Somewhere in the 2010s, “best-of-breed” stopped being a strategy and became a reflex in private equity, venture capital, family offices, everywhere capital gets allocated. Best CRM here, best data room there, best portfolio tool somewhere else, and a Zapier subscription praying in the middle. The suite vendors were cast as the enterprise dinosaurs, selling eleven modules where three were great, five were fine, and three existed mainly so the sales deck could say “end-to-end.”
 |
To be fair, the reflex was earned. Anyone who lived through an ERP rollout in the 2000shas scar tissue. A mediocre module inside a suite is a tax paid forever, and no singlevendor is excellent at everything. So the sophisticated buyer bought pieces andstitched.
That era appears to be ending, and at an awkward moment: global private markets have grown into a roughly $15 trillion industry, with credible forecasts pointing toward $25trillion by the end of the decade. The asset class is scaling far faster than the tooling underneath it. Meanwhile, AI has quietly shifted the economics of best-of-breed from two directions at once. One shift is well understood by now. The other, the industry is only starting to say out loud.
The first shift: context
This one is increasingly familiar, so briefly.
An AI system is only as smart as what it can see. An AI that reads a data room is useful.An AI that reads the data room, the deal flow pipeline, the LP and investor relationshistory, and the portfolio monitoring data isn’t 20% better. It’s a different animal. It cananswer the questions that actually matter to a fund: which LPs would want this deal,based on everything the firm knows about them? How does this company compare tothe last dozen that went through due diligence in the sector?
Now consider a typical fragmented stack. Five vendors, five copilots, each seeing a fifthof the picture. Five smart analysts who have never met each other, each answeringconfidently from partial evidence. Anyone who has managed people knows how thatmovie ends.
In private markets there is a nastier version of this problem: permissions. The industry runs on need-to-know. Stretch an AI layer across five vendors and five entitlement systems must be reconciled before any general counsel signs off. On a unified platform, the AI inherits one permission model. In an NDA-drenched business, that is frequently the difference between an AI that gets deployed and an AI that dies in legal review.
So far, so familiar. The second shift is the more interesting one.
The second shift: the moat itself
What did a best-of-breed vendor actually sell all these years? Strip away the branding and it was mostly this: years of accumulated workflow logic. Thousands of small automations across deal sourcing, due diligence, fund administration, investor onboarding, and LP reporting. Edge cases handled, integrations hardened, buttons in the right places. That depth took a decade to build, and it was real. It was the moat. It was why the point solution beat the suite’s module every time.
AI agents are draining that moat, and faster than most vendors want to admit.
A workflow that took three years of product iteration to encode (parse this document,extract these fields, chase these signatures, reconcile these records, notify thesepeople) is increasingly something an AI agent performs from a description of the task.Not all of it. Not the genuinely hard parts. But a large share of what made point solutionsfeel indispensable was never deep IP; it was patient accumulation of shallow logic. Andpatiently accumulated shallow logic is exactly what agents are good at replacing.
This flips the equation in a way that is a little brutal. The suite’s historic weakness was that its individual modules lagged the specialists on features. But if an AI layer can generate much of that long-tail functionality on demand, feature depth stops being the scarce asset. The scarce asset becomes the thing agents can’t conjure: unified, permissioned, historical data, and the trust infrastructure around it. Suites happen to own that. Point solutions, mostly, don’t.
A note of caution, because the claim is easy to overstate. Agents today are inconsistent, they need supervision, and “the AI can just do that now” is doing a lot of unpaid labor ina lot of pitch decks across this industry. But the direction is not ambiguous, and buyers signing three-year contracts should price the direction, not just the snapshot.
The strongest objection, taken seriously
There is a credible rebuttal to all of this, and it deserves a fair hearing: interoperability .Open agent protocols and standardized connectors are maturing quickly, and in theory they let best-of-breed tools share context with each other. If the CRM, the data room ,and the portfolio system can all expose their data to one orchestrating agent, the fragmented stack gets its unified brain back without anyone buying a suite. Best of both worlds.
