Why most AI projects fail before they start


How not to do operations

[Setting: An exec team is meeting for their monthly metrics review. It is 4pm on a busy Thursday afternoon. Several people sit around a large oval conference room table. A few of them type rapidly at their laptops. The rest are rapt with attention, listening to the speaker.]

CEO: So where are we with AI?

[execs look at each other uncomfortably, 60 very long seconds of silence go by]

CTO [sheepishly speaks up]: Well, we've got some great features in the pipeline!

CEO: Now we're talking. What are they?

CTO: Uh, we're adding a write-it-with-AI button?

[long, silent pause]

CEO [slowly nodding]: Okay. Is that it?

CMO [interjects, to the relief of the CTO]: My team found a cool outbound email tool!

CEO: [...]

Rest of room: [...]

CHRO [cheerfully pipes up]: And my team loves ChatGPT!

CEO [exasperated]: But what's the strategy?

[Scene]

This scene is playing out in boardrooms everywhere right now:

  • The CEO says, "Okay everyone, time to get with AI!"
  • The rest of the leadership team scrambles to figure out how to inject AI into their usual systems and processes
  • Business results stay exactly the same

This is not how businesses thrive, but it is how nearly every company is approaching AI transformation. Like AI transformation is a goal unto itself, for its own sake.

It isn't.

You're missing the plot

A month ago, I made a popular LinkedIn post that asserted:

Behind every CEO saying "your job is to reinvent your operations with AI!" is a CFO saying "your job is to hit the same metrics with half as many people."

I was being kinda glib, but also kinda not really.

It's true that there are a few building-oriented people who can start with AI and think outward; they are able to begin with the technology and spot valuable use cases for it. If you're lucky, maybe you even have someone like this on your team. They automate things that are currently manual. They'll spot a bottleneck in your process and build a neat custom GPT to help.

These people are awesome and you should do everything in your power to keep them. But also, this isn't AI or system transformation, and if you think it is, you're missing the plot.

AI is the means, not the end

The most effective organizational and process reinventions always begin with constraint. This far predates AI.

Constraint is sometimes financial, like when that CFO tells you to hit your metrics with half as many people. But there are other kinds of operating constraints too, e.g.

  • Onboard three times as many customers per month
  • Review and return contracts 30% faster
  • Hire 500 engineers before end of quarter

Effective constraints are always (always!) rooted in your fundamental business performance: revenue attainment, customer satisfaction, speed of delivery, talent retention, cost. They are the metrics you already use to assess your business performance.

Great leaders know how to manage to this kind of operational constraint. When leaders are accountable to operate to the constraint, they explore the full range of tool, process, personnel, and system changes that it takes to deliver. This was true long before AI, and it will be true long after the AI hype has died down. Good leaders are open-minded and explore all fruitful possibilities. It's just that, these days, the possibilities often include AI.

If you're serious about AI transformation, start with clear business constraints that you're working in pursuit of. Leaders should be accountable for delivering the end result within the constraints, not for using a particular tool (AI or otherwise) to get there.

Kieran


From now until August 31, my Viral Data Stories 101 course is half-price for nerd processor subscribers. Learn to tell your own data stories clearly and effectively!

My latest data stories | Build like a founder | nerdprocessor.com

kieran@nerdprocessor.com
Unsubscribe · Preferences

nerd processor

Every week, I write a deep dive into some aspect of AI, startups, and teams. Enter your email address to subscribe below!

Read more from nerd processor

How the AI gold rush is going Last year, I wrote a popular piece about why it is hard to sell AI software. I discussed the AI gold rush dynamic, where AI companies initially find it easy to raise money but then much more difficult past Series A. In the situation I described, businesses didn't know how to buy AI tools yet, and for startups, competing with big incumbents was even harder than before. It's now one year later, and most organizations have figured out how to buy AI tools by now. As...

A change is gonna come Over the last three weeks, we've looked at career change patterns among executives. We saw that most CEOs and C-suite leaders of the AI 50 have changed careers several times, while Fortune 100 execs have changed much less often. We also saw that, among execs, career changers are more likely to hire other career changers, and that execs hired in the last 12 months are more likely to be career changers than in previous years. This got me curious about two things: We've...

The Emergency In the book The Mysterious Benedict Society, all the action takes place in the backdrop of The Emergency, a constant news cycle of one frenetic catastrophe after another. It's a fantastic book series, double plus recommend. But though I read it several years ago, I can't get this one detail out of my head right now. Because real life has never reminded me more of The Emergency. There are many ways people try to cope with a steady diet of stress and chaos: political activism,...