When to Move Your Data Out of Spreadsheets?
Yuri Pontes | Jun 10, 2026
Somewhere in a boardroom right now, a private equity Operating Partner is presenting a slide deck with the words “AI-Driven Value Creation” on the cover. The financial model shows 200 to 400 basis points of EBITDA expansion. The hold period is 36 months. The investment committee is nodding.
Forty-eight hours later, that same Operating Partner walks into a portfolio company and sits down with the VP of Technology. The conversation that follows is the moment everything either starts moving or quietly stalls.
What the Operating Partner finds, more often than not, is a company running on a 12-year-old ERP system, customer data split across three platforms that don’t talk to each other, and an engineering team that has never built an AI application in production. The financial model looked clean. The infrastructure underneath it does not.
This is the gap that is defining private equity’s AI moment in 2026 is the implementation gap.
What the data says about AI readiness in PE-Backed companies
The PE industry’s commitment to AI is not hype. It is board-level doctrine backed by real financial data:
| 95% | of PE firms have begun or plan to implement agentic AI in their operations in 2026. |
| 200–400bps | EBITDA expansion achieved by PE funds that deploy AI correctly across portfolio companies within 12 months. |
| 58% | of PE-backed companies have no formal AI strategy at the point of acquisition. |
| 72% | of PE Operating Partners now consider IT modernization a top-three lever for value creation — up from 49% three years ago. |
Read those numbers together and a picture emerges. The ambition is nearly universal. The mandate is backed by financial evidence. And the majority of portfolio companies arrive at the starting line without a strategy, without modern infrastructure, and without the engineering capability to build what the operating model now demands.
“Most funds get AI wrong because they treat it as 25 separate technology projects instead of one operating model.”
The failure mode is consistent. A PE firm acquires a company, installs a value creation plan, and tasks the portfolio company CTO with “implementing AI.” What follows is a series of disconnected pilots: a chatbot for customer service, an automation tool for accounts payable, a dashboard built on top of a data warehouse that was never properly designed.
Eighteen months later, the board asks why EBITDA hasn’t moved. The answer is always the same: the initiatives never added up to an operating model.
Why standard AI tools fail mid-market portfolio companies
This is the year that every major technology company launched an AI product for business. Anthropic launched Claude for Small Business with integrations for QuickBooks, HubSpot, and DocuSign. Microsoft deepened Copilot across the Office suite. Salesforce shipped Einstein across its CRM platform. Bain partnered with OpenAI to build an enterprise deployment practice targeting Fortune 500 companies.
All of this is real and all of it matters. But none of it solves the problem facing a $150M specialty chemical distributor running on a 2008 ERP system with customer data in three separate spreadsheets and a field service team logging work orders in a system that hasn’t been updated since 2015.
A DocuSign integration does not fix fragmented data architecture. A CRM AI layer does not replace a warehouse management system that was never built for predictive inventory. A QuickBooks connection does not modernize a 50-location distribution network’s supply chain intelligence.
The companies that cannot solve their data fragmentation, compliance requirements, legacy infrastructure, and workflow complexity with a toggle are precisely the mid-market businesses PE firms are acquiring and trying to modernize. That gap is where the real implementation work happens.
There are three categories of solution in the AI market right now:
- Enterprise consulting firms (Bain, McKinsey, Accenture) – capable of deep transformation work but priced for Fortune 500 engagements, not $100M to $400M portfolio companies.
- Off-the-shelf AI tools (Anthropic, Microsoft, Salesforce) – designed for standardized workflows in uncomplicated environments, not regulated industries with legacy infrastructure.
- Custom engineering firms with vertical depth – the category that can actually do the work for companies in the middle. This is the most underserved segment in the market.
Mid-market PE portfolio companies live squarely in that third category. They need custom-engineered solutions built by people who understand their industry, their compliance environment, and their data reality.
