In March 2002, after some prodding from my friend and coworker Topher, I published my first blog post. It was about Flash, the shiny new toy every web designer wanted to play with. Splash pages. Intros that took a ghastly amount of time to load on a P133 over a 28.8k connection.
The post was titled “Caught Flashing?” Its main point holds up better than its title:
“As with many exciting new technologies, people stampede to be first in line to use it often without asking the most fundamental question; why?”
Swap Flash for AI and you could publish that sentence this morning. That’s a little embarrassing for our industry. It’s also a little reassuring, because the questions that kept me out of trouble in 2002 still work.
Three lessons have followed me the whole way. Each one has a moment from before AI and a moment from building AI strategy assistants, and in at least one of them I’m the person who forgot the lesson.
Ask why before how
The 2002 version was simple. Before you build the splash page, ask why you need one. If the answer supports what you’re trying to communicate, go ahead. If it doesn’t, skip it.
The AI version is where I tell on myself.
I was building an assistant aimed at business model innovation. I’m deep in Strategyzer’s methods, so I knew the territory. Then we started testing. As I watched people work with it, I kept hearing the same thing: “I guess I don’t know what a key activity is.” That’s business model language, and without some reps or experience, nobody has a way of knowing what it means. Most users didn’t have a working vocabulary for business models.
That gap turned out to be a binding condition. We couldn’t design around it. No clever prompt was going to teach someone a new vocabulary in the middle of a strategy conversation.
So the question became: why would someone start here at all?
They wouldn’t. So we changed the approach. I moved the assistant from business model innovation to disruption and built assessment tools that help a leader see where their business is under threat. That threat turned into the on-ramp. Once someone can see where they’re exposed, thinking about their business model differently stops being an academic exercise.
I’d like to say I asked “why” before I wrote the first prompt. I didn’t. Testing forced the question for me.
The user isn’t always the customer
A few years back I watched a B2B salesman at a distribution company drive forty minutes across town to help a customer load two trucks, while the retail side of the same company treated that customer as a workflow exception. The retail experience had been optimized for the person standing in the store. The salesman was serving the business owner whose quarter depended on that order. Same customer, two very different ideas of who mattered.
Drucker would say the purpose of a business is to create a customer. In UX we tend to design for whoever is touching the screen. Those are often different people, and it’s easy to forget which one you’re serving.
Strategy tools have the same split.
The person in the chat is usually one leader who has taken some ownership of a problem. They use the assistant to sharpen their thinking, work through the process, and get to a choice. But strategy touches every part of a business, and nobody makes a strategic choice alone. At the end of the day they have to bring their team along.
Those people weren’t in the chat. They didn’t co-create anything. They’re seeing the choice cold.
That’s why so many of the assistants end with an artifact built for that team. It names the choice at stake and the choice being made. All the thinking that happened in the conversation has to be explicit and written down, so someone can come up to speed and see the rationale without replaying the whole exchange.
The person typing is the user. The team they have to convince is the customer.
Decide how you’ll know
At Gordon Food Service, I led UX on a digital product team carrying a compounding target: grow adoption of a $4 billion B2B ordering platform 20% quarter over quarter. The team hit it three quarters in a row, roughly 73% cumulative growth. The number was our feedback loop, and it showed up every quarter.
The AI work has been humbler about this. Our feedback loop was manual and still is. We have usage numbers, and they show that some assistants get far more engagement than others.
The most-used ones are the assistants that teach. The suite puts a strategy team’s expertise in the hands of people who are new to the discipline, so a lot of strategy language has to be explained first. The busiest assistants show people how to think about strategy and business models in terms that make sense in their own professional services work. From there, people move into the deeper, more specialized tools.
The signal I trust most is whether the person will take the artifact the assistant produced and put it in front of their team as the defense for a specific choice.
If they won’t attach their credibility to it, the artifact isn’t pulling its weight.
That’s a harder test than a usage chart, and a more honest one. A tool can get opened every day and never change a decision. The test I care about is whether someone was willing to stake their name on what came out of it.
Still asking why
When I reread that Flash post now, the technical advice is ancient. Nobody’s worried about the plug-in or the 28.8k modem anymore. The questions underneath are the ones I’m still asking: why are we building this, who is it really for, and how will we know it worked.
AI made the material faster and stranger to work with. Those questions still come first.
I just hope I’m a little quicker at asking them than I was in 2002.
