Start With Why AI

Stop poisoning your business with AI solution it doesn’t need

Cassie Kozyrkov
4 min readAug 13, 2024

A little while ago, I posted about the irony of so many enterprise leaders coming to me for AI advice when, in reality, AI wasn’t what they needed at all. Luckily, I’m equipped to help with all manner of complex decisions. But let’s address the AI elephant in the room.

Let’s address the AI elephant in the room. The fastest way to derail enterprise value is to throw AI at problems that are either poorly defined or better suited to non-AI solutions — solutions no one took the time to consider.

The business challenges these leaders face are complex and significant, but more often than not, traditional programming approaches — offering greater control and reliability — would be better suited to solve their issues.

I do indeed believe this is the best image AI has ever generated for me, it gets funnier the longer I look at it.

What followed were hundreds of comments of a particular flavor. This one sums it up nicely:

‘So often, clients ask me “What kind of AI can I use?” And the answer is, “You don’t need AI for this, and it will be counterproductive for your aim[s].” But the response is always, “But we won’t get any funding if our product doesn’t have AI in it!’

This reflects a growing trend: leaders and managers feel pressured to “find” some AI, regardless of whether it’s the right solution. As I mentioned in the original post, the fastest way to derail enterprise value is to throw AI at problems that are either poorly defined or better suited to non-AI solutions — solutions no one took the time to consider.

How to combat this? Start with the business problem you’re trying to solve.

Start with why AI

You’ve probably heard that famous Simon Sinek quote, “Start with why.”

Sinek’s advice is geared more toward finding purpose in business, so allow me to offer my own spin on it: “Start with why, then why AI.”

“Start with why, then why AI.”

Imagine, for example, trying to write a program that will teach a computer to identify a cat. The subtle nuances and edge-cases are endless; it’s nearly impossible to solve it with traditional programming. So instead, Fei-Fei Li created ImageNet and pioneered the practice of teaching a computer through examples rather than instructions. Millions of examples later, a computer could identify cat/not cat.

But there’s always a price… even beyond the wild computing costs involved. And that price is mistakes.

AI isn’t a silver bullet. I’ve often said that Enterprise AI is like medicine: AI is like medicine — it can be a life-changer to those who need it, but everyone else should know better than to snack on it out of boredom. (Or board-om. Yes, I know there’s so much top-down pressure from boards these days.)

AI is like medicine — it can be a life-changer to those who need it, but everyone else should know better than to snack on it out of boredom.

AI is not magic. In fact, AI will always introduce complexity and remove reliability relative to the traditional approach. It should only be used when the problem you’re solving is so intricate that you can’t possibly solve it the old way, by defining the solution in clear-cut, written instructions (otherwise known as code.) And if you’re going to use it, the project leader needs to be extremely well-versed in how to play their role.

The only time it’s safe to ignore this advice is when failure is a really good option, which is why spam bot farms do so well: they’re in the business of garbage at scale that doesn’t need to work every time to work overall. But even here, the leader is key. (Not that I’m condoning this kind of pasttime. Just mentioning it so that you don’t get to thinking that the leader gets a free pass.)

To make pretty much every other class of use case work, you need skilled participation from someone who’s trained for the decision intelligence part. But the great news is that you’ll begin to see great use cases everywhere in your business after you learn AI strategy.

So stop solving the wrong problem with AI and start creating impact through revolutionary automation.

Developing your AI strategy

How do you identify use cases suitable for enterprise-scale automation? And when you’ve identified those use cases, how do you design, test, and deploy AI systems so that they drive growth and minimize risk to your organization?

My new course walks you through a 12 Step Framework to teach you all this and more. You’ll learn how to:

  • Identify the best use cases for planet-scale automation vs individual AI
  • Understand different approaches to AI strategy
  • Create a shared vocabulary across your team
  • Avoid the phenomenon of ‘Death by a Thousand Pilots’
  • Assess AI vendors
  • Manage complexity
  • Defeat the ‘Four Horsemen of the Enterprise AI-pocalypse’
  • Build safety nets for safe and effective AI systems
  • Use data to ask the right questions and make better decisions
  • … and much more

The next cohort runs on August 23 and 24. You can sign up for the course here: https://maven.com/cassie-kozyrkov/enterprise-ai

--

--

Cassie Kozyrkov
Cassie Kozyrkov

Written by Cassie Kozyrkov

Chief Decision Scientist, Google. ❤️ Stats, ML/AI, data, puns, art, theatre, decision science. All views are my own. twitter.com/quaesita

Responses (21)