The Box Nobody Knew What to Do With

We put an open box in front of people this summer and told them to ask anything.

They stared at it.

Nobody could think of a question worth asking.

That's the whole problem with most AI rollouts.

I keep having the same conversation with business owners, and it often starts the same way: somebody tells me their team is falling behind and they know that it's on them.

Here's my honest read after years of building this stuff inside real companies.

It has nothing to do with which AI tool you picked. The answer is connecting AI to the work your team already does every week.

Forget the prompts you use or the tool you picked. We're talking how you implement.

Let me show you what the 5% are doing.

🧱 Outlive Every AI Tool

The software changes every few weeks. These methods haven't changed once.

Key Facts

🔁 Repeatable beats shiny - Everything I set up for a client this month did not exist a year ago, and the process underneath it has barely changed.
🔌 It only knows what you show it - AI with no connection to your systems knows one thing, which is whatever somebody typed into the box that session.
🎯 Narrow beats broad - Simple, specific tools keep beating do-everything assistants, in my builds and in the research.

You might feel behind.

You bought tools and never changed how the work gets done.

You have the accounts, in fact, somebody on your team has three+ of them.

You watched the demos, sat through the webinar, forwarded the article to the group chat.

And the work still moves through your business exactly the way it did eighteen months ago, with people copying information from one system into another, rewriting the same five emails every week, and digging through old meetings for an answer somebody already said out loud.

MIT looked at 300 companies that rolled out AI 🔗 plus interviews with 52 executives and a survey of 153 leaders.

They found that 95% of those rollouts made no measurable difference to the bottom line, after $30 to $40 billion in spending.

The 5% who got paid were using the same software as everybody else.

⏩ Moore's Law

Nearly everything I build today is old in 6 months, and that used to bother me.

But now I build with that in mind on purpose.

I stopped designing for features and shiny releases and started creating around six questions that hold up no matter whose logo is on the software.

Where does the information live?
Who is allowed to see it?
What work happens every single week?
What does an excellent result look like?
Who reads the output before it goes anywhere?
What happens next once the output exists?

Answer those six for one workflow and you own something that lasts, because the AI underneath becomes a part you can swap out, which is exactly what it should be.

McKinsey surveyed 1,719 executives 🔗 across 97 countries and found only 6% of companies credit AI with at least 5% of their profit.

The one thing that separated them was rebuilding how the work flows, because adding a tool on top of the old process did nothing.

💻 Your Prompt Was Never the Problem

Then there's the debate on the prompt. Everybody wants to argue about how to write prompts, but the problem starts long before the prompt.

If AI has no connection to your meetings, your customer records, your project board, or your inbox, it doesn't know a damn thing about your business.

All it knows is what you've pasted in, uploaded, or reminded it for the 19th time.

And the fear that runs the other way costs just as much, because I talk to owners every week who avoid connecting anything, worried the AI will read all their emails and train itself on the company.

That is not how it works.

It looks at what a specific job needs, when it needs it, and only where you gave it permission.

Here's how that plays out for me. I have my AI go back through my client meetings and pull out the themes I keep repeating to people, and those themes become some topics for this newsletter.

This edition came out of that.

If I'm saying something out loud twelve times a month, that's a topic, and pretending otherwise is just ego wasting time.

🎯 Ask Me Anything Is a Bad Idea

Going back to the box…

Once we swapped the open question box for the questions people were already asking and let them pick one, the whole thing worked better by that afternoon.

Imagine how much better your customers will feel when the work you're doing supports them individually, not generically.

Every open box has three or four specific jobs hiding inside it, and those jobs are the actual product. It's more about how rich data supports the reason for the questions, than just answering a question.. for question's sake.

To put that into perspective…

Here's what that looks like across a few businesses that have nothing in common.

A home services company. "Tell me what this repair usually costs and how long it takes."
An accounting firm. "Show me which documents I still owe you."
A fitness studio. "Build my first week of workouts."

The same fix works inside your own walls, where an empty box quietly wastes your team's time every day.

Sales. "Get me ready for tomorrow's client call."
Operations. "Show me the jobs that stalled this week."
Support. "Sum up the last three calls with this account."

None of those will match your business exactly, and that's fine, because the move underneath is the same one every time.

→ Name the three questions your people are already asking you (or you find asking yourself all the time as an internal tool) and turn each one into a button.

If you're deploying AI on your site, product or service, an empty box hands your customer the hardest job in the whole thing, which is figuring out what to want.

Start showing them what questions they should ask and everything will change.

👉 Sit With These, Whatever Business You're In

What is the one task your team still does by hand that everyone agrees should be easier by now?
Where does information live that your people cannot get to without asking somebody else first?
If a customer opened your best AI tool today, would they know what to type?

✍️ Try This, This Week

Pick one thing your team repeats every week.

Do not start with "where could we use AI," because that question is too big to answer and it's how you ended up with four subscriptions.

Ask this instead: Where are people losing time hunting for, copying, sorting or rewriting information that already exists somewhere in this business.

Write down the answer, decide who checks the work, and run one test.

That's your assignment.

It won't impress anyone at a conference. It's just the thing that actually works the same way three questions beat one empty box, every single time.

If your team has years of knowledge sitting in meetings, inboxes and chat history with no system around it, an AI opportunity audit is a good place to start 🔗.

Play your cards right, and those answers turn into your customers questions.

P.S. VIP day spots are OPEN for Sept! Check out this link to apply.

ALSO. If you have a podcast, have you seen my latest startup https://cast.house?