Perspectives · AI & Automation
The Fragmentation Gap Behind Your AI Tools' Blind Spots
Private markets firms are buying AI tools, but fragmented data keeps them blind. Why the access gap exists and how to make investment data AI-ready.
· 10 min read
You bought the AI tools. Your team has licenses for Copilot, maybe Claude, maybe ChatGPT. You have seen the demos. The productivity gains look real. But when someone on your deal team asks the AI to pull together portfolio company performance across your last three funds, or to summarize the capital call notices from the past quarter, the results come back incomplete, inconsistent, or just wrong.
The problem is not the AI. The problem is what the AI can see. And in most mid-market private markets firms, the AI cannot see most of your data.
This is the fragmentation gap. It is the gap between what AI tools could do for your firm and what they actually do, because your data lives in places the tools cannot reach. Capital call notices sit in PDFs. Borrower financials arrive in Excel attachments. Valuation reports come from half a dozen administrators in half a dozen formats. Your portfolio monitoring system does not talk to your accounting system, and neither one talks to the shared drive where your deal team keeps their diligence files.
The promise of AI in private markets is enormous. But that promise hits a wall when the models cannot pull from the data that matters most. This article breaks down why that gap exists, what it costs your firm, and why fixing the data layer, not adding more AI tools, is the real path forward.
What Data Fragmentation Looks Like at a Mid-Market PE or VC Firm
If you work at a private markets firm with under $10 billion in AUM, you already know this story. Your data does not arrive clean. It arrives as capital call notices in PDFs, financials in Excel, and valuation reports in formats that change from one portfolio company to the next. Every fund cycle brings new sources, new formats, and new manual work to stitch it all together.
The Financial Times put it bluntly: "Excel spreadsheets, email and PDFs are still the most-used tools in private markets," according to Oliver Albers, Chief Product Officer for Nasdaq's Capital Access Platforms division. "And that really hasn't changed in 10 years."1
This is not a fringe problem. Research from Mergermarket, sponsored by S&P Global, found that 77% of firms are using at least 50% more data sources than they did five years ago, with 37% saying their number of data sources has more than doubled.2 More sources means more formats, more reconciliation, and more places where data can fall through the cracks.
At a mid-market firm, the fragmentation problem has a specific shape. According to Preqin's Future of Alternatives 2028 report, the average PE or VC fund uses between 8 and 12 distinct software tools across the deal lifecycle, none of which were designed to work together.3 Your deal team tracks pipeline in one system. Your portfolio management team monitors company performance in another. Your fund admin sends reports in a third format. Your investor relations team keeps LP data somewhere else entirely. Each tool does its job reasonably well in isolation. The problem is the seams between them.
Consulting research on private equity technology stacks has found that junior deal team members spend roughly 40% of their time on manual "stare and compare" data reconciliation across systems, cross-referencing company data, validating ownership structures, and updating records by hand.4 The problem is not that individual tools are inadequate. The problem is that they were never designed to work together, and the investment workflow they collectively serve requires continuity that no combination of point solutions can provide on its own.
The result is a data landscape where no single system holds the full picture. When someone asks a question that spans funds, vintages, or portfolio companies, the answer requires pulling from multiple sources, reconciling differences, and hoping nothing was missed. That work is manual, time-consuming, and error-prone. And it is exactly the kind of work that AI should be able to help with, if it could only see the data.
Consider a concrete scenario. Your head of portfolio operations wants to understand exposure to a specific sector across your last two funds. The portfolio company names are stored differently in each system. One uses legal entity names, another uses trade names, and a third uses abbreviations that only the associate who set them up three years ago understands. Before any analysis can happen, someone has to manually map those entities, reconcile the financials, and build a view that should already exist. Now multiply that across every question your team asks in a given week. The hours add up fast.
Why AI Cannot Bridge the Gap on Its Own
Here is the hard truth that most AI vendors will not tell you: AI tools can only reason over the data they can access. When your data is fragmented across PDFs, spreadsheets, legacy systems, and disconnected tools, the AI simply cannot reach it. It does not matter how powerful the model is. If the data is not accessible, the AI is working blind.
McKinsey made this point directly in its research on generative AI in private markets. The firm found that 67% of investors believe gen AI will have a "transformational" impact on their business within five years, and 82% view its use as a high priority. But McKinsey also cautioned that "gen AI may accelerate the discovery of information, but it will not safeguard for quality without user guidance."5 In other words, speed without a trustworthy data foundation just gives you wrong answers faster.
