Five Questions with an OG, John Shaw: GenAI Diligence and Value Creation in Private Equity

MissionOG is fortunate to be supported by a deep network of experienced operators and entrepreneurs. This entry is part of a blog series where we share perspectives from “OGs” — original innovators from specific market segments and/or business disciplines. In this installment, we look at how generative AI is changing the way companies assess opportunities, deploy technology, and measure impact, from evaluating AI readiness to turning promising use cases into measurable operating results.

 

John Shaw is CEO and Co-Founder of PraxisIQ, an AI consultancy and software company that works with private equity firms and mid-market companies to turn generative AI from a board-deck talking point into realized value.

 

PLEASE PROVIDE A BRIEF OVERVIEW OF YOUR BACKGROUND AND CURRENT ROLE.

I founded Nimbo, a cloud services firm, in the early days of enterprise cloud migration and sold it to Equinix in 2015. That was my education in what it costs to actually change how a company operates, versus what it costs to write a strategy for changing how a company operates. Very different numbers.

I then spent several years at AWS, leading the Global Healthcare and Life Sciences AI/ML team first — where I learned what it takes to put machine learning into environments where being wrong has consequences — and then the Private Equity team. That second role was the formative one: a close look at the gap between what firms said they would do with technology in the value creation plan and what actually happened in months four through twelve.

Today I’m CEO and Co-Founder of PraxisIQ, based in Austin. We work with PE firms and mid-market companies on GenAI diligence and post-close value creation, delivered through a forward-deployed engineering model.

 

WHAT INSPIRED YOU TO LAUNCH PRAXISIQ, AND WHAT MARKET GAP ARE YOU AIMING TO FILL?

Watching the same pattern repeat. A firm gets excited about AI in the thesis, commissions a strategy engagement, receives a very good deck, and then discovers nobody in the company can build the thing in the deck. Eighteen months later the AI line item has produced a chatbot and a stalled pilot.

The gap is a delivery gap, not an insight gap. Private equity is well served by strategy consultancies at the top and development shops at the bottom, and badly served in the middle. Strategy firms hand you a recommendation; dev shops build what you specify but have no context on the thesis, the hold period, or where the exit multiple comes from. Neither is accountable for a value creation number. And a $60 million revenue company can’t solve it internally — it cannot hire a chief AI officer, a data platform team, and an ML bench.

So we built PraxisIQ around three things: engineers embedded with the company rather than consultants presenting to it; delivery measured in weeks; and software underneath the services. Our FuseIQ platform is designed to track AI spend, use-case ROI, and training adoption, with those measures rolling up into a companywide view of AI maturity.

 

HOW DOES GENAI DUE DILIGENCE DIFFER FROM TRADITIONAL TECHNOLOGY DILIGENCE, AND WHAT INDICATORS DO YOU LOOK FOR WHEN ASSESSING A TARGET’S AI READINESS?

GenAI diligence requires a different set of questions because the output is probabilistic rather than deterministic. Instead of asking only whether the technology works, ask how often it fails, how those failures are measured, and what controls exist before an error reaches a customer.

A good diligence process should focus on four areas.

First, test data rights. Determine whether the company has the contractual right to use customer and proprietary data for training, inference, and product improvement.

Second, test the economics at scale. Model inference costs at several times current usage and understand how those costs affect gross margin.

Third, test defensibility. Ask whether a capable team could replicate the product quickly using third party models, or whether the company benefits from proprietary data, embedded workflows, integrations, and feedback loops that improve with use.

Fourth, test whether the product is truly in production. Look for real usage, measurable business outcomes, and a systematic evaluation process that tests model quality as the product changes.

One of the strongest indicators of AI readiness is whether the company can clearly answer four questions: How is quality measured? Who owns the business outcome? What does each transaction cost? What happens when the model is wrong?

If those answers are clear and supported by operating data, the company is usually much further along than one that can only point to pilots, demos, or model capabilities.

 

ONCE A TRANSACTION CLOSES, WHAT SHOULD THE FIRST 100 DAYS OF GENAI IMPLEMENTATION LOOK LIKE, AND WHICH EARLY WINS TEND TO BUILD THE MOST INTERNAL CREDIBILITY?

