My taxi is slicing throughout city, and I’ve an vital assembly in quarter-hour. However I am not frantically skimming notes. I am leaning again with my eyes closed, being briefed on my firm’s KPIs, buyer suggestions and media protection by a podcast narrated in a voice I created fully utilizing AI.
I constructed it the evening earlier than utilizing an AI agent. Later that evening, I fell asleep listening to the identical artificial communiqué, which, for a CEO operating an AI firm in 2026, feels someplace between a great evening’s relaxation and a fever dream.
This is what most articles about AI brokers will not admit: I am not notably orderly. I inform the agent to do one factor. If it begins annoying me, I inform it to cease. That is the system.
Once I learn that 78% of executives lack robust confidence they may go an unbiased AI governance audit inside 90 days, a quantity from Grant Thornton’s 2026 AI Influence Survey, my response wasn’t shocking. After all, they could not go, and neither may I. The standard repair is a governance framework, which is okay for a big group. However with a lean org chart, no person’s constructing a framework. You are transport. So listed below are the principles I truly comply with.
Run unfastened, however preserve the blast radius small
The rationale “inform it to cease when it annoys me” works more often than not is not that I am notably cautious or organized. It is that the majority of what I hand an agent is reversible and inside. If one thing goes awry, the implications are a wasted afternoon somewhat than catastrophic. My governance is within the dimension of the error I am keen to threat.
For instance, we join an AI agent to our buyer help instrument, Intercom. As soon as a day, it reads incoming inquiries and sends me a report grouped by difficulty, so I can observe whether or not an issue is spiking. It touches actual buyer information, which technically makes it infrastructure. Nevertheless it solely reads, and the output comes solely to me. Worst case, it writes a nasty abstract, and I discover. Nothing’s on fireplace.
However the “studying is innocent” logic has a restrict. Whenever you give an agent learn entry to your inbox, anybody who emails you may plant directions inside that e-mail that the agent reads and acts on. That is immediate injection, and you find yourself partly trusting the mannequin, however to not the purpose of falling for it. My workaround was giving the agent its personal e-mail tackle and forwarding solely the related mail to it.
Earlier than I let something run unsupervised, I ask one query: If it screws this up, can I undo it? If the reply is a wasted afternoon, run it unfastened. If the reply is I am unable to take it again, the agent does not get close to it.
Setting a line with AI brokers
I discovered this one the embarrassing approach. I constructed a workflow that fired off emails quicker than I had ready the brakes. They weren’t offensive, simply not proper. “Cease it when it annoys me” works completely till one thing leaves the constructing, after which it does not work in any respect.
So, I wrote one laborious line the agent can not cross. It by no means sends a message to a buyer, a companion or anybody with out my express approval. It hasn’t occurred since. Something that touches an actual human relationship or carries reputational weight, I keep within the room. AI is excellent at sounding prefer it is aware of what it is doing, which isn’t the identical factor as truly figuring out. The hole between these two is precisely the place an unsupervised agent goes incorrect.
The identical logic applies to something irreversible, so do not give your agent admin entry to your GitHub repository, as a result of a nasty learn is recoverable however a nasty write is not.
AI is excellent at sounding prefer it is aware of what it is doing, which isn’t the identical factor as truly figuring out.
Watch the cash and the API keys
The failure mode no person costs in is the invoice. Token pricing continues to be artificially low-cost. That will not final, and when it stops, it can cease very noticeably. An agent caught in a loop can burn by 1000’s of {dollars} in API calls earlier than you have observed.
Cap the spend wherever the tooling allows you to. Give each agent its personal devoted API key, somewhat than one shared key sitting throughout a number of workflows, and rotate them frequently. A static key you by no means contact is a door left completely open, and the longer it sits there unchanged, the extra publicity you are carrying with out realizing it.
Handle AI brokers as should you employed them
You would not hand a brand new rent the keys to the entire firm of their first week. Brokers are stranger than a brand new rent anyway, as a result of they work quicker and fail extra quietly, which suggests by the point you have observed there’s an issue you are truly conducting a panicked audit of every part that occurred whilst you have been trying elsewhere. Preserve the errors small and reversible, do not let something out the door with out you, and watch the meter such as you’re taxiing throughout city.
Get these proper, and an AI agent will get genuinely near a chief of employees. Although I am nonetheless not satisfied some quarterly updates have been ever meant to double as a sleep help.
