For: Founders who feel behind and aren't sure what they're behind on

What should I actually be doing with AI this quarter?

Where the line actually sits

2 workflows

Per quarter, no more

One on the money-making side, one on the admin side. Run them weekly until they're unremarkable before adding a third. Ten pilots is the same as zero, but with more subscriptions.

30 days

Before a tool is cancelled

Go through the billing page quarterly and cancel anything that hasn't changed an hour of someone's week. The pile of unused seats is the tell that you're buying instead of changing.

Drafts yes, decisions no

The line that doesn't move

First versions, summaries, research, cleanup — hand those over. Judgment, pricing, hiring calls, and hard conversations stay yours. Nobody remembers who wrote the draft; everyone remembers who made the call.

Every quarter

How often this answer expires

What was impossible last quarter is table stakes this one. This is the one page here with a genuine shelf life.

What changed in the last year

• The bottleneck stopped being capability and became standards. Most bad output is a missing definition of good, not a missing model — write the standard once, reuse it, and the quality problem mostly disappears.
• Cost per unit of work keeps falling, which means your competitors' prices can fall whether or not yours do. Even if AI doesn't touch your product, it is already touching your market's price floor.
• Being visibly human became a differentiator rather than a default. Speed is cheap now; the company that clearly still has people in it is the rare one.

What most founders get wrong

• “We need an AI strategy.”

  You need two workflows that got faster and stayed faster. Strategy documents about AI are usually a way of not doing either.

• “We'll automate support first.”

  It's the most tempting and the most visible to customers. Start where you're capacity-constrained internally, where a bad output costs you an hour rather than a relationship.

• “We're behind.”

  Almost everyone is at the same place: a few useful habits and a lot of unused subscriptions. Behind is a feeling produced by other people's announcements.

When I'd tell you the opposite

• Your customer's job genuinely gets done a new way now. Then this isn't an operations question at all — stop optimising workflows and rebuild the product, urgently.
• You sell trust, discretion, or accountability as the product. Moving faster is worth less to you than being the one who didn't, and the restraint is the offering.
• You're pre-revenue. Ship the thing. Internal efficiency has nothing to be efficient about yet.

Related questions

Where should a small company start with AI?

The work you'd hire for if you had the money. That list is the shortlist, because it's where you're already capacity-constrained.

Will AI replace my team?

It replaces tasks, and tasks make up jobs unevenly. Map your functions as gone, ten times cheaper, or unchanged, using your own payroll rather than headlines.

Should we build our own AI features or buy tools?

Buy for anything that isn't what customers pay you for. Build only where the output is the product, because that's the only place the maintenance is worth it.

How do I stop AI output from sounding generic?

Write down what good looks like — real examples, in your voice, with the rules stated — and reuse it. Most generic output is an unwritten standard, not a weak model.