The biggest lesson we have learned is that companies do not need another AI strategy document. They need a practical sequence for moving from strategy to production.

The first 100 days generally follow four phases.

• Days 0 to 15: establish the baseline. Understand the systems, data, existing AI usage, and potential use cases. Then narrow the focus to two or three priorities.
• Days 15 to 45: put one narrow use case into production with real users and real data. Companies often learn more from a focused deployment than from months of additional analysis.
• Days 45 to 75: measure usage, cost, accuracy, adoption, and return. Governance matters, but it should enable responsible deployment rather than become another approval layer.
• Days 75 to 100: determine what worked, what did not, and where to invest next.

One area companies often underestimate is instrumentation. Management may have limited visibility into what AI is being used, what it costs, and whether it is producing measurable value. We encountered that problem often enough that it led us to build FuseIQ, but the broader lesson is more important: leaders need a clear view of AI spending, adoption, and business outcomes.

That visibility builds confidence. When management can see what is working and why, the conversation shifts from whether to invest in AI to where to invest next.

The early wins that build the most credibility are usually practical ones that remove repetitive work rather than threaten jobs. Contract review, invoice matching, claims intake, and RFP preparation are good examples.

Ultimately, the specific use case matters less than the execution. A focused solution that reaches production quickly, involves employees, measures results, and is transparent about its limitations will usually create more internal momentum than a more ambitious project that remains stuck in pilot mode.

 

WHERE HAVE YOU SEEN GENERATIVE AI UNLOCK THE FASTEST, MOST MEASURABLE ROI ACROSS THE VALUE CREATION LEVERS?

Operational efficiency, and it isn’t close — specifically the document-heavy back office. The reason is structural rather than technological. These use cases have a measurable pre-existing baseline: cycle time, cost per document, error rate, hours. The workflow owner controls the change end to end, with no customer dependency, no pricing decision, no sales cycle, no migration risk. You reach a defensible number in 60 to 90 days, and finance accepts it because it’s denominated in units they already track.

Go-to-market acceleration is next, on a two-to-three-quarter horizon; the impact can be larger, but attribution gets contested — sales leaders have many explanations for a good quarter and only some involve your project. Product innovation is where the largest value lives, because it moves the exit multiple rather than this year’s EBITDA, and it’s also the slowest.

So the advice that I give sponsors is to use operational efficiency to buy credibility and fund the product work. Efficiency shows up in EBITDA during the hold; product innovation shows up in the multiple at exit. You rarely get permission for the second without delivering the first.

 

HOW DO YOU ENVISION PRIVATE EQUITY EVOLVING AS FIRMS INCREASINGLY INTEGRATE AI AND GENERATIVE AI INTO THEIR INVESTMENT AND VALUE-CREATION STRATEGIES?

AI readiness becomes a standard diligence workstream, sitting alongside quality of earnings and cyber, with findings priced into the model. Value creation teams change shape too: the operating partner model was built for people who had run functions, and it’s shifting toward people who can build.

Portfolio-level leverage becomes a significant opportunity and remains underused. A firm with 30 portfolio companies has something no single company has — repeated exposure to the same problems and permission to build once and deploy many times. Firms treating AI as thirty separate initiatives are leaving most of the value on the table.

And the risk side, which I’d underline for anyone underwriting deals right now: AI-native competition is a diligence question in its own right. When you’re pricing a business whose moat is proprietary process, accumulated document expertise, or headcount-based service delivery, you have to ask what it looks like against a competitor with a tenth of the people. Some of those moats are real. Some are being repriced as we speak.

The larger point is that hold periods and AI cycles are badly mismatched. Capability changes materially every few quarters; you own the company for five years. That argues against betting a thesis on any specific technology and for building durable capability — data foundations, the ability to evaluate and ship, and a measurement layer that survives the next model migration rather than being rebuilt around it. The firms that internalize that will compound. The ones buying point solutions will be re-buying them in eighteen months.