Coverage is not access. As AI narrows the information gap in private markets, advantage moves to what your firm knows but never recorded.
By Ismail Badereldine, Chief Executive, FinBursa
 |
The most consequential finding in this year’s private markets research has little to do with efficiency. In its 2026 Global M&A Outlook, KPMG observes that as analytical capacity expands across the industry, informational asymmetry narrows, and execution gaps surface earlier in the process.
That deserves more attention than it has had, because it describes something other than a productivity gain. If analysis becomes cheaper and better for everyone at once, then analysis is a weaker source of differentiation than it used to be. Advantage has to come from whatever fails to commoditise at the same rate. Most current spending assumes the reverse.
Adoption is broad, reported returns are not
Grant Thornton’s 2026 AI Impact Survey, in its private equity analysis, reports that 46% of private equity leaders say they are scaling AI across functions, and 80% say they are exploring or piloting agentic AI. Within the same responses, 24% report revenue growth from AI. Nine percent say they are very confident they could pass an independent AI governance audit within 90 days, against 22% across industries. Forty-five percent describe themselves as still piloting, eleven points above the cross-industry rate. Grant Thornton’s own framing is that private equity is seeing AI activity but not returns.
Those are self-reported perceptions, drawn from nearly 1,000 senior US business leaders surveyed in early2026, and Grant Thornton states that reported correlations should not be read as evidence of causation. Taken as directional rather than definitive, they still describe an industry that has bought more than it has demonstrated.
KPMG’s data points the same way from a different sample: 700 M&A decision makers, 519 corporate and181 from private equity firms, across 20 countries and ten sectors, fielded between 19 December 2025 and27 January 2026. In competitive intelligence and market analysis, which KPMG identifies as a leading use case, 66% of organisations report at least early efficiency gains. KPMG characterises most of those gains as incremental rather than transformative, with fewer than one in four organisations reporting significant efficiency improvements. In valuation modelling and scenario planning, 17% report gains above 25% and3% above 50%.
Two samples, two different questions, one shape of answer.
The convenient explanation is unavailable
You could conclude that firms have simply pointed the technology at the wrong stage of the deal. KPMG closes that off. It reports adoption already concentrated in the early lifecycle, supporting screening, market analysis and risk identification, with agentic deployment running at 56% in due diligence and valuation,53% in deal sourcing and strategy, 45% in post-merger work, 40% in execution.
The front end is where this technology has already been aimed. Moderate returns there are not a story about neglect.
What survives commoditisation
KPMG decomposes deal work into a repeating loop: reading, thinking, writing, verifying. Its assessment is that AI now handles substantial portions of reading and writing, and increasingly of verification, while thinking, meaning the application of judgment, context and experience, remains decisively human.
Put that next to the narrowing asymmetry point. The parts being automated are the parts every competitor is also automating. As those converge, position depends on whatever is not converging.
Which raises an obvious question: what exactly is the non-converging part? “Judgment” is a large word. It could mean sector expertise, operating capability, pattern recognition built over decades, or the quality of a firm’s relationships. Any of those could be the answer, and a piece like this should not pretend the evidence picks one cleanly.
But KPMG does say something specific about what blocks progress, and it is not a shortage of judgment. In early deployments, it finds the technology is rarely the bottleneck. The harder problem is codifying the unwritten rules, institutional shortcuts and tacit knowledge that experienced practitioners carry in their heads. Deal processes lean on incomplete information, undocumented precedent, and judgment embedded in individuals rather than systems. KPMG’s conclusion is that firms capturing the most value will be the ones that make their processes legible to machines.
When KPMG lists the categories of analysis that become viable as the marginal cost of structured work falls, one of the four is mining a firm’s own transaction history for precedent terms, playbooks and lessons learned. It describes this as institutional memory that tends to dissipate when deal teams rotate.
So the constraint identified is not a deficit of judgment. It is that the judgment already exercised was neverrecorded anywhere the firm can reach.
The same conclusion from the sourcing side
Bain arrives nearby by another route. In its 2026 Global Private Equity Report it observes that the averagedeal process now involves dozens of private equity firms examining the same deal book and the samemarket information. What proactive firms do differently, in Bain’s account, is not to search harder. Theyidentify the companies they are uniquely suited to own and track them over years rather than months, wellbefore any indication the asset will trade.