Read more: How to Use AI Classification to Build More Efficient Apps
The five conditions for AI readiness in a PE portfolio company
After 12 years of building technology in regulated, data-heavy industries, like payments, financial services, energy, logistics, industrial operations, we have learned that AI readiness is not about having the right tools. It is about having the right infrastructure underneath them.
Before any portfolio company can capture the EBITDA expansion the model promises, five things need to be true:
1. Data is centralized, clean, and queryable.
Not scattered across three platforms and a folder of exports. AI systems are only as reliable as the data underneath them. Fragmented data doesn’t just slow down implementation, and it produces confident answers built on wrong inputs. Fix the foundation first.
2. Core workflows are mapped. Not assumed.
The way work actually moves through the business, every handoff, every exception, every workaround, is documented before a single model is deployed. AI applied to an unmapped workflow doesn’t optimize it. It automates the dysfunction.
3. Legacy systems are assessed, not ignored.
The ERP from 2008 is either integrated, modernized, or replaced. Working around it is not a strategy — it is a debt that compounds with every new layer added on top. The assessment has to be honest about what the current architecture can and cannot support.
4. Technology infrastructure can handle what is being asked of it.
APIs exist. Data pipelines are built. The architecture is stress-tested before deployment, not after. Most AI initiatives stall not because the model failed, but because the infrastructure was never built to run it in production.
5. The organization knows what it is building and in what order.
Not 25 pilots. One operating model, sequenced by impact and dependency. The roadmap is specific enough that the delivery team can execute it and the board can measure it.
The companies that get this right do not start with AI. They start with the diagnostic: an honest assessment of where they are, what is broken, and what needs to happen before the AI initiative can produce anything real. The value creation plan follows from that assessment, not the other way around.
This is what the 100-day plan should actually contain. Not a slide deck about AI strategy. A structured diagnosis of data quality, technology infrastructure, workflow complexity, and organizational readiness, with a sequenced roadmap that the delivery team can execute and the board can measure.
Read more: Conversational AI: How to Design and Build Scalable AI-Powered Features
Why the first 18 months after acquisition define the AI outcome
One more thing worth saying plainly: timing matters enormously in PE-backed technology modernization.
The first 12 to 18 months after acquisition are when the operating model is being built. The 100-day plan is being executed. The value creation roadmap is being sequenced. Operating Partners are making vendor decisions. Technology partners are being evaluated and selected.
After that window closes, incumbents are in place. Contracts are signed. The modernization roadmap is locked. Getting in at month 24 is structurally harder than getting in at month three, not because the work is different, but because the organizational momentum is already pointing in a different direction.
The best time to address AI readiness in a PE-backed company is the moment after the acquisition closes. The second best time is now.
For PE firms actively managing portfolio companies that were acquired in the last 12 to 24 months: the implementation window is open. The question is whether the right engineering partner is in place to actually close the gap between what the model projects and what the infrastructure can currently support.
For Operating Partners building a repeatable value creation framework across multiple portfolio companies: the AI Readiness Assessment, a structured 2 to 4 week diagnostic of data quality, technology infrastructure, workflow complexity, and organizational readiness, is the instrument that makes that framework repeatable.
Closing the implementation gap: what the right engineering partner does
Private equity is not wrong to bet on AI. The financial evidence is real. The EBITDA case is documented. The urgency is justified.
But the bet only pays off if the implementation is right. And implementation, the actual work of diagnosing data, modernizing systems, mapping workflows, building pipelines, and delivering AI applications that run in production, is not something a strategy deck or an off-the-shelf tool can provide.
It requires engineers who have done this before in regulated environments. Practitioners who understand what it means to build for compliance, for data fragmentation, for legacy constraints. People who know the difference between a demo that impresses the board and a system that actually moves EBITDA.
The companies that close the implementation gap in the next 12 months will collect the returns the model projected. The ones that don’t will be explaining to the investment committee why the AI initiative stalled.
Your implementation gap has a cost, measured in months and in EBITDA points. If the AI initiative is running without a harness, we’ll tell you exactly what’s missing and how to fix it. Book the AI Readiness Assessment.
About the author.
