A PE and M&APE and M&A fund's revenue engine, on Attio and Deepline

The problem: a valuable network nobody could query

The fund invests across private equity and M&A, and like most funds its single most valuable asset is its network: founders, intermediaries, co-investors, LPs. On paper that network looked large. In practice it was unusable, and the raw count was never the point. Contacts had poured in for years from sources that did not talk to each other, an export here, a conference list there, enrichment runs, calendar invites, and the result was a base full of duplicates, dead roles and half-filled records. The real question was not how many contacts there were, it was how few of them were real, current and worth a partner's attention, and nobody could tell.

The symptoms were the ones every fund eventually hits. Dealflow lived partly in a spreadsheet and partly in three partners' heads. Introductions were rediscovered by accident, months late, because nobody could see who already knew a target. Deal stages were a column someone forgot to update before the Monday meeting. And the CRM itself had crossed the line every CRM can cross: the partners no longer believed what it told them, so they stopped putting good data in, which made it less trustworthy still. A CRM nobody trusts is worse than no CRM, because it costs money and lies to you.

The mandate was not "configure us a CRM". It was: make this network queryable, make the pipeline real, and automate the manual work so the fund can run on the system instead of around it, without hiring an internal engineering team to babysit it.

The build, part one: Attio as the source of truth

Automation on top of a bad data model just produces wrong answers faster. So the first work was the model itself, designed around how this fund actually invests, not a generic sales funnel. The full reasoning is in our Attio for funds guide; here is what shipped for this firm.

One human, one record

People became a single typed object: deal side (buyer, seller or intermediary in an M&A context), ticket range, sector focus, source. The discipline that mattered most was one human, one record, which sounds obvious and is the hardest thing to hold in a fund where the same founder reaches you through a Calendly address, a work email and a LinkedIn profile. Holding that line is a pipeline, not a setting, and it is covered below.

Companies and deals on the fund's real process

One Companies object covers prospects, portfolio, LP institutions and co-investor firms, separated by classification rather than by silos, because the firm that passed on one vehicle is the LP prospect for the next. Deal stages mirror the investment committee's real path, sourced, first meeting, partner review, term sheet, diligence, closed. Two attributes were made non-negotiable: an owner on every deal and a dated next step, because an unowned deal is a deal that dies quietly between two partner meetings.

Lists and relationships as live views

Saved views over the same records, never copies: an LP pipeline for the active raise, sector thesis lists, a conference-prep list rebuilt before each event, and a priority board for the Monday meeting. Native Gmail and Calendar sync gave every record a live interaction history and a connection-strength signal, the relationship intelligence funds usually buy Affinity for, included rather than billed separately.

The build, part two: 24 Deepline workflows in production

With a model worth automating, the automation layer lives on Deepline, which lets the whole thing run as versioned code the fund owns rather than as a no-code tool's credit meter. Twenty-four workflows run in production. Rather than list them all, here is the one that earns the partners' trust.

Post-meeting extraction, with an adversarial check

After every recorded call, a pipeline extracts the substance into structured CRM data: who was in the room, amounts discussed, mandate and ticket signals, agreed next steps. The part that makes it trustworthy is the second pass. An adversarial verification step re-reads the transcript and challenges every extracted claim, and anything that was not actually said in the meeting is rejected before it touches the base. In production, that check regularly blocks fabricated or inferred facts that a single-pass extraction would have written straight into live records. That guardrail is the whole difference between a CRM that fills itself and one you can trust.

The same engine runs the rest without a person in the loop: identity-level deduplication across Calendly and LinkedIn so one human stays one record, enrichment refresh through Clay so the network never goes stale, and pipeline reporting rebuilt through the Attio API, all orchestrated with n8n alongside Lemlist for outreach. The CRM is the nervous system, not the whole body.

What changed for the fund

Forget the counters. What changed is how fast the fund moves, and how far it trusts its own data.

The headline is an outcome, not a counter. With the structure in place, the fund closed a €30M allocation in under three weeks: the work that usually slows a deal down, the warm path, the full interaction history, the qualified counterparties, was already done by the system instead of by a partner the night before. Automation turned a scramble into a query.

Underneath it, the base earns its trust by being ruthless about what gets in and honest about what it records. The network is queryable, the pipeline is real, the busywork is gone, and the fund moves faster on live deals, without an internal engineering hire.

What it runs on

A deliberately small, owned stack, each piece doing one job and the fund holding the keys to all of it.

• Attio — the source of truth: data model, pipeline, relationship intelligence, and the API the reporting was built on.
• Deepline — the automation layer: the 24 production workflows as versioned code, from extraction to dedup to enrichment.
• n8n — orchestration and secure connections to inboxes and external systems.
• Clay and provider APIs — the enrichment waterfall behind the refresh.
• Lemlist — outreach, fed by and feeding back into the CRM.

No platform lock-in, no black box. Everything is documented and handed over, which is the same way every build we ship works.