The Dealmakers' OS

The partner who is the system

In most small investment and advisory firms, there is one person who knows everything.

They know that the founder of the target company fell out with his co-founder in 2019, and why that matters. They know which of the firm’s contacts will actually pick up the phone, and which will let it ring. They know that a conversation at a conference eighteen months ago is the reason a mandate is now live. None of this is written down anywhere useful. It lives in their head, in a handful of email threads, and in a spreadsheet that only they update.

What that person is really doing, several times a week, is spotting a connection nobody else in the firm can see: this buyer and that asset, this introduction now rather than next year, this business worth a second look when everyone else has moved on. It is the most valuable activity in the firm, and it depends entirely on one person happening to hold the right things in mind at the right moment.

“It is the most valuable activity in the firm, and it depends entirely on one person happening to hold the right things in mind at the right moment.”

This works. It has worked for decades. It is also the single largest concentration of risk in the firm, and the reason the firm cannot grow without becoming worse at the thing it is good at.

Every firm of this size knows it. Most have tried to fix it, usually by buying software. And most have discovered that the software they bought made a record of what already happened without making a single connection anyone would have missed.

I have some history with this. At a firm I worked for years ago, we put "create connections" on the wall. It was the promise, and everyone in the building believed it. The data behind it was a mess, and we quietly missed nearly every connection we had promised to make. I left partly because I wanted to work on the version where the promise was true, found a platform flexible enough to build it on, and have spent the years since doing that for other people's firms.

That gap - between a system that remembers and a system that shows you something - is what this piece is about.

Why the gap costs more than it used to

The complaint is not new. Firms have always worked this way, and for a long time the cost of it was tolerable. What has changed in the last two or three years is the price of carrying on, for three reasons that have nothing to do with software.

The easy returns have gone. For a decade, a great deal of value came from buying at one multiple and selling at a higher one, with cheap debt doing much of the work in between. That is over. Rates are higher, multiples are flat, and exits are slower than anyone planned for. Holding periods have stretched, distributions have slowed, and investors are asking harder questions about cash returned rather than paper gains. The consequence for a firm of any size is that value now has to be created rather than captured. That is operational work, and operational work runs on information.

The broad process is breaking. Wide auctions with dozens of invited parties are converting badly. Buyers will not commit real diligence resource to a process where they believe they are one of forty. Sell-side advisers are responding by running narrower, better-matched processes: fewer buyers, chosen deliberately, approached with a thesis rather than a teaser. That only works if you genuinely know who the right five buyers are, and why. Which is a data problem dressed as a relationship problem.

Speed has stopped being a differentiator. AI has compressed the mechanical parts of deal work considerably: reading data rooms, summarising contracts, drafting first-pass memos. This is real, and it is now widely available. Anything widely available is not an advantage. What it does is push the advantage upstream, into knowing about the opportunity earlier than the other side, and downstream, into what you do with the business once you own or advise it.

Both ends of that shift reward the same thing: seeing the connection before the other side does. Which is no longer a question of who has the sharpest instinct, but of whether the firm can put what it already knows in front of that instinct in time.

“It’s no longer a question of who has the sharpest instinct, but of whether the firm can put what it already knows in front of that instinct in time”

Why the available tools do not solve it

Three options are usually on the table, and each fails in a specific way.

The specialist data platform. Tools that surface signals - funding events, executive moves, hiring patterns, filings - are genuinely useful, and the good ones are getting better. But they know the market and they do not know you. They cannot tell you that one of your partners sat on a board with the new CFO, or that you passed on this company two years ago for a reason that has since ceased to apply. They produce a stream of things that might matter, with none of the context that determines whether they do. Used alone, they become another tab nobody opens.

The relationship CRM. The established relationship-intelligence products solved a real problem: they take the manual work out of logging who spoke to whom. That is worth something. But they impose a fixed shape on the firm. Contacts, companies, deals. A pipeline that moves in one direction. Real work in this world does not have that shape. A single relationship might be a buyer on one mandate, an introducer on another, and an investor in a third. A deal might involve a platform, several targets, a lender, two advisers and a co-investor, all of which are entities in their own right with their own histories. Force that into a fixed schema and the parts that do not fit go into a spreadsheet, which is where the institutional memory quietly leaks away.

The spreadsheet. Which is, to be fair, honest about what it is. It fits the firm perfectly because the firm built it. It also cannot capture anything automatically, cannot be trusted by anyone who did not write it, and stops existing the day its author leaves.

The pattern is consistent. The rigid tools do not fit the work; the flexible tool does not do any work. None of the three ever surfaces a connection you had not already thought of, which is the only thing that would justify the licence fee. Firms end up with all three, and with the real system still sitting in one person’s head.

The first row is worth sitting with. A specialist data platform will always know more about the market than anything you build, and that is an argument for using one, not for doing without. The question is what it feeds.

One category is deliberately absent from that table: the investor system of record. These tools handle commitments, capital calls, distributions, statements, the audited trail of who is owed what. That is administration and reporting, it has a regulator attached, and a firm that has one should keep it. But it knows what an investor committed; it does not know why they said yes, what they declined last year, or who made the introduction that started it. The two need connecting, not merging.