Consider what that means in practice. (The following is illustrative rather than drawn from any particularfirm.) A partner meets a founder at a conference in 2022. Nothing is for sale. Over four years there areperhaps nine conversations: a view on a competitor’s pricing, an introduction to a CFO candidate, a frankexchange about why the founder will never sell to a strategic buyer, an offhand comment about aco-investor she would refuse to work with again. None of it is written down anywhere except in fragmentsacross the partner’s inbox and his own recollection. In 2026 the company runs a process, and the partnerhas moved to another firm. His replacement inherits a name in a CRM, a logo, and a stale note. Everycompetitor in that process now has the same starting position, and the four years bought nothing.
A reasonable objection: anyone can pick up the phone. In a broad auction they will, and there therelationship advantage barely exists to lose. The argument holds at the proprietary and semi-proprietary end of the market, where a founder takes one call before deciding whether to run a process at all, or acorporate sounds out two buyers it already trusts. Bain’s own observation about dozens of firms readingthe same book suggests a good deal of the market is intermediated, and for those processes coverage maybe all anyone needs.
Where it does hold, the difference is less about reachability than about what a conversation carries.Anyone who calls receives the deck. Four years of relationship produce the material that never enters adeck: why the last process collapsed, what the founder wants after close, which number in her own modelshe does not believe.
If analytical capability converges across the industry while that context stays trapped in individuals, thething that was supposed to be proprietary becomes the thing least likely to survive.
The strongest objection
The serious counterargument is that all of this is early.
Agentic systems have been running in deal workflows for quarters, not years. Moderate gains are what theearly diffusion of a general purpose technology would be expected to look like, well before its eventualimpact is visible. The KPMG sample also skews toward large organisations, and self-reported perceptiontends to miss diffuse productivity effects that never appear as a line item. On that reading, “fewer than onein four report significant gains” means only: wait.
Much of that is right. Notice, though, that it argues for a higher ceiling rather than against the direction oftravel. If capability improves for everyone, asymmetry narrows further and the pressure onnon-commoditising advantages rises. There is also a data point pulling the same way, though a soft one:Grant Thornton finds organisations with fully integrated AI roughly four times more likely to reportAI-driven revenue growth than those still piloting, 58% against 15%. That figure is cross-industry andcorrelational, and Grant Thornton warns against reading causation into it. It is at least consistent withintegration rather than adoption being where the difference sits.
Where this leads
If asymmetry is narrowing, and the binding constraint is knowledge nobody wrote down, then thenearer-term contribution of AI to origination looks less like prediction and more like memory. Holdingcontext across hundreds of counterparties over multi-year horizons. Ensuring a relationship built by oneperson survives that person changing seat. Making the state of a long conversation available to a teamrather than to an inbox.
Unglamorous work, and precisely what Bain’s account of proactive sourcing assumes a firm can alreadydo.
What would change this view
Independent outcome data, not published by anyone selling software, linking AI-assisted pipelines torealised results: entry multiples, conversion rates, return dispersion against a comparable baseline. As faras I can establish, none is public. Most of the striking performance figures in circulation come fromvendors, which does not make them false but does mean they cannot carry an argument.
Three questions worth asking
1. Does this give me information my competitors will not also have? Where the underlying data is ashared subscription, the answer is no, whatever sits on top of it.
2. Which part of the work am I buying? Reading, writing and checking are converging across theindustry. Judgment is not for sale.
3. If a partner resigned tomorrow, what would the firm still know about the relationships built over thelast four years?
The last one gets asked least, and on the available evidence it is where the technology is most likely toearn its cost.
The author is chief executive of a company that builds relationship and workflow infrastructure for private-market firms, andtherefore has a commercial interest in the conclusion argued above. The author’s firm does not act as a broker, advisor, placementagent or intermediary. The evidence is cited so readers can weigh it independently.
This article is published for general information and discussion. It does not constitute investment, legal, tax or professional advice,and it is not an offer, solicitation or recommendation in respect of any security, fund or transaction. Views expressed are theauthor’s own. Third-party survey findings are reported as published by their authors and have not been independently verified.
Sources: KPMG International, 2026 Global M&A Outlook, March 2026 (700 M&A decision makers; 519 corporate, 181 privateequity; 20 countries and ten sectors; fielded 19 December 2025 to 27 January 2026). Grant Thornton, 2026 AI Impact Survey andits private equity analysis (nearly 1,000 senior US business leaders, early 2026; self-reported, non-causal). Bain & Company,Global Private Equity Report 2026, 22 February 2026.
