Is Your AI Task Force Still Meeting?

This week my website has me in a small wooden boat in a stormy harbor, holding up a laptop like a lantern, while something the size of a parking garage stands up out of the water behind me, because Godzilla Minus Zero opened on Friday (after $6.7 million in Thursday previews, which is a lot of people who apparently also enjoy watching a city get stepped on). Every image of me on there was generated by the AI that rebuilds the site at 5AM every Saturday, from my real headshot, and every one of them has my blue fedora in it, because a Roske never removes his fedora in the presence of others, and I’m counting a 300-foot lizard as others. (Edward, you’ve now spent an entire sentence establishing hat protocol for a sea monster, in a post that’s supposed to be about AI strategy… yes, I know, I’m getting there.)

Pretty much every one of these movies, going back to the 1954 original, has a scene where a dozen very serious people sit around a long table and discuss what to do about the monster. They form a committee and commission a report (and then argue for a while about whether to tell the public), and while they’re arguing, the monster is already in the harbor, and the whole audience is yelling at the screen, because the audience can see the window and the committee apparently can’t.

A lot of companies are in that scene about AI right now. They’ve formed an AI task force, it meets every other Thursday, and it’s currently in Phase 1 Scoping (it’s always Phase 1 Scoping). Meanwhile, the thing the task force is scoping walked into the building a while ago. Back in 2024, Microsoft and LinkedIn surveyed 31,000 people and found 75% of knowledge workers were already using AI at work, and 78% of those people were bringing their own tools to do it. Then last year, an MIT study found workers at more than 90% of companies using personal chatbot accounts for their daily work, while only 40% of those companies had bought an official subscription. So if your task force is still deciding whether to let AI in, I’ve got some news about the people in the cubicles outside the conference room (and the ones on their phones in the conference room, during the task force meeting, under the table).

I’ve been around for 4 of these landfalls in my career, so I’m not saying this to be smug (well, not only). I started out in 1995 doing Essbase consulting in the client-server days, when the scary new thing was a database that ran on a server instead of a mainframe. I founded interRel in 1997, right as the web showed up and everybody’s company needed “an internet strategy,” whatever that was. When the cloud showed up, I wrote a whole book about Oracle’s EPM Cloud (part of the Look Smarter Than You Are series), and I based it disturbingly loosely on The Martian, so the premise was that I was stranded and the only way home was to implement EPM all by myself, with a dated log entry at the top of every chapter like Mark Watney’s (I can’t believe I just told you that in a post about task forces, and I can’t believe that book got published, and yet both of those things happened). AI is the 4th one, and it’s the biggest by a mile, and it’s moving faster than the other 3 put together.

And in every one of those, the companies that came out fine weren’t the ones with the best committee. I’ve told the story before about the CEO who held up a methodology book in 1997 and thanked it for the company’s great year, and how I leaned over to Eduardo Quiroz and asked whether, if we all quit, the methodology would implement itself. (I never found out. We quit. I’m still a little proud of that question.) The same question works on an AI task force: if everybody on it went home tonight, would the AI strategy implement itself? If the honest answer is “nothing would change, because nothing has happened yet,” then what you’ve got is a committee, and I’d try asking it that question at the next meeting, if you’re feeling brave.

So who actually saves the city? In the good monster movies, it’s a handful of ordinary people with the right skills (somebody who knows boats, and somebody stubborn enough to keep asking questions when everybody else wants to adjourn) who get organized around one specific job, and as far as I can tell it’s never the generals. At a company, that looks roughly like this: find out who’s already using it, and don’t punish anybody for the answer (the people with personal ChatGPT accounts are your scouts, and they already know which jobs it’s good at), then give everyone an approved tool this month with sensible rules about what data goes in, because a ban mostly moves the AI onto personal phones where nobody can see what goes into it, then pick 1 real problem with a number on it, something you already measure (hours, errors, turnaround time, tickets per week), so in 30 days you’ll know whether it worked instead of having a feeling about it, and put the people who do that job every day in charge, since they know which part of it is worth automating (it’s usually the part they complain about at lunch).

After that, connect it to your real data. An AI working from memory will give you a number that sounds about right. One that can read your actual systems can give you the actual number, which is why I’ve been coding the plumbing for exactly that (AgentCube, if you’ve got Essbase or Oracle EPM). (Edward, you just plugged the software you coded in the middle of a post about a 300-foot lizard… yes, I know, but it’s relevant, so I’m leaving it in.) Then put a date on the calendar, and when it comes, keep it or kill it based on the number (or roll it out wider, if the number’s good), and whoever liked the demo most gets exactly 1 vote, same as everybody else.

You’ll probably need a budget for some of it, and you’ll definitely need somebody willing to go first, which is honestly the hard part.

I’ll go ahead and put a date on this one: by this time next year, “we’ve formed an AI task force” is going to sound the way “we’ve formed a Y2K committee” sounded in 2001, which is to say, like something you’d only bring up if somebody asked what you did back then. I could be early on the timing, and if so, write me next November and tell me your task force made it to Phase 2. I’ll send flowers.

If your task force is meeting again on Monday, feel free to forward them this (anonymously, I won’t tell anybody where you got it). If you’d like a room full of your leadership team to hear it out loud, with slides and a fedora, that’s what the speaking page is for. Or just write me at Edward@Roske.AI and tell me what Phase 1 is scoping this month. I promise I’ve heard worse, and I’d genuinely like to know!

Asking good questions, Edward