The AI Manhattan Project won't work (but yours can)


Hey there Reader!

I'm in the midst of a final flurry of travel, preparing clients to start strong in 2026. I can't believe how fast it becomes the time of year when the elves are on the loose and we finish the workday in the dark, but here we are!

This week's Deep Dive feature is focused on the U.S. government's much-feted AI Manhattan project - a fascinating endeavor. Beyond that, in the world of PE and data:

  • Adobe’s $1.9 billion acquisition of Semrush signals that the future of marketing cloud is owning the “front door” of brand visibility in AI and search, fusing classic SEO with generative engine optimization so CMOs can manage how their brands surface across LLMs, traditional search, and the broader web from a single stack.
  • Real Madrid’s move to sell a 5% non‑voting stake to outside investors signals that even member-owned giants now see private equity and media-savvy capital as essential to compete, cementing a new commercial paradigm in soccer that clubs like Wrexham, with their early Hollywood-led, content-first ownership model, anticipated years ago.

I was joined on this week's Transformed With Data podcast by Ian Alden Russell. In his experience working with PE firms in Asia, Ian has found that better data doesn't always mean better decisions. This was a fascinating exploration of the value creation that can lie beyond the numbers.

video preview

Enjoy the rest of this week's Data Signals! If you know a company or a firm that needs help with their data silos, please pass the edition on to them!

DEEP DIVE

What the government's impending billion-dollar data integration failure can teach PE firms about portfolio intelligence

The federal government just announced what they're calling the "AI Manhattan Project" - a massive initiative to break down data silos across agencies and build a unified platform for AI-driven insights.

It sounds ambitious. It will probably fail.

Not because the vision is wrong, but because you can't retrofit data governance onto decades of technical debt, inconsistent standards, and systems that were never designed to talk to each other. The government is trying to unify datasets that use different schemas, operate under conflicting legal frameworks, and live in legacy systems that predate the iPhone.

But here's what's interesting for PE firms: you're facing the exact same challenge across your portfolio companies, except you actually have the mandate to fix it.


The Problem Is Identical (Just Smaller)

Think about your portfolio right now:

  • Company A runs NetSuite, Company B is on SAP, Company C built everything in Excel
  • Revenue recognition policies differ across portcos despite being in the same sector
  • Customer data lives in six different CRMs with no common taxonomy
  • Your deal team needs 72 hours and three analysts to answer: "What's our actual SaaS ARR across the portfolio?"

This is the same fragmentation problem the government faces. Multiple autonomous entities, different systems, sensitive data, and a sponsor trying to create shared intelligence.

The difference? You can impose standards. They can't.


Why This Actually Matters for Exits

When you can't quickly aggregate and analyze data across portfolio companies, you lose three things that directly impact valuation:

1. Cross-portfolio pattern recognition

Individual portcos can't see pricing power trends, churn drivers, or working capital levers that only become visible across multiple businesses. You're sitting on insights that could inform better pricing, better retention strategies, better capital deployment - but only if the data can actually be compared.

2. Speed and confidence in buyer conversations

Strategic buyers will ask cross-cutting questions during diligence: How does customer acquisition cost trend across your portfolio? What's the real margin profile when you normalize accounting? How sticky is revenue across different customer segments?

If you need three days and manual spreadsheets to answer, you've just signaled risk. Buyers discount what they don't trust.

3. The "enterprise-grade data maturity" premium

Increasingly, buyers care about whether a business can be integrated post-acquisition. Fragmented, ungoverned data isn't just an IT problem - it's an integration risk that shows up as earnouts, price cuts, and lower multiples.


The PE Advantage (That Most Firms Aren't Using)

Here's where you have a massive advantage over the government's doomed Manhattan Project:

You can standardize from day one post-acquisition.

The government has to negotiate with entrenched agencies, navigate conflicting regulations, and respect decades of operational autonomy. You don't.

Within 100 days of close, you can:

  • Define a core KPI model (revenue, margin, pipeline, churn) that every portco reports into
  • Build centralized pipelines that ingest data in a governed, standardized way
  • Create a "group analytics layer" while preserving domain-specific detail at the portco level
  • Establish role-based access and clear purpose limitations so data isn't just thrown into one big bucket

This isn't about forcing Company A to rip out SAP and adopt Company B's systems. It's about creating a conforming layer that lets you see across the portfolio without flattening important nuance.


What Actually Works (From Firms That've Done This)

The pattern that works isn't "build one massive integrated platform." It's more modular:

Treat each portco as a domain responsible for its own high-quality data products - clean, documented datasets they contribute to a shared environment with common contracts and governance.

Separate your monitoring layer from your experimentation layer - regulatory reporting and board decks pull from one system of record; AI experiments and what-if scenarios happen in a sandbox with different controls.

Build a library of reusable accelerators - reference pipelines, standard KPI definitions, feature templates that can be cloned and adapted per portco rather than rebuilt from scratch every time.

This is basically a data mesh approach, and it's how you get the benefits of integration (cross-portfolio insights, shared AI capabilities, faster diligence) without the concentration risk of putting everything in one giant, ungoverned bucket.


The Thing Nobody Wants to Hear

The AI Manhattan Project will struggle because the hard part isn't the technology. The hard part is getting autonomous entities with different incentives to adopt common standards and contribute clean data.

Sound familiar?

Your Portco CEOs are running their own businesses. They have their own priorities. "Feeding data to the parent company's analytics platform" is rarely at the top of the list unless you make it part of the mandate from day one.

This is an organizational problem that requires executive sponsorship, clear governance, and accountability - not just a technology fix.

The good news? You control the incentives.

FINAL SEND OFF

That's it.

Here's what you learned today:

  • The government's AI Manhattan Project faces the same data fragmentation you see across portfolio companies
  • Unlike government agencies, PE firms can impose standards and governance in the acquisition
  • Unified portfolio data isn't just "nice to have" - it de-risks diligence, speeds decision-making, and protects multiples
  • The pattern that works is modular: domain ownership + central governance + reusable components

Thank you for reading. Let us know how else we can help!

Cheers,

Graeme

Graeme Crawford

CEO at Crawford McMillan

Helping PE firms protect and grow company valuations with clean, reliable data.

CRAWFORD McMILLAN
Professional Data Consultancy of 25+ years

The content provided in this newsletter, including discussions regarding data architecture, operational efficiency, valuation readiness, and business strategy, is intended strictly for informational and educational purposes only. Decisions regarding capital allocation, investments, acquisitions, or business strategy should always be made in consultation with qualified professional financial, legal, and investment advisors.

113 Cherry St #92768, Seattle, WA 98104-2205
If you wish to change the emails you receive, please visit Preferences. If you wish to no longer receive communication via the newsletter, please Unsubscribe here.