Matchmaking in Attio: M&A, Brokerage and Lending

1. The short version

Capture every conversation, note and email without asking anyone to fill in fields. Surface the buyers, sellers, investors or lenders that fit a brief, with a reason and a rating. Compound what the firm learns, so every rejected match makes the next one better.

Most deal firms already hold the knowledge they need to make great matches. It just lives in partners' heads, old emails and a spreadsheet called buyers v7. This guide shows how to turn that into a working system in Attio, from a version you can build yourself in an afternoon to one that reads your whole relationship history. It is written for M&A boutiques, business brokers and placement agents, with notes for lending and corporate finance.

2. What matching means in a deal firm

Every firm that moves companies or capital does the same job: holds a mandate on one side and a pool of appetites on the other, and finds the fit. The names change by sector; the shape does not.

Corporate finance houses raising capital and syndicators placing a deal with co-investors follow the same pattern. If your firm holds mandates and knows people with appetites, this applies to you.

3. Why it breaks today

The knowledge exists. The system for using it does not. The same four failures show up in almost every firm I speak to:

• Criteria live in heads. The partner who knows a buyer wants healthcare services under £20m is the only one who knows it, and only until they forget.

• Lists go stale. The buyer list is a spreadsheet, updated before each process and never between them.

• Conversations evaporate. Someone mentioned the perfect acquirer on a call six months ago. Nobody wrote it down, so it never resurfaces.

• Relationship strength is invisible. Two buyers look identical on paper. One you have spoken to eleven times this year, one never. The spreadsheet cannot tell.

"We keep our buyer list in a spreadsheet." The most common sentence on my scoping calls, from firms of every size.

A London placement agent and an M&A adviser I have known for years both talked, long before it was possible, about wanting The Brain: one place that remembered everything the firm knew and could be asked a question. It is possible now. It still takes skill to build, and it only works if the raw material is captured first.

A family office I spoke to this year was the ambitious version of this: a huge network, serious deal volume, and a clear need for a system that could keep up with both. The same design serves a three-partner boutique. The difference is scale, not shape.

4. Capture: get the knowledge in without asking anyone to type

The principle: structured data is the destination, not the price of entry. Partners will not fill in fifteen fields after every call. So the system reads what already happens - emails, meeting notes, call transcripts, the odd WhatsApp forwarded in - and proposes the structured updates itself. A human approves them.

Sources sync into the timeline; AI proposes field values from it; a reviewer approves before anything becomes structured data.

In practice this means three layers inside the CRM:

1. The timeline. Everything that happened with a company or person, synced automatically. Attio does most of this out of the box for email and calendar; call recorders and note tools connect in.

2. The reading layer. AI reads the timeline and extracts appetite: "Interested in UK healthcare services, £5-20m EBITDA, not before Q2." It writes that into the fields, flagged as proposed.

3. The review habit. Once a week, someone with judgement reviews proposed changes. Ten minutes, not ten hours.

The standard to hold: nothing the AI writes should be unsourced. Every proposed field value should point back to the email, call or note it came from, so a partner can check it in one click.

5. Surface: a brief goes in, a ranked shortlist comes out

A matchmaker should be AI, and it should weigh three things in order.

The filters run in order: hard criteria first, context second, relationship last, and each row of the shortlist carries its reason.

The sequence matters. Hard filters remove what can never fit. Context separates the plausible from the keen. Relationship strength decides who gets the first call. A system that skips straight to AI scoring without the hard filters produces confident nonsense; a system that stops at the filters produces a phone book.

Each match carries a score, an A/B/C rating, a two-line rationale that cites real fields, and a note of what is missing from the record. Partners read fifteen rows with reasons. That is the product.

6. Compound: every decision teaches the system

The matches a partner rejects are worth as much as the ones they shortlist. Each shortlist, rejection and outcome goes back into the record, so the next brief starts from a smarter position.

Decisions and outcomes flow back into appetite data, so each new brief starts from what the firm has already learned.

Three habits make the loop turn:

• Rejections need a reason, picked from a short list (wrong size, wrong sector, timing, relationship), not a free-text box nobody fills.

