I’ve been through a few technology waves.
Not watching them from the sidelines, but actually building through them, shipping systems, running engineering teams, helping customers deploy systems, and the pattern is surprisingly consistent.
First comes the platform shift happens. Then comes the services wave that actually makes the platform usable. We’re at that moment right now with AI.
Every Platform Shift Looks Revolutionary at First
When I started in tech in the late ’90s, basic e-commerce still felt like magic.
The idea that you could sell something online securely and reliably was new. A lot of companies knew they needed to be doing it. But very few knew how to actually make it work.
Then mobile arrived. Software suddenly moved into everyone’s pocket and teams had to rethink interfaces, distribution, performance, and almost everything.
After that came SaaS. Software stopped being something you installed once and became something that continuously evolved.
Then the cloud changed the infrastructure layer underneath all of it. Servers became programmable, infrastructure moved from procurement cycles to API calls, DevOps replaced ticket queues, and speed suddenly mattered in a way it never had before.
Each of those shifts produced massive platform companies. But they also produced enormous services ecosystems around them, because once the excitement fades, companies run into the same reality: Knowing a technology matter is very different from successfully deploying it.
What Every Wave Had in Common
The gap between those two things is where services live.
Platform companies ship the building blocks — APIs, models, infrastructure, and tools. They move incredibly fast, and capabilities evolve every few months. Meanwhile, the average enterprise is dealing with very different questions:
Where does this actually fit in our business?
What systems does it connect to?
What data does it need?
How do we keep it safe?
Who owns it when it breaks?
None of those problems get solved by a model release. They get solved by people who understand architecture, workflows, operations, and what it actually takes to run software inside complex organizations. And solving those problems is where services come in.
Every time a major platform shift happens, the demand for that work explodes.
AI Feels a Lot Like Cloud Did in the Early Days
The AI landscape right now reminds me a lot of the cloud between about 2008 and 2012.
A small set of providers dominated the platform layer. The capabilities were impressive and evolving quickly with best practices still emerging. Most organizations were still trying to separate real opportunity from hype. I saw this up close during the early AWS days.
Back in 2012 at Sturdy, which later became Onica, we were helping companies move real systems into AWS when many enterprises were still debating whether the cloud was something they should even consider. The work wasn’t writing strategy decks, it was migrating workloads, refactoring applications, and building systems that actually ran in production.
That work created an entire industry of cloud services companies.
AI is starting to look very similar, but the scope is bigger.
AI Isn’t Just Infrastructure. It Changes How Work Happens.
The cloud mostly changed where software ran. But AI changes how decisions get made and how work gets done.
It shows up in product experiences, customer support, operations, finance workflows, internal knowledge systems, engineering productivity, basically anywhere humans interpret information and make decisions.
That’s a much larger surface area than infrastructure alone. And when you step inside real organizations, you quickly see that the hard problems aren’t model capability. They’re usually operational – the data is messy, the workflows exist in people’s heads, the systems you need to integrate with are 20+ years old, and compliance and governance are non-negotiables.
None of that shows up in demos, but all of it shows up in production.
Where Most AI Projects Actually Stall
When companies struggle with AI adoption and projects stall, it’s rarely because the model wasn’t good enough. The problem usually shows up earlier. Teams start with broad ambitions: “AI strategy,” “AI transformation,” or things like that, but never narrow down the work to a specific workflow that they can actually ship. Others build something impressive in a demo environment that never gets integrated into the systems people use every day.
Production systems require discipline, clear scope, real data, integration into systems of record, guardrails that make behavior predictable, and monitoring and evaluation once it’s live.
It’s not glamorous work, but it’s the work that turns capability into outcomes and something the business can rely on.
The Winners Will Be Builders
Every technology shift creates a lot of commentary. You see it today with AI: strategy frameworks, trend predictions, endless panels about what the future might look like, but they’re just not where the value gets created.
The companies that win these waves are the ones willing to ship. They start with a narrow workflow AND get something into production quickly, then they iterate and expand from there. And over time, the teams that build real systems develop something much more valuable than a point of view, they develop operational experience. That’s incredibly hard to replicate.
This Is Going to Be A Very Big Wave
AI isn’t a passing trend. It’s a structural shift in how software participates in work. That means the platform companies will grow very large, but it also means that the services ecosystem around them will be enormous, because organizations don’t just need access to AI, they need help turning it into something that actually works inside their business.
I’ve seen this movie before.
And the biggest opportunities usually go to the teams that stop talking about the future, and start building it.
If you want to move from AI interest to AI in production, start with one workflow, one measurable outcome, and a 4–6 week sprint that ships something real.
FAQs
Why is AI a services wave, not just a software wave?
Because the hard part is turning models into outcomes: workflow design, integration, guardrails, measurement, and adoption.
How is this similar to the cloud transition?
Early cloud required migrations, refactors, and new operating models. AI requires new decision layers, workflow redesign, and production discipline.
What types of companies will buy AI services?
Mid-market and enterprises that need AI embedded into real systems with governance, security, and measurable ROI.
What separates successful AI services firms from “AI consultants”?
Execution: shipping production systems, integrating into workflows, measuring outcomes, and hardening through iteration.
What’s the fastest way to start?
Pick one workflow with clear pain, define the outcome, and run a short sprint that ships a bounded capability into production.
