrichment
Fills each record with firmographics, role, seniority, and a verified business email. No more half-empty rows.
Fit scoring
Scores every lead against your criteria so reps work the best-fit accounts first instead of top-of-list.
Deduplication
Checks every lead against your CRM so you never pay to re-source or re-contact someone already in the pipeline.
CRM write-back
Writes clean records straight to HubSpot, Salesforce, or a sheet: segmented and tagged by campaign.
Handoff to outreach
Best-fit leads can flow straight to a sales outreach agent for personalised first-touch, closing the gap between list and pipeline.
Integrations
Standard integration stack.
- Data providers: Apollo (preferred), Clearbit, ZoomInfo, LinkedIn Sales Navigator (where terms permit)
- Email verification: NeverBounce, ZeroBounce, provider-native verification
- CRM: HubSpot, Salesforce, Pipedrive, Attio
- Output: Google Sheets, Airtable, CSV export
- Handoff: direct pass to a sales outreach agent or sequencing platform
Custom integrations with proprietary systems are quoted as add-ons. Not sure if yours fits? Describe your stack and we'll confirm.
What it looks like in practice
A typical run.
- Monday 06:00, Agent pulls this week's ICP segment and sources ~600 new candidate accounts from data providers.
- 06:00–07:30, Enriches each record, verifies emails, scores fit, deduplicates against the CRM. ~180 drop as bad-fit or duplicates.
- 07:30, Writes ~420 clean, scored leads to HubSpot, tagged by campaign and segment.
- 07:35, Sends a Slack digest: leads sourced, enrichment coverage, top-scoring accounts, anything flagged for review.
- Rest of week, Best-fit leads flow to the outreach agent; reps work the flagged-for-review queue (usually 20–30 accounts).
FAQ
Questions about the Lead Generation Agent.
How is this different from just buying a list?
Bought lists are static, stale, and undeduplicated. The agent sources continuously against your live ICP, verifies emails at source, scores for fit, and removes anyone already in your CRM, so reps work fresh, relevant accounts, not a decaying spreadsheet.
Does it comply with GDPR?
Yes. It sources only business contact data with a legitimate-interest basis under GDPR Article 6(1)(f), respects do-not-contact lists, and a DPA is executed before deployment. It does not scrape personal/consumer data.
Can it feed my existing outreach tool?
Yes, it writes to your CRM or sequencing platform, or hands best-fit leads directly to a sales outreach agent for personalised first-touch.
What does it cost to run?
Two cost lines: data-provider fees (your existing Apollo/Clearbit plan) and Claude API usage (typically €40–€200/month at this volume). Build is delivered as a Spark engagement.
Related