• Outcomes get recorded: who was approached, who engaged, who bid. These are the strongest appetite signals you will ever have.

• The review happens weekly. A loop that turns quarterly is a report, not a system.

Note: this is where a firm's own moat forms. Two boutiques can buy the same software. Only one has three years of recorded decisions about who wanted what.

7. Where to keep the brain

You can store this anywhere. Attio is the best place I have found, for two reasons. It holds structured fields and unstructured history on the same record, which is exactly the blend matching needs. And it connects to the sources - email, calendar, call recorders, data providers, automation tools - without a developer in the room.

A note on Boardy. If you have heard of Boardy, the AI that phones you, learns what you are working on and introduces you to people in its network, you already understand the instinct behind this guide. The difference is ownership. Boardy matches you inside its network. The system described here matches inside your firm's private relationships and deal history, which nobody else can see or sell.

The brain is not the software. It is the structured appetite, the captured context and the record of decisions. The software is just where it lives.

8. Design choices

There is no single right build. Four decisions shape it, and each has a sensible default.

My default for a boutique: Brief as a custom object, Matches as a custom object linked to both Brief and Company, AI run outside Attio and written back, partners working in a saved Attio view. It is the simplest version that still compounds.

9. The data model and a readiness check

The hard part is not the matching algorithm. It is structured appetite on both sides. AI will rank ten thousand companies against a brief in minutes. What most firms lack is a consistent record of what each buyer, investor or lender actually wants.

One brief and one company can each have many matches; the same appetite fields sit on both sides so they can be compared.

The appetite fields that matter, on both the Brief and the Company:

• Sector tags (a controlled list, not free text)

• Size range (revenue or EBITDA, min and max)

• Geography

• Deal type (majority, minority, bolt-on, growth, debt)

• Hard excludes (never show me)

• Soft preferences (nice to have)

• A short narrative, in the partner's own words

Is your estate ready? Score one point for each:

• Sector is a tagged field, not a free-text box

• At least 70% of companies have a description longer than one line

• Duplicates have been merged in the last six months

• Email and calendar sync covers every partner

• Someone owns data quality, even part-time

Four or five: start matching. Two or three: clean up first. One or none: begin with a migration and enrichment pass, or the matches will be as thin as the records.

10. Three ways to do it

The right level depends on the size of your estate and how the team wants to work, not on ambition.

Each step up handles a larger estate and costs more to set up, but less per brief to run.

Note on credits: Attio's AI attributes and workflow steps consume workspace credits. Scoring 150 candidates against one brief at Level 2 can use 150 or more credits; a research agent step is around ten times that. Check the run viewer on a real brief before you budget. Above a few hundred candidates, Level 3 is cheaper as well as faster.

11. DIY: a first version in an afternoon

This is Level 1 with a step towards Level 2. It needs a Pro workspace for custom objects, an AI assistant you trust, and about three hours. Screenshots on the live page are from a demo workspace, dated, because the interface changes.

Part A: set up the objects (45 minutes)

1. Create a Brief object with the appetite fields from section 9, plus Status (Draft / Ready / Matched / Closed) and a link to the client record.

2. Add the same appetite fields to Companies if they are missing. Make Sector a multi-select, not text.

3. Create a Match object with: Brief (record link), Company (record link), Score (number), Rating (A/B/C), Rationale (text), Data gaps (text), Review (New / Shortlisted / Rejected), Rejection reason (select).

4. On the Brief record page, add the Matches as a related-records panel.

Part B: write one real brief (15 minutes)

Fill in every field for a live or recent mandate. Write the narrative the way the partner would say it out loud. This is the test case for everything that follows.

Part C: run the first match by hand (60 minutes)

1. Filter Companies in Attio by the hard criteria. Export the result; aim for 50 to 150 rows.

2. Give your AI assistant the brief, the export, and these instructions:

You are a deal analyst. For each company, score fit to the brief from 0 to 100, give a rating A/B/C, write a two-line rationale citing only fields present in the data, and list any missing information. Rank the output. Do not invent facts.