What the system has to do

An operating system - we can call it the Dealmaker’s OS - is not a better CRM. It is the layer everything else runs on, and its job is singular: to make the connections in your firm’s own knowledge visible without anyone having to go looking for them. Four properties get you there.

A Dealmaker’s OS is the layer everything else runs on

It holds the actual shape of your business. Not contacts and deals, but whatever your firm genuinely trades in: mandates, buyers, targets, funds, introducers, lenders, portfolio companies. Each a first-class thing, related to the others in the way they are actually related. If the model does not match the work, people will not maintain it, and an unmaintained system is worse than none.

It captures without being asked. Email, calendar, notes, documents, flowing in by default. Anything that depends on a busy principal remembering to log it will not be there when it matters. This part is unglamorous and it is the foundation of everything above it.

It is open at both ends. Signals from outside come in; the firm’s structured knowledge goes out to whatever needs it. A system that cannot be connected to is a silo with a nicer interface. This is where the specialist data tools stop being competitors and start being inputs: they are good at watching the market, and they belong feeding a system that knows your firm.

It acts. Something changes in the world, and the system does something about it - surfaces it to the right partner, with the relevant history attached, at the point it is useful. The connection arrives rather than waiting to be noticed. This is where structure earns its keep, and it is also why the AI question resolves the way it does: an AI answering questions over unstructured, half-maintained records will give you confident answers that are wrong. The same model over a well-built graph of your firm’s actual relationships and history is a genuinely different proposition. The model is not the differentiator. What you point it at is.

Why build it on Attio

An honest answer, because the alternative arguments are not necessarily wrong. A bare database gives you the flexibility and none of the adoption; a purpose-built product for your corner of the market gives you the opposite, and works well right up to the point where your firm stops looking like the one it was designed around.

Attio sits in the useful position between the two. It has the relational modelling of a database - you define the entities, and the relationships between them, without a developer - alongside the communication capture, interface and workflow engine of a modern CRM. It is open enough at the edges to take feeds from elsewhere and to be queried by whatever comes next.

Attio has the relational modelling of a database, alongside the communication capture, interface and workflow engine of a modern CRM

An example Dealmaker's OS platform constructed within Attio

It is materials rather than an answer, and that is the point of calling it an operating system. An operating system on its own does nothing much. Designed around how the firm actually works, fed by the right sources and connected to the things that already run the business, it becomes the place the work happens: the mandate, the buyer list, the introducer, the history and the next action all in one place, kept current without anyone being asked to keep it current - and close enough together that the link between two of them is obvious rather than lucky. The difference between those two outcomes is entirely in the design: what the entities are, how they relate, what happens automatically and what deliberately does not.

What it looks like built

Picture a firm a year or two past the decision.

The mandates, the buyers, the funds, the commitments and the people who made the introductions are all things in their own right, related the way the firm actually relates them. Email and calendar flow in without anyone being asked to file anything. Market information from outside arrives attached to the companies it concerns, rather than in a separate tab someone means to check.

A partner opening a target’s record sees more than its history. They see the ways in: who in the firm has dealt with whom, which of those relationships is warm and which has gone cold, what the firm concluded the last time it looked at this and whether that reasoning still holds. When something shifts - a leadership change, a raise, a filing - it reaches the person most likely to know what it means, with the context already attached.

The associate who joined in March asks better questions than their tenure should allow, because the firm’s memory is available to them rather than resident in a colleague who is in a meeting.

Built in that order - the shape, then the capture, then the connective work on top - what the firm ends up with is not a better record of the past. It is a firm that notices things: partners walking into meetings knowing what the firm knows, a pipeline that reflects reality without being maintained into shape, and an advantage that compounds with every conversation instead of leaving when someone does.

Why now?

Dealmaking has always rewarded the same thing. Not the firm with the most information, but the one that saw what the information meant: that these two parties belonged together, that the moment was now, that a business everyone else read as ordinary was worth a second look.

The information itself has never been cheaper or more evenly distributed. What only you know about who - the history, the context, the reason a particular introduction will work - cannot be bought, and until recently could not be held anywhere but in somebody’s head. That is what has changed. But it only accrues from the day something starts holding it. Every quarter a firm runs without that is a quarter of context that quietly fails to survive: the reason a buyer passed, the introduction that almost worked, the conversation that would have mattered eighteen months later. Nothing visible goes wrong, which is exactly why it gets deferred.

The usual objection is that this is a large undertaking. It is a good deal smaller than it was. Modelling the actual shape of a firm is now configuration rather than development, and it commits you to less, not more: a platform that changes as the firm changes, rather than a product whose roadmap you inherit and whose limits you discover in year three. Starting is a smaller decision than it looks, and it is the only version of this that compounds.

Do it and the firm’s knowledge stops being a personal asset and becomes a firm one. It remembers what it has learnt. New people arrive into a system rather than an oral tradition. And the firm gets better at what it is really for: standing between capital and value, and seeing before anyone else does where the two should meet.

—

The firms that start implementing their Dealmaker's OS propoerly now will know things in three years that their competitors cannot buy.