Want AI for your business? Start here.
There are multiple ways to get your company started on its AI journey. The simplest: hand everyone access to ChatGPT, Microsoft Copilot, or whatever tools you're already using, turn on all the AI features, and call it a day. But let's focus on AI projects meant to solve one specific business pain. Where do you start so you can get your feet wet without drowning?
Here are the questions we'd work through with a client:
Is there a burning problem?
Something constantly draining your resources, or massively getting in the way? Why does it need solving, and what would "solved" look like and unlock in your business?
This avoids doing AI for AI's sake—applying it to something just because you can, or optimizing an area that doesn't need optimizing.
If you don't have such problems, don't be sad you can't throw AI at them. Be glad you don't have big problems. ("I really wish my house was on fire so I could use this cool firefighting robot I built…")
What have you considered so far?
Assuming you have a burning problem, what solutions have you attempted, and why didn't they work? Have you tried more people, off-the-shelf software, consultants? What were the outcomes?
If you haven't tried anything, we have to question whether it was actually a big problem. It's possible you haven't tried because, until AI, it was considered intractable—something you just lived with. Fair enough. If you have tried things and they've failed, that gives important intel into the assumptions, surprises, and unexpected twists. Even a failed attempt provides valuable learning.
Can we identify a small, self-contained piece that delivers some value?
Big burning problems that defy simple solutions often span multiple areas. Still, for an initial project, find an initiative with a small blast radius. Solving the entire problem might require a huge effort, but we can define something with limited scope and reduced features to prove out value and build momentum.
The pitfall: something that's only useful once all its many components have been built. Be suspicious of long projects with late payoff. Seek out short projects with early payoff, even if small. The 80/20 rule works in our favor—80 percent of value unlocked with 20 percent of effort. And after that, we don't slavishly work until 100% completion. We inspect, adapt, use feedback from a successful rollout to define a new project where we play the 80/20 game again.
If you're on board in principle but find this too generic, let's have a chat about how this could work for your business. And if you know someone curious about where between AI hype and AI doom there might be some value for them, send them my way.
