From mandate to buyer: what AI can and cannot do in corporate finance

Corporate finance is a memory business dressed as a deal business. Ask any partner how the last sale actually found its buyer, and the story is rarely the long list. It is the remembered conversation: the strategic who mentioned an appetite two years ago, the fund that lost the last auction in this space and still wants in, the founder someone met at a closing dinner. The deal went to the firm that remembered.

So when AI arrives promising to transform M&A, the right question is not whether it can generate a buyers list. It is whether it can make a firm's remembering better. Those are very different products.

What the generic version gets wrong

Database-driven buyer search already exists and every firm uses it. Adding a model on top produces longer lists faster, and longer lists were never the constraint. Every advisor in the market can name the fifty obvious buyers for a given asset; the mandate is won by the three unobvious ones and by the credibility of knowing which fifty to leave out. That discrimination comes from what the firm has learned across every process it has run: who bid and why, who withdrew and why, what a buyer said in a management meeting three years ago. No external database holds any of that, because it is nobody else's to hold.

Where AI genuinely moves the needle

The mandate's first week. Teasers, information memoranda and process letters are assembled from the firm's own precedents against the current numbers. Structured drafting from your own documents, with every figure traced to its source, takes real days out of the most compressed phase of a process without touching a single judgement about positioning.

Matching on substance. A mandate's characteristics held against everything the firm has recorded: past processes, bidder behaviour, stated appetites, relationships and their history. The output is not a list but a set of arguments: this buyer, because of this history, with this caveat. The unobvious candidates surface because the firm did, in fact, once know them, in a mailbox nobody thought to search.

The process under control. A live transaction is hundreds of parallel threads: NDAs at different stages, data-room permissions, bidder questions, deadlines. Work that is administrative in nature and catastrophic when dropped. A system that tracks every thread and flags what stalls is unglamorous, and unglamorous is where hours actually go.

What stays human, permanently

The call on price and the read of a negotiation. The moment of advising a founder to take the lower certain offer over the higher conditional one. The judgement that a strategic's interest is tactical intelligence-gathering rather than intent. These are made of things models do not hold: responsibility, relationships and the willingness to be accountable for advice. Any system in this business must present its work as preparation for those judgements, with reasoning attached, and leave the judgements where the liability lives, with the people.

The confidentiality bar

An M&A firm's records are other people's secrets under mandate: who considered selling, who looked at what, what a business was nearly worth. The dedication test is absolute here. One firm, one system, one memory; no pooling, no cross-customer learning, a full log of every access, and a contractual exit with everything returned. A vendor whose model improves from your processes is redistributing your mandate history to the market, and should be shown the door with the courtesy that deserves.

Where we stand

We build this as UNIVERSE, the Corporate Finance Brain: the firm's own process history made present, mandates matched on substance with the arguments attached, documents drafted from the firm's own precedents, and every judgement left with the partners. Nothing pooled, everything logged, shadow before live. The full shape is on the page for UNIVERSE, the Corporate Finance Brain.