The Four Words Nobody Wants to Hear

Dave: Big news. Leadership approved the new data platform. Seven figures. Full enterprise license.

Jen: Nice. What data are we putting in it?

Dave: That's Phase 2.

Jen: What are the data standards?

Dave: Phase 3.

Jen: Who owns the data?

Dave: That's more of a governance question.

Jen: Right. So when does governance start?

Dave: It's not in scope.

Jen: So we bought a seven figure tool with no data, no standards, and no governance.

Dave: Yes but the demo was incredible.

Jen: ...

Dave: They had a dashboard with live charts.

Jen: Of what data?

Dave: Demo data.

Jen: Outstanding.

If that conversation made you wince? Good. It should. If it made you laugh because you've lived it? Welcome. Pull up a chair.

I'm giving myself a birthday gift: my first LinkedIn post and editorial series. Ever.

I will opt out of posting about "10 Ways to Be a Leader Before Breakfast" and "why I'm successful and you should feel bad about it". Not my wheelhouse, but here's why I'm finally writing.

Myself and most of my colleagues, people who have spent decades inside federal programs, have been making the same observations for years. In hallways, on whiteboards, in post mortem "Lessons Learned" which nobody reads. The same patterns. The same mistakes. The same outcomes. And we keep watching talented people and capable organizations fall into the same traps with the best of intentions. So this is Part 1 of an ongoing series to crystalize and discuss these patterns. To help us all recognize them more often and ask some honest questions to any Federal Agency out there who has seen them up close.

WARNING: It's going to be direct. Along my career, the best mentors I have known have all been brutally honest at all times. None of them ever won a personality contest but they were (1) authentic, zero time for platitudes, (2) DEAD ON about the subject at hand and (3) had an unquestionable track record of success in their wheelhouse. If this series comes across like the relative at Thanksgiving who pokes a hole in your Grandma's Etsy business plan?

Guilty as charged.

But like that person, the goal is to save the business (OR) kill it as quickly as possible for Nana's sake. Stick around. We're rooting for you Federal agencies out there.

Why I'm writing

First, let's be clear about something. We all love technology to some degree or another. My career started building and testing the very first "personal computers" (google "Columbia Data Products"). Then a Formula 1 speed of tech waves ensued for networks (Novell), the early email (Crosstalk and WordPerfect Office), client server (Lotus Notes), then "the interwebs", cloud, analytics, and so on. Every wave was exciting and every wave changed everything. Now it's 2026 and AI?

AI might be the most important paradigm earthquake since the printing press or, I don't know, perhaps the discovery of FIRE by humans. In 2026, building these early AI systems and agents on government infrastructure right now is EXCITING. Far beyond any previous tech wave. So the themes in this series are not coming from myself or others who are afraid of the future role of AI in Government systems. In fact, it's quite the opposite.

This is coming from people who have watched the future get sabotaged, repeatedly, by skipping one critical step. AI could be the next tech casualty to fall short of expectations.

Here goes.

Four words, in this exact order

There are four words that matter in enterprise data, and they go in this exact order:

Data Governance → Data Standards → Data Management → Data Exchange.

Your favorite software tool? The one you've been dreaming about installing for your org? The one where the vendor demo was so good you've been operating as their pro bono sales rep with your cubicle neighbors? That tool belongs in the last two categories. Never the first two. Ever. Read that over and over again as needed.

Still there? I need to check in, as this is where we lose half the room. Why? Because most federal agencies start with tools. They need them always and lots of them. An RFI goes out. A shiny platform gets procured. Best of intentions, real budget, even aligns with strategy. Eighteen months later it's sitting in the software landfill next to the last platform, the one before that, and the one before that. And here's the part nobody talks about: the vendor didn't want that outcome either. No commercial product company wants the reputation of "that platform failed us" before their platform and tools even had a chance to impact the agency's mission.

Federal software vendors build capable products. They deserve a fair shot. But getting a purchase order and a few installs is like getting a tee time. It doesn't mean you can play golf.

When the tool is aligned with the agency mission instead of just the building advocate, it's a much better long term value proposition for everyone: the agency, the vendor, the taxpayer. Vendors don't want to end up in the landfill any more than we want to put them there. But it requires both sides to understand what the platform and toolset is walking into.

The pantry is the problem

I used AI to ask our database who the authoritative source was. It replied "you should ask the other database."

Why do so many capable enterprise software platforms and tools fail? Because we handed world class products ungoverned, unstructured, non authoritative data and asked them to perform miracles. That's like handing Gordon Ramsay a kitchen full of expired ingredients and then blaming him when dinner is terrible. The chef isn't the problem. The pantry is.

"Garbage in, garbage out" is the Mick Jagger of IT catch phrases and is making a comeback. We must admit it still rings true with modern APIs and absolutely rings true with AI today. Across the Federal IT landscape, the technology changes every few years but guess what? The failure mode hasn't changed once.

[side note: In case you're wondering, the boy-band-one-hit wonder catch phrase in Federal IT was "Big Data" and saying it as often as possible to sound smart. Most of us in this industry are relieved it's gone from the tech lexicon. Word of advice: stop saying that phrase please, it makes the kids "cringe".]

