# How to conduct an AI due diligence on a portfolio company AI due diligence for portfolio companies: a structured framework for PE operating partners to assess data maturity, AI capability and value creation potential. Published: 2025-01-13 Author: Rodan Analytics Most private equity firms now ask some version of the AI question during diligence. The problem is most of them ask the wrong questions — and the management teams they are interrogating know it. The typical exchange goes something like this: the deal team asks whether the company uses AI, the CEO says yes, someone mentions a pilot or a chatbot or a recent Microsoft Copilot rollout, and the box gets ticked. What nobody examines is whether any of it creates value, whether it is defensible or whether the organisation has the data infrastructure to do anything meaningful with AI at all. This matters more than it used to. AI capability — or the absence of it — is beginning to separate companies at the same revenue scale in ways that compound over a three-to-five year hold period. An organisation that enters your ownership without the foundations in place will cost you time and money to fix. One that has quietly built genuine capability is worth more than the model suggests. This article gives you a structured approach to AI due diligence that goes beyond the surface. It covers what to look for, how to assess it and what it means for value creation. ## Why standard commercial due diligence misses the AI question Commercial due diligence is good at assessing markets, competitive position and management quality. It is not designed to assess data maturity, technical debt or AI capability. Most CDD providers will include a section on digital or technology — but it tends to be descriptive rather than diagnostic. It will tell you what systems the company runs. It will not tell you whether those systems produce data that can be used, or whether the organisation is capable of acting on analytical insight. The gap matters because AI capability is not binary. It exists on a spectrum, and where a company sits on that spectrum determines what is possible during your hold period and at what cost. A business with clean, integrated data and a technically literate commercial team can deploy meaningful AI systems in months. A business with fragmented CRM data, no data warehouse and an operations team that runs on spreadsheets will need eighteen months of foundations work before it can benefit from anything more sophisticated than a dashboard. Both of those businesses might describe themselves as "exploring AI opportunities." You need a framework that distinguishes between them. ## The four dimensions of AI due diligence Assessing AI capability in a portfolio company requires you to look across four distinct dimensions. They are not equally weighted — data infrastructure is foundational and should be examined first — but all four inform the investment thesis and the hundred-day plan. **1. Data infrastructure and quality** Start here. Every AI use case depends on data. The question is not whether the company collects data — most do — but whether that data is structured, accessible and trustworthy. Ask to see the data architecture. Where does operational data live? Is there a central warehouse or does each function maintain its own system of record? How is customer data linked across touchpoints? What is the data latency — are commercial teams working from yesterday's numbers or last quarter's? In a recent diligence on a mid-market B2B distributor, the management team presented an impressive-looking analytics dashboard. Closer examination revealed it was manually updated by the finance analyst every Monday morning from four separate spreadsheet exports. The underlying systems had no integration. The "data capability" the management team was proud of was a person doing a job that would not scale. **2. Current AI and analytics use cases** Separate genuine capability from vendor-installed tooling. A company that has deployed AI as part of a broader SaaS platform — say, demand forecasting inside their ERP, or a recommendation engine in their ecommerce stack — has not necessarily built AI capability. It has bought a feature. What you are looking for is evidence that the organisation understands its own data well enough to identify problems worth solving analytically, and has either the internal capability or the commercial relationships to solve them. Ask: what decisions does AI currently inform? Who makes those decisions? How do they know when the model is wrong? The distinction between a company that uses AI as a feature and one that uses it as a capability is partly technical and partly cultural. The cultural dimension is often the harder one to fix. **3. Technical talent and organisational readiness** You do not need a portfolio company to have a data science team. You do need it to have someone who can own the data agenda — a head of data, a technically fluent CTO or a senior analyst who has the mandate and the access to drive change. Assess the technical depth of the leadership team. Can the CEO articulate the company's data strategy without reading from a slide? Does the CTO distinguish between AI infrastructure and AI applications? Is there a budget owner for data and analytics, or does it sit as a line item under IT? Equally important: assess the organisation's tolerance for data-driven decision-making. Companies that say they are data-driven but override analytical recommendations with gut instinct will not extract value from AI investment regardless of what you build. **4. AI risk and governance** This dimension is underweighted in most diligence processes and will become harder to ignore as regulation tightens. The EU AI Act is already shaping how technology businesses operating in European markets classify and govern their AI systems. For UK-headquartered businesses, the regulatory picture is still evolving, but the direction of travel is clear. Ask whether the company has any documented AI governance — policies around model use, data handling, explainability requirements or bias monitoring. Ask whether any AI systems touch regulated processes: credit decisions, hiring, pricing in regulated sectors. Ask whether the company has assessed its exposure under current and forthcoming AI regulation. A logistics business that has deployed route-optimisation AI with no documentation of how it makes decisions is a manageable risk. A financial services business using an undocumented model to influence lending decisions is not. ## How to structure the AI diligence workstream AI diligence should run as a parallel workstream to commercial and financial diligence, not as an afterthought. It requires a different set of questions and, in most cases, a different set of interviewers. A data strategist or AI advisory firm conducting a focused technical assessment will surface issues that a generalist deal team will not. Structure the workstream in three phases: - **Document review** — data architecture diagrams, system inventories, any existing analytics or data strategy documentation, vendor contracts for AI-adjacent tools - **Technical interviews** — CTO, head of data or equivalent, plus one or two operational leaders who are actual users of analytical tools - **Rapid assessment** — a structured scoring of the four dimensions above, with a clear read on what is possible during the hold period and what foundation work is required The output should not be a qualitative narrative. It should be a capability rating, a gap analysis and a prioritised roadmap with rough cost and time estimates. That gives the operating partner something to act on from day one. ## What the findings mean for value creation AI diligence is only useful if it connects to the value creation thesis. Frame the output in commercial terms. A distribution business with clean transactional data and a fragmented pricing operation represents a near-term AI opportunity: dynamic pricing or margin analytics could improve EBITDA within twelve months with the right deployment. A professional services firm with poor data infrastructure and a partnership model resistant to process change represents a longer-term play, where the first eighteen months need to go into foundations before any AI application creates value. The diligence findings should answer three questions directly: what AI-driven value is available during this hold period, what does it cost to access it and what are the risks if we do not address the gaps? If the answers to those questions are not in your investment committee memo, you are making a decision with incomplete information. ## Turning diligence findings into a hundred-day plan The companies that extract value from AI investment in PE-backed environments share one characteristic: they do not wait for perfect conditions. They identify the highest-value, lowest-friction use case available within the existing data environment and build from there. That requires a specific kind of AI advisory support — not a firm that will spend six months writing a strategy document, but one that can assess the situation quickly, recommend a starting point and deploy alongside the management team. Rodan's diagnostic engagement is designed for exactly this moment. For a fixed fee, we assess AI and data capability across a portfolio company, deliver a prioritised roadmap and give the operating partner a clear view of what is realistic during the hold period. It typically takes three weeks and informs both the hundred-day plan and the longer-term value creation plan. If you are entering ownership of a business and do not yet have a clear read on its AI and data position, that is the right place to start. --- **Meta description:** AI due diligence for portfolio companies: a structured framework for PE operating partners to assess data maturity, AI capability and value creation potential. HTML: https://rodan.io/insights/how-to-conduct-an-ai-due-diligence-on-a-portfolio-company