3. Review the top 20. Note where it is wrong and why - this is your first tuning data.

4. Create Match records for the ones worth a conversation.

Part D: automate one step (30 minutes, Level 2)

Build an Attio workflow: when a Brief's status changes to Ready, run an AI step over the filtered companies and write Match records. Watch the credits it uses on the first run before you leave it on.

Note: if Part C takes you more than an hour, the problem is almost always the data, not the AI. Go back to the readiness check.

12. What a good match looks like

A partner can act on this without opening the record: the fit, the evidence, how warm the relationship is and what is still unknown.

A good match answers four questions in the time it takes to read it:

A worked example. Brief: UK founder-led healthcare services, £5-20m EBITDA, majority sale, no PE-backed platforms. A good card reads: "A, 86. Sector and size fit; three UK healthcare acquisitions since 2023 per enrichment. Told us in March they would stretch for founder-led. Last contact 3 weeks ago. Gap: EBITDA range not confirmed this year." A bad card reads: "Strong strategic fit, highly recommended." One cites evidence; the other asks you to trust it.

13. Making it stick with partners

A matching system fails when it is a tool nobody opens, not when the scores are off. Three things keep partners using it:

• Put it where they already look. Matches appear on the Brief record and in the weekly pipeline view. No new tab, no new login.

• Make rejection cheap and useful. One click, one reason from a short list. The reason feeds the next run.

• Show the relationship, not just the fit. Partners trust warmth more than scores. A card that says "you spoke three weeks ago" gets acted on.

A ten-minute slot in the Monday pipeline meeting to review new matches is worth more than any feature.

14. Costs, risks and client data

Callout: AI reading your firm's conversations is the most valuable part of this design and the part most likely to worry a compliance-minded partner. Decide the boundaries first, write them down, and only then connect the sources. A matching system built on data people did not know was being read will be switched off the first time someone notices.

Honest limits: matching is only as good as the descriptive text on each record; AI rationale can sound confident when it is guessing, which is why it must cite fields; and no system knows about the buyer a partner met at a conference yesterday until someone writes it down.

15. Want it built, or want the app?

Two routes from here, depending on how far you want to go.

Have it built. I design and build matching systems for investment, advisory and brokerage firms on Attio, from the data model and migration through to the engine and the partner review habit. It starts with a scoping call and, for most firms, a short paid audit of the existing estate. Book a scoping session.

Register interest in the app. I am building a matching app that runs inside Attio: a Brief record, a Find matches button, a ranked widget with reasons and ratings, and a workflow block your team can drop into any automation. It is being piloted with a small number of firms. If you would like to be told when it is ready, or to be considered for a pilot, leave your details here.

Written by Will Sawney, founder of Sideways CRM and an Attio Expert Partner with 35+ builds for deal firms.

—

16. FAQ

Can Attio do deal matching out of the box? Partly. Attio can filter companies by structured fields and run AI steps in workflows, which covers small estates. Ranked matching with rationale, relationship weighting and feedback needs a data model designed for it and, for larger estates, an external engine or app.

Do I need a custom object for Briefs? If you run several mandates at once, yes. A custom object gives each brief its own fields, status and linked matches. For occasional matching, a structured note on the client's company record is enough to start.

How much does AI matching cost in Attio? It depends on the level. Attio's built-in AI steps consume workspace credits, roughly one per candidate scored and more for research agents. An external engine using an AI provider's API typically costs pennies to a pound or two per brief in tokens.

Is it safe to let AI read our emails and call notes? It can be, with boundaries set first: which mailboxes are in scope, which provider processes the data and under what terms, and which briefs are restricted. Every AI-written field should cite the source it came from so a partner can check it.

Does this work for placement agents and lenders as well as M&A? Yes. The pattern is the same: a mandate on one side, a pool of appetites on the other. Placement agents match funds to LPs; lending brokers match borrowers to capital sources. Only the field names change.

Can I build this myself? A first version, yes, in an afternoon, using the walkthrough in section 11. A version that reads your whole relationship history, scores at scale and improves with feedback is a build project, and usually worth doing with someone who has done it before.