Large agencies managing hundreds of systems and thousands of databases almost always have hundreds of data silos. And everyone treats the silos like the disease. They're not. They're a symptom. The real diagnosis? Zero enterprise data governance. Nobody has documented the authoritative sources for their own data. Who owns it? Who maintains it? Who publishes the authoritative version? What's the data lineage? Policy enforcement? Where's the configuration management? Silence.

Here's a question worth sitting with: does your agency leadership regard data governance as a critical mission enabler, a 1st Class discipline embraced by all? Or as a geeky administrative exercise that is (inexplicably) unrelated to critical operations?

Be honest. Because that answer explains almost everything.

QUESTION: When did any critical agency discipline which we now take for granted move from the kids table to the adults table?

There was a time when many Federal agency enterprises viewed master scheduling, configuration management, Verification & Validation or even program management the same way. It was seen as overhead, bureaucracy, paper pushing that "real operators" didn't need. Until projects started collapsing, requirements failed beyond repair, budgets evaporated, and entire programs were publicly cancelled. It took a generation of expensive, painful lessons before program management earned its seat at the table. Data governance is fighting that exact same battle today. And the agencies that figure this out first will modernize years ahead of everyone else. Think simply throwing everything "AI" at literally everything you have will fix known and unknown data challenges?

Artisanal data processing

And every organization has a hero. The spreadsheet jockey who whips up a complicated Excel file at 2am just in time to brief leadership during the latest crisis. Everyone applauds. Seriously. This is happening in your building right now and can be witnessed daily by all of us. It's like watching a matador take a bow after the eighth pass. The crowd goes wild. The bull is still in the ring. And sooner or later, the math catches up. (This topic gets its own deep dive later in the series.)

That is not automation. That is the exact opposite of automation. Your critical agency processes and operations are not going to GET to automation if your mission ops suffer from a systematic reliance on "Human Middleware". So ask your doctor if an emergency, painful tattoo removal of human middleware is right for you.

Does that impact your favorite support contractors? Yeah. It probably does. Stay focused.

(Not to worry, we are about to see an army of agency spreadsheet jockeys transform into "AI jockeys" by the thousands.)

Meanwhile, the metric that actually matters? What some call Data Excellence, what I refer to as Data Maturity, never moves. Not once. And the cycle repeats for every Line of Business. You could set your watch to it, except the watch is also ungoverned.

Walk towards the light

Here's what we keep coming back to. AI is too important, too transformational, to let it FAIL SPECTACULARLY the way too many previous generations of technology platforms have in government.

So the goodness of AI or automation modernization cannot be transformational if you skip governance.

Let that sit for a second.

What an agency CAN do, and this is the genuinely exciting part, is give modern platforms and capable tools exactly what they need to succeed: authoritative, structured, well governed, mature data. When that happens? The software platforms and tools don't just work, they thrive. Down in the Star Trek engine room, APIs deliver clean, trustworthy data which maintains verification and validation from end to end.

...Suddenly dashboards stop lying to leadership.

And AI? We're already seeing quick, reactive AI solutions spit out unreliable garbage for our most important data. Wait, what happened? I thought our new chatbot was magic? Heard somewhere in an AI demonstration: "Ha! You see? I told you AI was no good, it doesn't even understand our data governance!"

...Well, the question is... do you?

Here's the difference that matters. There's a world of gap between a RAG-based chatbot that ingests unstructured PDFs and spits out a different answer every time you ask the same question, and an AI model built on governed, structured, authoritative data that provides the correct response, consistently, every single time. Same question Monday. Same question Friday. Same answer. The right answer. That's not a theoretical difference. That's the difference between a novelty demo and a production system people trust with mission critical decisions. One is a party trick. The other is governed infrastructure.

The good news? This is fixable. And it's not as complicated or expensive as the industry has made it sound. If you're STILL here? Wow, thank you. Here's my favorite phrase with my colleagues...

You cannot "AI" or outsource what you cannot govern. Stand in the mirror and repeat that to yourself.

Every stakeholder from federal agencies, vendors, contractors and engineers has a chance to get aligned here. The vendor who understands the mission and governance and helps the agency mature its data excellence isn't just selling software. They're building a partnership that lasts. The agency that manages its own data governance and feeds tools clean data isn't just buying technology. They're building a foundation that every future investment benefits from. Everyone wins. And the landfill gets a little emptier.

One use case I am working on is building this concept into a 200,000+ record sandbox. Siloed legacy federal systems consolidated into a single governed registry, which is like giving Red Bull to AI powered components. Once you see an AI responding with structured responses you cannot un-see it. Early results are great, no hallucinations. It works because the data governance and data standards came first. Bonus: we then began testing to see if local AI models can help highlight and measure where the data quality gaps exist across silos. Try comparing a bull fighting event to watching a Blue Angels show.

In 2026, the future is incredible. The tools are incredible. AI is going to change everything. But four words have to come first. In order.

Data Governance. Data Standards. Data Management. Data Exchange.

Get the first two right and the last two take care of themselves. Skip them and no tool, not even AI, will save you. You can't microwave your way to a Michelin star. And you cannot explain to your leadership why simply adding a 7 figure AI tool to your Data Lake failed miserably.

We've all watched this movie too many times. This series is about writing a different ending together. Now that the car is on the road, buckle up and let's go.

Part 1 of The Governance Series by daitagov. First published on LinkedIn.