Forget the team of AI agents. Think software.
There is a lot of terrible AI advice right now, and most of it is about agents. If you think the answer for your business is a team of AI agents, you are probably missing the signal for the noise.
I have spent nearly two years helping small and medium-sized companies put AI to work. From that vantage point, one thing is clear: we are a long way from effective teams of AI agents doing real work on their own. The demos are impressive. The reality, in ordinary businesses with messy data and real consequences, is not there yet.
More than that, I think the agent framing often hurts more than it helps. It invites you to imagine hiring digital employees, which sets the wrong expectations, points you at the wrong problems, and quietly wastes money on the wrong things. It is a compelling story, and it is mostly not what your business needs this year.
What is actually happening
AI is not replacing your workers in the short term. That is the honest position, even if it is less exciting than the headlines. What AI is doing is changing the economics of software.
A custom app or automation that would have been too expensive to bother with now moves onto the worth-doing pile.
You still, in most cases, need someone with expertise to build it well. I will admit a bias here, since that is what we do. But the point stands: the cost of building the right thing has fallen dramatically, and that changes what is worth building.
Software, solutions, and services
So forget the AI worker framing entirely. Three plain questions get you further than any amount of agent talk.
What software are you overpaying for?
Clunky, bloated SaaS you tolerate because switching hurt and building your own was out of reach. That calculation has changed.
What do you need that does not exist yet?
What you cannot buy off the shelf, or that exists but does not quite fit how you work. The gaps you have learned to work around.
Is it time to revisit build vs buy?
For a long time the answer was almost always buy, because building was expensive. For the things that matter most, that is worth reopening.
And then: who do you need?
Finally, the human question, which the agent framing conveniently skips. What people do you need to help you make and implement these changes? The technology is rarely the hard part now. Deciding what is worth doing, building it well, and getting it adopted still takes judgment and people. That has not changed, and I do not expect it to soon.
Stop thinking about hiring AI agents. Start thinking about software you overpay for, needs you cannot buy, and the build-versus-buy calls worth reopening now that building is cheap.