Almost every target now has an AI story in its management presentation. Some of those stories add value, some add risk, and a few are simply marketing. Before you close, you need two answers: what could AI cost you in this business, and where could it move the P&L during your hold? This guide gives you the questions for both, grouped by area, plus the red flags that should change your price or your plan.
The short answer
- Diligence AI through two lenses: risk in the target (claims, data rights, model dependence, security, regulation) and upside (specific workflows where AI can change revenue, margin, or cash).
- Ask for evidence, not descriptions: contracts, usage logs, architecture diagrams, and before-and-after metrics.
- Carry the upside into the 100-day plan as a short, ranked list of initiatives with owners, not as a line in the investment memo.
Why AI deserves its own workstream
Regulators have already acted on overstated AI claims. In March 2024 the SEC settled charges against two investment advisers, Delphia and Global Predictions, for saying they used AI in ways they didn't, with $400,000 in combined penalties. Six months later the FTC announced Operation AI Comply, a sweep against deceptive AI claims. If a target's growth story leans on AI, you're buying that disclosure risk along with it.
The upside is real, but early. Bain, which advises PE firms, surveyed private investors with $3.2 trillion in assets under management in September 2024. Most said their portfolio companies were testing generative AI, and nearly 20% of companies had operationalized use cases with concrete results. So treat AI value as something you'll have to build, not something you're acquiring.
Lens one: AI risk in the target
What could hurt the deal
- AI claims
- Data rights
- Model and vendor dependence
- Security and controls
- Regulation
Lens two: AI value-creation upside
What could move a P&L line
- Where the work is
- Where the data is
- Whether the team can execute
Lens one: AI risk in the target
AI claims
- Which products, services, or internal processes does management describe as AI-powered? Show us each one running.
- What share of revenue depends on those features, and can you prove customers use them?
- Is any "AI" actually rules, scripts, or offshore manual work?
- Who wrote the AI language in the CIM and sales collateral, and was it reviewed by counsel?
Data rights
- What data trains or feeds each model, and do customer contracts permit that use?
- In field services, do service agreements allow job photos, call recordings, and site data to be used beyond the job itself?
- In energy infrastructure, do customer specifications, drawings, or test data carry confidentiality terms that limit reuse?
- Are there data processing agreements with every AI vendor, and do vendors retain or train on the company's data?
Model and vendor dependence
- Which third-party models and platforms does the company rely on, and on what terms?
- What happens to cost and service if a provider raises prices, changes a model, or exits?
- Who owns the prompts, fine-tuned models, workflows, and code built around them?
- How many people actually understand these systems? Is that knowledge in one contractor's head?
Security and controls
- Is there an inventory of AI tools in use, including ones employees adopted on their own?
- What stops sensitive data, such as customer records or engineering drawings, from being pasted into public tools?
- Are AI outputs that reach customers, such as quotes or service recommendations, reviewed by a person?
- Has anyone tested the customer-facing AI for prompt injection or data leakage?
Regulation
- Does the company sell into the EU, or deploy AI that affects EU residents? The EU AI Act classes some critical-infrastructure and employment uses as high-risk. The European Commission now says those Annex III obligations apply from December 2, 2027.
- Is AI used in hiring, technician scoring, or performance management? That draws scrutiny in many jurisdictions.
- Does marketing describe AI capabilities that the product can't substantiate today?
Lens two: AI value-creation upside
The goal here isn't a long list of use cases. It's finding the three to five workflows where AI could change a P&L line within your hold period.
Where the work is
- Which workflows consume the most labor hours outside the core service or production work? Think scheduling, quoting, documentation, invoicing, and reporting.
- In field services, how much technician time goes to travel, paperwork, and repeat visits?
- In energy infrastructure, how long does it take to go from RFQ to quote, and from order to approved bill of materials?
Where the data is
- Is work-order, job, and customer history captured in a system, or spread across spreadsheets and email?
- Can the company produce a clean, monthly history of the KPIs you'd target, such as utilization, first-time completion, or quote win rate?
- Is there proprietary data, like years of engineering designs or service records, that a competitor couldn't easily replicate?
Whether the team can execute
- Who would own an AI initiative day to day, and do they have the time?
- Has the company ever shipped a technology change that front-line staff actually adopted?
- What would management do first with AI, and how would they measure it?
BCG's 2024 survey of 1,000 executives, published by a firm that sells AI consulting, found that 74% of companies had yet to show tangible value from AI. It attributed about 70% of the challenges to people and process, not technology. That's why the execution questions matter as much as the data ones.
Red flags
| Red flag | Why it matters | What to do |
|---|---|---|
| AI features can't be demonstrated live | Disclosure and revenue-quality risk | Strip AI-linked growth from your base case |
| No contractual right to use customer data | Models or products may need rework | Specific indemnity or remediation covenant |
| One vendor or one contractor holds everything | Cost shock or continuity risk | Transition plan and IP assignment before close |
| No inventory of AI tools in use | Unknown data exposure | Day-one audit in the 100-day plan |
| No baseline KPIs for target workflows | Upside can't be measured or credited | Build baselines before launching initiatives |
| AI value cited without numbers | The upside is a hope, not a plan | Ask for before-and-after metrics, or price it at zero |
A starting framework, not legal advice. Have counsel review data rights and regulatory exposure.
Turning findings into a plan
Diligence findings lose value if they sit in a report. Before signing, rank the upside opportunities by value, feasibility, and time to impact. Turn the risk findings into day-one actions. Then hand both to whoever owns the 100-day plan, so the first month starts with decisions rather than another assessment.
We do AI and operational diligence and 100-day planning for PE firms in field services and energy infrastructure, so weigh our view accordingly. The template below works whoever does the work.
Ten questions to answer before you close
- Can every AI claim in the CIM be demonstrated? If not, remove it from the value story.
- Does the company have the right to use the data its AI depends on? Check customer contracts and vendor terms.
- Which vendors could raise prices or change terms on you? Model the cost of switching.
- Who owns the AI-related IP? Confirm assignments from employees and contractors.
- Is there an inventory of AI tools in use? If not, plan one for week one.
- Is any AI use subject to the EU AI Act or employment rules? Get counsel's view.
- Which three to five workflows could AI change within the hold? Name them.
- Are there clean KPI baselines for those workflows? No baseline means no credible value case.
- Who would own execution? Name the person, not the function.
- What's in the base case versus the upside case? Keep unproven AI value out of the price.
Where to go from here
If you have a deal in diligence and want a second set of eyes on the AI thesis, tell us about the target. You can also see how we work across diligence, the 100-day plan, and exit on our Private Equity page.
Sources
- SEC: SEC Charges Two Investment Advisers with Making False and Misleading Statements About Their Use of Artificial Intelligence
- FTC: FTC Announces Crackdown on Deceptive AI Claims and Schemes
- Bain & Company: Field Notes from the Generative AI Insurgency (Global Private Equity Report 2025)
- European Commission: AI Act, regulatory framework for AI
- BCG: AI Adoption in 2024: 74% of Companies Struggle to Achieve and Scale Value