Almost every conversation we have with a general partner about artificial intelligence begins with a queue. A data room nobody has finished reading. A diligence pack that took three weeks to summarise and was stale by the time it landed. An LP request list that somebody junior is still working through at nine in the evening in the third week of January. The implied argument is that if the queue moved faster, the firm would be better.
We think that argument is incomplete in a way that matters commercially. In most private capital firms the binding constraint is not the volume of work produced. It is the number of decisions the firm can take, with genuine conviction, in a given period. Capacity and decision throughput are related, but they are not the same variable, and they do not respond to the same intervention.
Adding capacity to a firm whose decision process is the real constraint does not generate more good decisions. It generates more material in front of the same committee, on the same calendar, with the same people carrying the same unresolved disagreements. That is not neutral. It is usually worse than doing nothing.
Analyst hours were never what stopped the deal
Consider how a deal that should have happened actually fails. In our experience it is rarely because nobody read the quality of earnings report. It fails because the sponsor came back with a revised structure on a Thursday and the partner who had the relationship was in another process until the following Wednesday. It fails because two partners had a real and unspoken disagreement about the sector thesis, and the way that disagreement resolved itself was through delay rather than through debate. It fails because the firm required a level of certainty on one variable that nobody had authorised anyone to waive.
None of those are reading problems. They are questions of authority, sequencing and what the firm has decided in advance that it is willing to be uncertain about. A team with twice the analytical output would have failed in exactly the same way, slightly better informed.
The same pattern shows up on the investor relations side. When a fundraise stalls, the cause is very seldom that the firm could not produce material. Most firms can produce material. The cause is that the firm could not decide, quickly, what it was prepared to concede on fees, on co-investment rights, on an LPAC seat, or on a side letter clause that touches every other investor in the fund. Those decisions sit with a small number of people who are already the busiest people in the organisation.
More input into an unchanged decision produces noise, not conviction
There is a specific failure mode we now see often enough to name it. A firm deploys something capable against diligence or screening. It works. The volume of analysis rises sharply. Investment committee papers grow from twelve pages to forty. Screening memos that used to cover eight opportunities a month now cover thirty.
The committee still meets for two hours a fortnight. The partners still read the papers on the same train journey they always did. What changes is not the quality of the decision, it is the distribution of attention: more pages read less carefully, more flagged risks that nobody has ranked, more opportunities surfaced than the firm has the bandwidth to pursue properly. The firm has manufactured optionality it cannot exercise, and optionality you cannot exercise is just cost and distraction.
Worse, the additional material has a peculiar effect on conviction. When an analysis identifies forty considerations rather than twelve, and nothing in the process ranks them or says which ones could change the answer, the reasonable response of a careful fiduciary is to become more hesitant, not less. We have watched firms become measurably slower to commit after a successful capacity deployment. Nobody had done anything wrong. The decision process simply was not built to absorb what was now arriving at it.
Where capacity genuinely is the binding constraint
We should concede the strongest version of the counter-argument, because it is a good one and it holds in specific places.
In diligence-heavy commercial real estate, capacity is frequently the actual constraint and not a proxy for one. A portfolio acquisition with three hundred leases has a genuine floor of reading that has to happen, and that floor has historically forced real trade-offs: sampling rather than reviewing, accepting a rent roll at face value, scoping out a category of estoppel review because the timetable would not carry it. When the cost of complete review falls, that is not more noise. It is the removal of a compromise the team never wanted to make. The same applies to complex carve-outs and to portfolios where the risk genuinely lives in the tail of the documents rather than the summary.
Investor relations peak load is the other honest case. The weeks after a quarter end are not a decision-throughput problem. They are a queue with a hard deadline, staffed by people who cannot be hired seasonally because the work requires knowing the fund. Capacity applied there converts directly into fewer errors and less attrition among exactly the people a firm cannot easily replace. We would not talk a client out of that.
The distinction is this. Capacity pays when the work has a defined floor and a fixed deadline. It does not pay, and can cost, when the output feeds a judgement that a small number of senior people have to make and the design of that judgement is unchanged.
Decision design is a set of unglamorous, specific choices
Redesigning how decisions get made sounds abstract. In practice it is a short list of concrete questions that most firms have never answered explicitly, because the answers were implicit in who happened to be in the room.
What can be decided without the full committee, and by whom, up to what size?
Which two or three variables would actually change the answer on this deal, decided before the work starts rather than after?
What level of uncertainty on each of those variables is the firm willing to accept, stated as a number rather than a feeling?
Who is empowered to kill an opportunity early, and is that person rewarded or quietly penalised for doing so?
Answering those changes what analysis is worth producing. A firm that has decided in advance which three variables are decisive can direct enormous analytical capacity at those three variables, at depth no team could previously reach, and deliberately decline to produce the other thirty-seven pages. That is a different use of the same technology, and it produces conviction rather than hesitancy, because the output maps onto a decision somebody has already agreed how to take.
It also changes what a firm should measure. Documents processed and hours saved are inputs. The numbers that matter are the time from first look to a firm answer either way, the proportion of opportunities killed in the first fortnight rather than the second month, and how often the committee defers a decision for want of information that was always going to be unobtainable.
The compounding sits with the firms that redesign first
To put rough numbers on the difference, and these are illustrative rather than measured, take a mid-market fund that looks at four hundred opportunities a year, takes thirty to committee and closes eight. Doubling analytical capacity in the traditional shape might take the firm to six hundred looks and thirty-two committee papers, with the same eight closings and a more tired team. Redesigning the decision so that kill decisions happen in week two and depth is concentrated on the variables that decide the outcome might take the same firm to four hundred looks, forty committee papers of half the length, and ten closings with better-understood risk. The second firm did less reading. It made more decisions.
The reason this compounds is that speed and clarity change what gets shown to you. Sponsors and intermediaries route opportunities to firms that answer quickly and do not waste a management team's time. A reputation for a clean, fast no is worth something real in deal flow terms, and it is downstream of decision design rather than of analytical capacity.
Capacity is the entry fee, decision design is the return
We are not arguing that firms should be slower to adopt. Capacity is becoming genuinely inexpensive, and a firm that declines it will in a few years be paying people to do work that its competitors no longer pay for. That is not a defensible position, and the concessions in commercial real estate diligence and in investor relations peak load are real gains available now.
What we are arguing is that capacity is the entry fee rather than the return. It gets a firm to parity with everyone else who bought the same thing, which within a few years will be everyone. The return sits with the firms willing to do the harder and less purchasable work of examining how their own decisions actually get taken, where authority sits, what they have agreed to be uncertain about, and which questions were worth asking in the first place.
That work does not require a procurement process. It requires a few honest hours among the partners who already know where their decisions get stuck. The encouraging part is that the firms doing it tend to find that the constraint, once named, turns out to be far more movable than they assumed.
