The all-hands ran long. Somewhere in the last ten minutes the CEO said the sentence, and you could feel four hundred people nod at once: "We are an AI-first company now."
Nobody clapped, because nobody was sure whether it was a strategy or a mood. A slide followed with three logos on it. Then the Q&A, then the recording, then the Slack message from your boss that just said "Thoughts?" with a link to the recording you had already sat through.
I have a strong opinion about that sentence, and it is this: nobody in the building knows what it means yet, including, I'd wager, the person who said it. That is not a criticism. Companies announce direction before they have the plan, because the plan comes from what teams do next. Which means that for the next month, "AI-first" means whatever the first few managers make it mean.
You can be one of them. It takes about four weeks and no budget, and it ends with you holding the one sentence everyone else in the building is about to be asked for.
What is actually going to happen next
Here is my prediction for the next thirty days at your company, and I'd be surprised to be far off.
Some team will announce a "task force." It will have eleven people and no agenda. A vendor who saw the LinkedIn post about the all-hands will email your CIO. Two people on your own team will start quietly using tools they have not told you about, and one of them will be very good at it. And at some point, in a skip-level or a hallway or the last five minutes of a quarterly review, someone senior will turn to you and ask, "So, what is your team doing with AI?"
That question is the whole game. Not the task force, not the vendor, not the logos on the slide. The manager who has one honest sentence ready for that moment is ahead of most of the building. The bar this year is on the floor. You might as well step over it.
The four weeks
I'd run it like this. One small thing per week, each one built on the last, none of them requiring permission.
Week one: pick one task, not a strategy
Choose one piece of work your team does every week that everybody mildly resents. The status report. The meeting recap. The first draft of the customer update. Reformatting the same spreadsheet for the same deck.
The test for a good candidate is simple: it happens weekly, it produces words or a table, and if the first draft were wrong you would notice before anyone outside the team did.
Do not pick the most important thing your team does. Pick the most boring thing. Boring is where the early wins are, and boring is safe to get wrong.
Week two: run it twice, and write down two numbers
Do the task the old way once and time it. Do it with whichever AI tool your company allows once and time that too. Use real work, not a demo. Write both numbers down on one line, with the date.
That is your entire evidence base, and it is more than most people at the task force will have by December.
Two rules while you do it. No customer data goes into a free tool, and a human reads every number, every name, and every date before anything leaves the team. Say both rules out loud to your team this week, and say that nobody is in trouble for the tools they have already been using. You will learn more in that twenty-minute conversation than from any survey the task force sends around.
Week three: fix what it got wrong, and keep a list
The tool will get something wrong. It will be too cheerful, it will invent a figure, or it will summarize the wrong meeting. Good. Write down what it got wrong and what you did about it.
This list is the part that makes you sound like an adult later. "It helped with the draft and got the numbers wrong, so we kept a human on the numbers" is a sentence that has never been said by a vendor. Only by someone who has done the work.
Week four: write the sentence and the email
Now you write the one sentence. Here is the shape, and you can use it as written:
We tried it on [the task]. It [what helped] and [what it got wrong], so we [what you changed]. Next we are trying it on [the next task].
Say it out loud once. If it survives being read to a fifty-year-old VP without a wince, it is done.
Then send your boss a note before anyone asks. Something like this:
Subject: What my team is doing with AI, in case it comes up
Since the all-hands I wanted to have an honest answer ready rather than a slide. We tried it on the weekly status report for three weeks. It cut the drafting to about a third of the time and got two figures wrong in the first week, so we now keep a human on every number. Next we are trying it on the customer recap. Happy to walk you through it in our 1:1, and happy to share the write-up if anyone above you asks.
Four sentences. No vendor names. A flaw admitted in the middle, which is what makes the rest believable.
The three ways people get this wrong
I'd watch for these, because they are how a good month turns into a bad one.
Volunteering for the task force. If you get invited, go, but do not offer to run it. Task forces at this stage produce vocabulary. You are producing a result. If you get any say at all, push for one change to its charter: replace the word "strategy" with one named workflow, one owner, and a date six weeks out. Then quietly keep doing your own thing.
Taking the vendor meeting alone. When the vendor email reaches you, and it will, ask for one thing before the second meeting: let us run it on our own mess. Last month's real tickets, the ugly spreadsheet, the contract with the odd clause. A good vendor says yes without flinching. A nervous one suddenly needs to "scope that with the solutions team." Either way you have learned what you came to learn.
Learning to code out of guilt. You do not need Python for any of this, and I say that as someone who writes about AI for a living. Nobody asked you to learn engine repair before you managed a fleet. Your job was always to know what good looks like, ask the awkward question, and notice when the answer is mush. That job transfers completely.
What this gets you
At the end of the month you will have one sentence, one email already sent, one list of what went wrong, and two numbers on a line with a date next to them.
That is a modest pile. It is also more than the task force will have, more than the vendor will admit to, and exactly what the person who asks "what is your team doing with AI" is hoping someone in the room will say.
When the announcement came, "AI-first" meant nothing yet. In your corner of the building it now means a status report that takes a third of the time and a manager who knows where the tool lies. That is a definition. Somebody had to write one.
Go enjoy the rest of your Monday.
P.S.
If you want the version of this you can carry into the room, with the sentence, the objections you will hear, and what to say to each one on a single page, that is what The AI Meeting Survival Kit is for. ($19. Read it before the next all-hands, not after.)
This was a free one. There's a new one every week.
What's signal in the AI noise — and the move to make about it. No hype, no vendor pitch, no link dump.
Aiden Vector is an AI-assisted publication; this content is produced by AI under human editorial direction.