In most industries, that rebuttal will eventually win real ground. In private markets, it runs into three walls. First, permissions: a connector can move data, but it cannot reconcile five vendors’ entitlement models into one coherent need-to-know policy, and in a business governed by NDAs and side letters, moving data without moving its restrictions is not integration, it is a breach with good UX. Second, incentives: each point vendor’s enterprise value rests on owning its slice of data, and vendors do not enthusiastically build pipes that commoditize themselves. Third, liability: when an orchestrating agent hallucinates across five systems, five contracts point at each other. Interoperability is real and worth watching. But in a confidentiality-bound industry, “technically connectable” and “deployable past a general counsel” remain very different standards.
Where best-of-breed still wins
An honest assessment has to argue the other side properly, because the answer is notone-sided.
When a single workflow is the firm’s actual edge, say a private credit fund whose alpha lives in a proprietary analytics pipeline, no suite module should ever sit there. Buy the suite for commodity workflows; protect the differentiation with the best specialized tool available, or build it in-house.
Scale changes the answer too. A mega-fund or sovereign wealth institution with two hundred engineers can be its own integrator: buy the pieces, build the connective tissue, capture the context advantage without the vendor dependence. At the other end, a two-person VC fund needs a data room and a spreadsheet, and honestly, godspeed. The suite’s real territory is the enormous middle: mid-market private equity firms, growth-stage venture capital funds, multi-family offices. Too big for duct tape, too small to employ a platform team.
Then there is the fake suite problem, which buyers should be ruthless about. Plenty of“suites” are three acquisitions stapled together: three codebases, three data models,one invoice. That is best-of-breed’s weaknesses wearing a suite’s price tag, and everyargument in this article collapses for them. The test worth putting to any vendor,without exception: show how a permission set in one module changes the AI’s answersin another. Live, in the demo. If they can’t, the “platform” is a brochure.
And a firm in the middle of a fundraise should not rip out five systems because a blogpost said the physics changed. Timing is a legitimate reason to stay fragmented foranother cycle. Software strategy that ignores the calendar is just theory.
Different markets, different instincts
The decision also splits along geographic lines that are visible to anyone selling into all three regions.
American PE and VC firms lean best-of-breed almost culturally: empowered teams, expensed tools, fragmentation tolerated as the price of speed. European institutions weight governance and vendor risk, and after GDPR they increasingly want architectures where data residency and access control can be explained in one meeting, which pushes toward platforms. And in the Gulf, something else is happening: institutions with no twenty-year legacy stack to defend are skipping the fragmentation phase entirely and adopting unified, AI-native platforms directly. It’s the same leap frog that let these markets skip checkbooks and jump to mobile payments. Some of the boldest platform adoptions in private markets right now are happening in Riyadh and Abu Dhabi, not New York. That still surprises people. It shouldn’t.
None of these instincts is wrong. They optimize different things: speed, governance, trajectory.
A checklist for the skeptical COO
Stripped of vendor noise, the decision reduces to four checks.
Write down the five questions the firm most wishes it could answer instantly. Ifanswering them requires data from more than two systems, fragmentation is costingmore than the line items show.
Name the one workflow that is genuinely the firm’s edge. Never let a suite module sitthere.
For any suite under evaluation, apply the unity test above: one data model or a stapledbundle.
And put a slightly cruel question to each incumbent vendor (the CRM, the data room, the portfolio management tool): how much of what this contract pays for could an agent do next year? The answer matters less than the length of the pause.
The disclosure
Full transparency: The author runs a company that builds a unified, AI-native platform for private markets, which makes him precisely the person you would expect to write this article. Discount accordingly. But the argument never asked for trust; it asked for a test. Firms whose five specialists are genuinely talking to each other should keep them. Firms whose specialists have never met might consider introducing them to one system that has read everything.
Ismail Badereldine is the CEO of FinBursa, the AI-native platform for private markets.
This article reflects the author’s views on industry trends and is provided for informational purposes only. It does not constitute investment, legal, or tax advice. AI generated analysis, including outputs from any software platform, can contain errors and should always be reviewed by qualified professionals; investment decisions shouldnever be based on AI outputs alone