The data backs this up. State Street's 2025 Private Markets Study surveyed nearly 500 senior executives at buyside investment institutions globally. They found that 83% of institutions recognize the value of generative AI and large language models for generating analyzable data from unstructured information. But when it came to actual implementation, the numbers tell a different story. Only 33% said they are using GenAI or LLMs to generate consistent, analyzable data from unstructured information, and just 8% had actually implemented these solutions in core workflows. A further 41% were still at the pre-investment planning stage.6
That gap between recognition and implementation is the fragmentation gap in action. Firms see the potential. They buy the tools. But the tools cannot reach the data because the data is not centralized, standardized, or governed. The AI is sitting on top of a fractured foundation, and the fractures show up in every query.
Think about what happens when someone on your team asks Copilot to summarize portfolio company revenue trends across your current fund. If the revenue data lives in five different Excel files, each formatted differently, with no common entity identifiers, the AI cannot pull it together. It might find one file. It might hallucinate a number based on partial data. Or it might just say it cannot find what you are looking for. None of those outcomes is useful.
The core issue is access. AI tools need structured, connected, governed data to produce reliable results. Fragmented data creates blind spots, and blind spots produce unreliable outputs. As one industry analysis put it, "if you don't control how documents, entities and relationships are structured, you're essentially renting insight from vendors and risking fragmentation."7
The Cost of the Access Gap
The fragmentation gap is not just a technical inconvenience. It has costs that compound over time.
Inconsistent and unreliable outputs
When AI tools pull from partial or fragmented data, they produce inconsistent answers. Ask the same question twice and you may get different results because the tool is reaching different subsets of your data each time. For a firm that needs to defend its NAV calculations or report to LPs, inconsistency is not acceptable. McKinsey's research on institutional investors identified three foundational steps for AI success, and the first two are establishing data governance and building a data platform that integrates data across sources into a single source of truth.8 Without that foundation, AI outputs cannot be trusted.
The AI illusion
There is a dangerous stage in AI adoption where early productivity gains create the appearance of progress while structural gaps quietly widen. One industry analysis of private equity firms described this as the point where "early productivity gains create the appearance of progress while the structural gap to Stage 3 and Stage 4 firms quietly widens."9 You get value from AI on the easy tasks, like drafting emails or summarizing a single document. But the hard tasks, the ones that would actually transform your operations, remain out of reach because the data foundation is not there.
Manual workarounds that do not scale
When AI cannot reach the data, your team falls back on manual processes. Someone exports data from three systems, reconciles it in a spreadsheet, and then feeds the cleaned version to the AI. That work is invisible in most firms but it eats hours every week. The MSCI 2026 General Partner Survey found that 42% of GPs still operate on partially connected tech stacks, and only 12% have achieved centralized data flow. Nearly half rely on manual or semi-manual processes.10 At a mid-market firm where headcount is lean, manual overhead is a direct drag on the work that matters most: sourcing deals, supporting portfolio companies, and managing investor relationships.
A widening competitive divide
The firms that fix their data foundation are pulling ahead. McKinsey's Global Private Markets Report 2026 noted that "alpha is less likely to emerge from market dynamics alone. Increasingly, it will be made," and identified how successfully participants "navigate and leverage AI" as a key differentiator.11 FTI Consulting's 2026 Private Equity AI Radar found that while 95% of PE funds say their AI initiatives are meeting or exceeding expectations, only 17% say they are significantly exceeding them. The difference came down to a subset of top-performing funds whose advantage was "not how much they invest in AI. It is where and how they deploy it."12 The firms winning with AI are the ones that put the data foundation first.
For a mid-market private markets firm, the stakes are specific. You are competing for deals and LP capital against firms that may have larger teams and bigger technology budgets. AI should be your lever to do more with the team you have. But that leverage only works if the AI can see your data. Right now, for most firms, it cannot.
The Real Fix: Making Your Data AI-Ready
The bottleneck is data accessibility, not AI capabilities. You don’t need a bigger model or a different AI tool. You need your data centralized, standardized, and quality-assured, so whatever AI you choose can actually reach it and produce results you can trust.
This is where Aquata can help. The Aquata Investment Data Platform is built to transform disjointed data into action. Aquata centralizes and standardizes all investment data, including unstructured data from PDFs, Excel and email.
While you consider Aquata, there are practical ways to make your data more AI-ready. These are makeshift solutions that you can implement today with tools you already have. They require work, but they can offer immediate relief and will only accelerate results from a solution like Aquata.
Our next article, Why Your AI Tools Can't See Half Your Data (And What to Do About It), walks you through these stopgap solutions.
