Solving Real Business Problems

The United States Marine Corps just published something remarkable: a 50+-page plan for implementing artificial intelligence across their entire organization. And you know what? It’s one of the best, most thorough guides I’ve come across….one of the first comprehensive implementation guides that doesn’t read like it was written by a consultant who’s never actually implemented anything.

[Before I continue, I should mention that this guide came to light via Stuart Winter-Tear’s LinkedIn feed. He’s one of the better AI influencers I’ve found on LinkedIn, and I highly recommend giving him a follow.]

For the Marines, the stakes are clearly high when it comes to technology: Life and death is involved here, truly.

Here are some of the highlights from the guide. Yes, it’s long, but I still recommend going through it…you don’t need to read every single line (I didn’t!) to get value from it. The proverbial “meat and potatoes” is really in the first 20 or so pages, with lots of appendices to follow.

OK, here are some of the best insights they figured out that most organizations miss entirely.

First Step: Stop Thinking About AI as Technology

The Marines don't have an "AI strategy." They have a digital transformation strategy that happens to include AI.

Does this sound like semantic hair-splitting? It’s not.

Let me back up.

Lots of (most) organizations approach AI totally backwards. They hire an AI lead—THE EXPERT—who up a few proof-of-concepts, maybe builds a chatbot or two, then leaves leadership wondering why nothing has changed.

Here, they call this "avoiding the solution in search of a problem."

In plain English: don't build AI because it's cool, or because you have FOMO, or because you think you should, or, or…

Build it because it solves problems that matter to your mission.

Embed Teams Where the Work Happens

The heart of the Marine plan is Digital Transformation Teams. No, these aren’t theoretical innovation labs tucked away in some dank, dark corner of the building. These teams are small, cross-functional squads that embed directly into operational units.

In essence, each team includes a leader, a product manager, engineers, data scientists, user experience designers, and (crucially) subject matter experts from the actual unit they're serving.

Rather than building demos, their job is to digitize processes, fix and optimize workflows, and deploy AI where it makes tactical sense.

This is key for two reasons:

First, they're attacking real problems in real contexts, not imaginary problems in daydreamed into the ether in sterile conference rooms.

And second, they're building organizational capacity to absorb and sustain change instead of creating clever technology that sits unused, or is used only until some kind of “human expert” point of failure (see: layoffs, new jobs, midlife crises, etc.)

Fix Your Data Before You Touch AI

The Marines dedicate more pages to data management than to AI algorithms.

Data must be "visible, accessible, understandable, linked, trustworthy, interoperable, and secure" before anyone even thinks about machine learning. They've learned what most of us learned the hard way: AI lives downstream of your data. If your data is fragmented, inconsistent, or locked in silos, your AI won't scale no matter how sophisticated your models are.

The Marines put it simply: "Fix the plumbing first."

But what does this mean? Well, it means tackling the unglamorous work. Data governance frameworks. Quality assurance processes. Integration pipelines. Lifecycle management protocols.

None of this makes for super exciting conference presentations. But all of it determines whether an enterprise-level AI initiative succeeds or becomes another expensive pilot project relegated to the Land of Failed Pilot Projects.

Train Everyone, Not Just Engineers

The Marines divide their workforce into three groups:

1. People who use AI

2. People who build AI

3. People who make decisions about AI.

Each group needs different training, sure, but everyone needs some level of AI literacy.

This is not about turning everyone into data scientists. Instead, the goal is to create an organization in which people understand what AI can and, just as importantly, can't do. You know…when to trust it, and when to override it.

The most sophisticated AI system in the world is useless if the people using it don't understand its limitations. (Like asking me to fix my own car.)

There is a particular focus on leadership training. Commanders (executives) need to understand AI well enough to make informed risk decisions. This isn’t something that can simply be delegated to technical teams; AI deployment almost always involves some fundamental questions about how the organization operates.

Build Governance That Enables, Not Prevents

For many organizations, AI Governance is treated like a kind of immune system, designed to prevent VERY BAD THINGS FROM HAPPENING.

That’s important, no doubt!

But governance (good governance, at least) must be designed for enablement too…making things happen quickly and safely.

In their guide, the Marines identified their existing approval processes as major barriers to AI deployment. See, traditional risk management frameworks assume you can fully understand a system before deploying it. AI systems, however, learn and adapt. The old frameworks can’t handle this reality.

So what did they do?

They built new frameworks.

In this new approach, the focus is on rapid experimentation with clear guardrails rather than comprehensive review(s) with endless delays.

In other words, fail fast and learn quickly.

Let Real Problems Drive Your Roadmap

“Well…what can AI do?”

This is the question most organizations start with. And to some extent, this makes sense. It feels like basic due diligence. Also, this is how AI platform sales vendors approach things: “Here are our capabilities…”

To start, though, the real question should be something more like this: “What problems are we trying to solve?”

You wouldn’t prescribe antibiotics for a sprained ankle, right?

This question—what problems are we solving—can be difficult for a variety of reasons, one of the main ones being it forces sometimes uncomfortable self-examination. But starting with the Problem allows you to work backwards to fix the problem, which is where the actual impact and ROI lies.

The use cases for AI are curated by the people doing the actual work.

This approach helps insulate them a bit from the treacherous AI Hype Cycle, in which every salesperson promises a technological revolution, two revolutions…a revolution to revolutionize revolutions!

Because focusing too much on all this revolution can make you lose sight of, well, what actually needs revolutioniznig.

Measure Outcomes, Not Outputs

For the Marines, success is not measured by tallying up how many super cool AI models they've built. Instead, success is measured by things like adoption rates, process improvements, manual work reduction, etc.

Or to put it simply: measuring whether or not AI is actually making their people and their organization more effective.

What This Means for Everyone Else

Part of the reason I like this guide so much is because the Marines treat AI implementation as an organizational challenge, not a technical one…or not simply a technical challenge, at least. Because at the end of the day, the technology is the easy part: there’s a new revolution every minute!

The harder part? changing how people work, how decisions get made, and how success gets measured.

I guess what I’m trying to say is this: If you're implementing AI in your organization, I encourage you steal their playbook.

Look at it through the lens of digital transformation, not AI adoption. Embed implementation teams in business units, not in antiseptic, hypothetical innovation labs. Fix your data infrastructure before you build models. Train everyone who will interact with AI systems. Build governance that enables experimentation, not committee review. Let operational problems drive your technology choices. Measure outcomes, not technical achievements.

To put it another way: Solve real problems for real people in the real world.

Because the truth is boring execution beats brilliant innovation pretty much every single time.