Benchmarking AI maturity across a private equity portfolio
Most private equity firms now ask about AI in due diligence. Fewer know what to do with the answers. A portco says it uses machine learning for demand forecasting - is that a competitive asset or a retrofitted spreadsheet with a new label? Another says it has "an AI roadmap" - who wrote it, and does the CEO know what is in it?
The mistake most firms make is treating AI maturity as a binary. Either a business has AI or it does not. That framing misses almost everything that matters: whether the capability is proprietary or vendor-dependent, whether the data infrastructure can support what the roadmap promises, whether the team can actually execute.
For operating partners managing a portfolio of eight to fifteen companies at different stages of transformation, the absence of a consistent benchmarking methodology is not just an inconvenience - it is a valuation risk.
This article gives you a practical framework for assessing AI maturity across a portfolio, identifying where value is being left unrealised and prioritising where to deploy resource.
Why standard due diligence misses AI risk
Traditional due diligence captures technology spend, system architecture and key-person dependencies. It does not capture the depth of a company's data capability or whether its AI investments are generating commercial returns.
Consider a distribution business acquired at 8x EBITDA. The management team presents a pricing optimisation model they built in-house. It looks credible in the data room. Post-acquisition, the operating partner discovers the model runs in Excel, depends on one analyst who is considering leaving and has not been updated in eleven months. The "AI capability" was real - but fragile, undocumented and completely unscalable.
This scenario is not unusual. The issue is that most diligence questionnaires ask whether AI exists. They do not ask whether it is institutionalised.
The questions that matter are different:
- Is the model embedded in a production system or does it live in a personal drive?
- Is there a feedback loop - does the model improve over time?
- Does the business track the commercial output of the model, or just its technical performance?
- Could this capability survive the departure of the person who built it?
A business that answers yes to all four is in a fundamentally different position from one that answers yes to the first question only. Diligence needs to reflect that distinction.
A five-level maturity framework for portfolio benchmarking
Maturity models are only useful if they drive decisions. The following framework is designed to produce a score that maps directly to a prioritisation and investment decision - not to sit in an appendix.
Level 1 - Unstructured. Data is siloed across systems, reporting is manual and there is no analytical capability beyond standard business intelligence. AI is absent or purely experimental.
Level 2 - Aware. The business has identified AI as a priority and may have one or two point solutions in place (a chatbot, a vendor-provided recommendation engine). There is no coherent data strategy and no internal capability to build or maintain models.
Level 3 - Building. The business has a functioning data team, a modern data stack and at least one AI application generating measurable commercial output. The roadmap is credible but execution is inconsistent.
Level 4 - Scaling. Multiple AI applications are in production. The data infrastructure is robust. There is a governance framework - the business knows who owns AI decisions, how models are monitored and how failures are handled. The capability is institutional, not individual.
Level 5 - Differentiating. AI is a source of structural competitive advantage. The business generates proprietary data that competitors cannot replicate. Models improve with scale. AI capability influences product, pricing, customer retention and operational efficiency simultaneously.
Most mid-market portcos sit at Level 2 or early Level 3. The diagnostic value comes not from the score itself but from understanding why - and what is blocking the next level.
Scoring consistently across different business models
The challenge with a portfolio is that a Level 3 B2B software business and a Level 3 logistics business look completely different in practice. Applying the same criteria produces misleading comparisons.
The solution is to score against two dimensions separately: absolute maturity (the five-level scale above) and sector-relative maturity (how the business compares to its direct competitors and peers).
A professional services firm at Level 2 may actually be ahead of its market. A retail business at Level 3 may be dangerously behind. Both scores matter for different reasons - the first informs value creation planning, the second informs competitive risk.
When Rodan conducts AI maturity assessments across a portfolio, we use a structured diagnostic that covers six domains: data infrastructure, analytical capability, AI deployment, governance, talent and commercial integration. Each domain is scored independently before a composite is calculated. The domain-level breakdown is more useful than the headline number - it shows exactly where the gap is.
A consumer goods portco might score well on data infrastructure (it has a modern warehouse and clean product data) but poorly on commercial integration (the insights sit with the analytics team and never reach the category managers making ranging decisions). The fix in that case is not a technology investment - it is a workflow change and a different reporting structure.
Translating maturity scores into capital allocation decisions
A benchmarking exercise is only worth the investment if it changes how you allocate resource. Here is how to use portfolio-level maturity data.
Prioritise Level 2 businesses with strong data assets. These are the highest-return opportunities. The infrastructure work is already partially done - the gap is analytical capability and deployment. A focused six-to-twelve month intervention can move a business from Level 2 to Level 3, which in the right sector has a direct impact on EBITDA multiple at exit.
Stabilise Level 3 businesses before adding capability. The common mistake at Level 3 is building new models before existing ones are properly institutionalised. Scaling fragile AI is worse than having none - it creates operational dependency without reliability. The priority here is governance and documentation, not new projects.
Protect Level 4 and 5 businesses from regression. Key-person risk, vendor lock-in and data quality degradation are the main threats. At this level, the operating partner role is closer to risk management than transformation.
Use portfolio benchmarks to set board expectations. When every portco has been assessed using the same framework, you can present a portfolio-wide AI maturity heatmap to your LP base. That changes the conversation from "what is your AI strategy" - a question that currently produces very different quality answers - to "here is our methodology and here is where each business sits against it."
The cost of benchmarking without acting on it
Firms that run maturity assessments and file the results have wasted the money. The output of a benchmark is a prioritised action list - and that list has a shelf life.
AI capability gaps compound. A portco that sits at Level 2 today while its sector moves to Level 3 does not stay at Level 2 relative to peers - it falls behind. The gap between a business that has institutionalised AI and one that is still experimenting is not static. It widens.
For an operating partner managing a portfolio through a three-to-five year hold period, the window to close that gap and have it reflected in exit valuation is narrower than it looks. The businesses that will command a premium multiple at exit in the next cycle are the ones where AI capability is demonstrable, documented and defensible - not the ones where it is promised.
If you are building or refreshing your portfolio AI assessment process, Rodan offers a structured diagnostic engagement - typically completed within three to four weeks - that produces a domain-level maturity score, a competitive benchmark and a prioritised roadmap. Book a diagnostic conversation with our team.
