Build or buy deal tracking software? The AI case for biotech BD

Deal tracking tools do not fail because they are badly designed. They fail because a form only knows what somebody typed into it, and in a twelve person company nobody has time to type.

Business development at a small biotech does not fall over on strategy. It falls over in the gaps. A term sheet comes back with a changed royalty and it sits in one inbox. A partner asks for the Phase 1 safety package and it is assumed to have gone. A confidentiality agreement runs out three weeks before the data room is due to close. None of that is difficult. It is just spread across mail, calendar, decks and one person's memory, and it is exactly the ground the wider build vs buy software decision is fought on.

What buying deal tracking software gets you

A pipeline tool gives you a shared vocabulary, a stage model, and a place a board member can look without asking you. Those are real, and if you have a BD team of five with a sales operations person feeding it, buy one.

Below that headcount the economics invert. The tool is a form, somebody has to fill it in, and that somebody is the person who least has the time. So the pipeline goes stale, and a stale tracker is worse than no tracker, because people act on it. The same trap governs the biotech CRM decision, which is the same argument with the contact record swapped in for the deal record.

What building it yourself gets you

A capable person with a coding agent can build a serviceable deal tracker in a couple of evenings. It will fit your stage names, it will not cost a subscription, and it will do exactly what you asked. That is a genuinely reasonable answer, and some companies should take it.

Two things tend to get scoped out. The first is that the build captures state and not upkeep, so you have moved the typing from a vendor's form into your own form. The second is the second year: whoever built it now owns it, and they were hired to do something else. The wider version of that trade, and the five questions that actually settle it, is internal AI tool vs vendor.

The third shape: the deal record writes itself

There is an option that is neither a purchased form nor a home-made one. Instead of you reporting into a system, the system reads what already happened and writes the record. It is one seat rather than a shared pipeline the commercial team logs into, and the records are plain files on that person's own machine. The direction of travel is reversed, and that single change is what decides whether a tracker is still true in March.

In practice that means four things. Deals update themselves, because every email and call moves the record and drafts your next step, queued for you and never sent on its own. There is one true status per deal, so the board update and the tracker read the same file and cannot disagree. Nothing you are owed goes cold, because a promise made out loud on a call is held and surfaced when it slips. And the dates that actually hurt are caught: cost claims, confidentiality agreement expiries, notice periods, each with a heads-up scheduled before it comes due rather than on the day. What it does not do is value the deal or model the royalty. It holds the record and the dates around them.

What this looks like on a real deal

A partner's business development lead writes to say their committee slot moved forward, so the round two diligence answers are needed two weeks earlier, same fourteen questions, no new scope. That one message is not one change. The answer date moves. The data room loading window moves with it, because the answers have to be in before the committee reads them. The internal review of the answer pack moves forward, which means your clinical, chemistry and regulatory leads need to know sooner than they were told. And a reply needs writing that confirms the new date without conceding scope.

A form will record the new date if you remember to open it. What actually needs to happen is that the record moves, the downstream dates move with it, the open items close and reopen, and the reply is drafted and waiting for your approval. That is the shape of the work described in AI for business development in biotech, and it is the reason the useful comparison is not tool against tool.