Yesterday I was in a room with a group of nonprofit executive directors.

We were talking about AI.

There was a mix of curiosity + pressure + quiet overwhelm.

Because they’re all being told the same thing:

But no one tells them where to start.

And for nonprofits, it’s even heavier.

They’re already expected to do more… with less… with smaller teams… with tighter budgets…

So adding “figure out AI” to the list just feels like one more thing they’re behind on.

So instead of talking about tools…

I told them about Herbie.

From The Goal (published in 1984 and still taught in Operations classes in business school)

If you haven’t read it, here’s the quick version:

A group of boys are hiking. Herbie is the slowest kid in the group.

Everyone else walks faster, but they keep having to stop and wait. So even though most of the group is fast…

The entire group’s speed is determined by Herbie.

Not the average.

Not the best.

The slowest.

It's relevant. Because we all have functions, tasks or priorities which stand in our way.

Here’s the mistake:

Leaders are trying to apply AI everywhere…

Instead of asking:

In nonprofits, Herbie can show up as:

• Grant writing that takes weeks
• Reporting that eats entire days
• Manual data entry across disconnected systems
• Content creation bottlenecks for fundraising

In companies, it might be:

• Sales follow-up
• Customer support backlog
• Internal approvals
• Decision-making itself

If you don’t fix Herbie… Nothing else matters.

You can make everything else faster… But the system won’t move.

So the better way to approach AI is simple:

Don’t start with tools.

Start with this question:

That’s your Herbie.

Then (and only then) apply AI there.

Not everywhere.

There.

When you do this right:

• You don’t just get a 10% improvement.
• You unlock the entire system.
• The team moves faster.
• The work compounds.
• Momentum builds.

Most organizations don’t need more AI.

They need more focus.

If you’re thinking about where to start, try this:

Find your Herbie:

1. Where does work consistently pile up?
2. What do people complain about the most?
3. What delays everything else downstream?

Then test AI there:

• Can it draft the first version?
• Can it automate part of the workflow?
• Can it remove repetitive steps?

Ignore everything else (for now).

Because the goal isn’t to “use AI.”

The goal is to move faster as a system.

And that starts with Herbie.

Alex