The coming agentic AI services boom

Are we about to see a massive technology services boom from agentic AI?

The mainstream and industry press tell us that AI is going to cause massive job losses. And in part, I agree. AI is already reshaping the division of labor, automating routine tasks, and performing many forms of complex intellectual work.

New jobs may emerge, as they have after every technological revolution. But disruption and displacement are going to happen en masse. I do not know a single serious person who uses AI regularly and disagrees.

But there is a big difference between predicting mass automation and being the person responsible for implementing it.

When you stand at a whiteboard or sit at a keyboard, and accept accountability for an AI project and its results, the world suddenly looks a lot different.

Let's say you are going to deploy AI agents to perform a complex business process. The goals: reduce staffing levels, cut costs, reduce error rates, and maintain or improve customer satisfaction.

At a high level, what will your project plan look like?

First you need a strong understanding of the existing process, and how it might be reshaped or improved.

Then you need to select the right tools and architecture, from the current explosion of AI automation vendors.

Let's assume the tools you select do most of the build — a current strength of emerging AI solutions.

Now you have to deploy into production. For an agentic solution you'll need to consider orchestration layers, memory/state management, human-in-the-loop handoffs, API integration at scale.

Next you'll need to consider security. Or rather, you'll realize you should have considered it from the beginning. Your agentic AI will be subject to prompt injection, tool misuse, privilege escalation, and data exfiltration. Plus attacks waiting to be invented, discovered, or popularized.

Once in production, it will require constant monitoring. What decisions is your agent making? What actions is it taking? Is behavior drifting? If your agent is behaving oddly, can you determine why? Can you undo errors and restore systems to their previous state? Do you have forensic capabilities?

Are you resilient?

If and when AI starts making poor decisions — or fatal ones — how are you going to respond to the regulatory and political pressures that follow? Even if your own systems are not at fault?

You can delegate any of the above to AI. But ultimately you'll be accountable to other human beings, and will need to demonstrate and explain how you got the results you did.

Yes, we may be heading for mass job losses. But getting from A to B?

As my good friend Pavel Tsatsouline says, "The confidence of amateurs is the envy of professionals."

We are about to see a boom in AI-related technology services. Be ready.