Don’t Buy the Hype: Checking Under the Hood of "AI" Startups
AI has shifted from a post-deal integration chore to a make-or-break factor in M&A. CIO Jason James shares how deal teams can spot AI-washing in target companies, use AI to speed up due diligence and post-close integration, and keep human judgment at

“Don’t let an AI tool make the call on whether to close a deal. That’s the human decision.
Adapted from Chapter 18 of Make IT Work: An IT Playbook for Mergers and Acquisitions
Nearly half of all technology deals now have an AI angle, and Bain & Company's Global M&A Report 2026 shows AI adoption among dealmakers has more than doubled to 45%. With the risk of sellers, or targets, using “AI-washing” to jack up their price, private equity sponsors, and boards need their CIOs to act as the deal’s underwriter. In the era of SaaS, overlooking technical flaws may have meant a bad IT integration, but now it can mean the target’s core business model could collapse.
Auditing the target
Everyone and their mother says they’re an AI business now. In a November 2026 Gartner survey, risk leaders flagged "AI strategy miscalculation," defined as committing capital based on “anticipated AI benefits that fail to materialize at the expected scale, speed, or economic value,” as a top emerging threat.
Here are four areas to focus on in your audit:
Where did they get the data?
Ask if the data is proprietary. Is it licensed? Are there regulatory restrictions on how it can be used? Has it been collected with proper consent? These questions matter because a model trained on improperly obtained data is a lawsuit waiting to happen. Regulations like the EU AI Act's Article 50 transparency rules have turned data governance into a huge liability.
Can the target explain how their models work?
Is there documentation of the training process, the evaluation metrics, or the known limitations? If the answer is “it’s in the code,” that’s a red flag. Undocumented models are technical debt of the highest order.
Does the math work?
AI workloads have different infrastructure requirements than traditional applications. They need GPUs, large-scale data storage, specialized compute environments, and often significant cloud spending. Make sure your infrastructure planning accounts for these requirements. A CIO who acquires an AI company and then cannot provide the compute resources the team needs is going to have a very short honeymoon.
Will the builders stick around?
AI talent is among the most competitive hiring markets in technology. If you’re acquiring a company for their AI capabilities, you’re really acquiring their AI team. If those data scientists and ML engineers walk out the door post-close, you’ve bought an expensive pile of code that nobody knows how to maintain. Retention packages for AI talent should be generous and front-loaded.
AI in post-close
Merging two companies will always be a bit chaotic. If you read my essential CIO’s data room checklist, you know that due diligence used to require teams of analysts going through virtual data rooms in spreadsheets and cross-checking terms manually. Fortunately, AI applications can also help with the integration work that comes after the deal is done.
Part of the road to success will be making sure customers don’t feel any of that chaos. As I wrote when looking at bridging the post-close gap and crisis control and managing acquisition chaos in the first 72 hours, integration leaders need an early warning system right after an announcement. Running automated text analysis on all support tickets and internal emails can flag recurring bottlenecks quickly so your team can fix issues before they escalate.
Personally, I’ve used AI tools to analyze licensing contracts in recent deals that I’ve closed. A task that would take my team days of reading through hundreds of pages of license agreements now takes hours. The AI highlights the important terms, identifies potential problems, and summarizes financial obligations. My team reviews the AI’s findings and makes the final call. It’s not flawless, but it’s a huge productivity boost. But please, don’t, don’t, don’t let an AI tool make the call on whether to close a deal. That’s the human decision. You need experienced operators reviewing every output when valuation and deal terms are on the line.
My pro tip: If you ask to see the model architecture and you get a slide with a robot on it, you’ve found a marketing department, not a capability.
Today, M&A technology leadership now includes the CIO and rightfully so. It’s just one more table we now have a seat at. Ask blunt architectural questions early, verify the unit economics yourself, and keep experienced operators in charge of the final call.
Jason James is Chief Information Officer (CIO) at Acumatica, a provider of cloud ERP solutions for small and midsize businesses, where he oversees the company’s ever-evolving digital landscape and translates the company’s vision into an actionable IT roadmap. He has over two decades of leadership experience in SaaS, cybersecurity, IT infrastructure, and digital transformation. Earlier, Jason was CIO at Aptos and at several high-growth software and data analytics organizations in the healthcare sector, and held executive-level IT roles at Servigistics (now PTC), PRGX, and EarthLink. He is the author of Make IT Work: An IT Playbook for Mergers and Acquisitions.
If this caught your attention, that’s not accidental.
The best editorial systems don’t happen by accident. Outlever builds them.









