What Tony Stark Understood
The language we are settling on for AI is quietly redefining the people who use it. "Human in the loop" gets it backwards.
At a panel session at MSP Global this week on how AI can help managed service providers, the thinking was sharp and the enthusiasm real. But two words kept recurring that I have not been able to let go of since: people described as “humans”, kept “in the loop”. Both are worth stopping on, because the language we settle on for AI will quietly shape how we treat the people who use it.
Start with the first. They are not “humans”. They are people. It sounds like a small thing, but calling someone a “human” tells you whose point of view the sentence is written from. It is the machine’s. In that framing, the people in your business have become one more category of input to the system, sitting alongside the data and the prompts. The word choice gives the game away.
They aren't "humans". They're people. The word choice gives the game away.
The second phrase does the rest of the damage. “Human in the loop” sounds like a safeguard, and in its original, narrow sense it is one. But as a way of describing the role a person plays, it relegates them to a machine operator, there to push the button the system points them at. With a technology as powerful as AI, that is not a cost-free piece of jargon. A disengaged operator rubber-stamping an AI’s output is not oversight. It is how a bad decision gets waved through, into your business and out to your customers. And it drains the work of any satisfaction, which only makes the rubber-stamping more likely.
A disengaged operator rubber-stamping an AI's output isn't oversight. It's how a bad decision gets waved through.
The way I think about it is Tony Stark and Jarvis. They are a team. Tony brings the ideas, the direction and the judgement; Jarvis brings the knowledge, the skill and the reach. Tony leads, because the ideas and the responsibility are his, but together they build what neither could alone. And to be clear, this is not about treating AI as a colleague. Jarvis is a machine, and everyone watching knows it. The partnership is in what the two of them achieve together, not in any pretence that the tool is a person.
That is the relationship I want between my people and the tools we give them. Not a person kept in the loop of a process they do not drive, and not a person reduced to issuing instructions to a tool, but a team: capable people paired with the most powerful instrument they have ever had, leading it rather than operating it. The person sets the direction and owns the call. The AI helps them explore further and build faster inside it.
Lead it, don't operate it.
There is a hard-edged business case underneath the principle. The engaged person, the one whose ideas and judgement are genuinely in the work, is also the one who catches what the rubber-stamp misses. Ownership, not supervision, is what makes oversight real. Treat your best people as button-pushers and you lose the very judgement you most needed from them, at exactly the moment the technology makes that judgement most valuable.
That is how I use AI, and how I encourage my team to. Give capable people a powerful tool and let them lead it, and they do more, they explore further, they enjoy the work more, and the business and its customers are better for it. The companies that win with AI will not be the ones that kept their people in the loop. They will be the ones that kept them in the game.
Toby Forman, Group CEO