AI for founders who need to move fast.

What you'll be able to do

• Research a market, competitor, or customer segment in under an hour
• Turn a rough idea into a first-draft landing page, email, or deck
• Write prompts that give the model the context it actually needs
• Spot when AI output is confident but wrong, before it costs you
• Build a repeatable AI workflow for the tasks you do every week

Inside the path

A focused set of five-minute lessons. Each one ends with a hands-on exercise, not a quiz you can guess.

1. AI as your first hire 5 min

   Where AI genuinely saves a founder time, plus the jobs you should never fully hand off.

2. Research a market in an hour 6 min

   Prompt patterns for sizing a market, mapping competitors, and finding customer pain.

3. First drafts that don't sound like a robot 5 min

   Get usable copy for your site, emails, and deck by giving the model your voice and constraints.

4. Pressure-test your own thinking 5 min

   Use AI as a skeptical advisor: red-team your pitch, pricing, and assumptions.

5. Build your founder AI workflow 5 min

   Turn your weekly repetitive tasks into a set of reusable prompts and checks.

Try a sample exercise

This is the kind of card you'd practice inside Iro: you do the thinking, then get feedback.

Why founders get uneven results from AI

Most founders try AI, get a few great answers and a few useless ones, and can't tell why. The difference is almost never the model; it's the prompt and the verification. A good founder prompt gives the model the same context you'd give a sharp new hire: who it's for, what "good" looks like, what constraints matter, and what to do when it's unsure.

The other half is judgment. AI will confidently invent a market size, a competitor feature, or a legal detail. The founders who win with AI treat every output as a draft from a fast but junior teammate: useful, but checked before it ships to a customer, an investor, or the codebase.

The highest-leverage uses for a lean team

• Research: market maps, competitor teardowns, customer-interview synthesis, positioning options.
• First drafts: landing pages, cold emails, investor updates, job descriptions, help docs.
• Operational glue: summarizing calls, turning notes into SOPs, drafting support replies, cleaning data.
• Thinking partner: red-teaming your pitch, stress-testing pricing, listing risks you haven't considered.

Notice what's not on the list: letting AI make the decision. It drafts and researches; you decide.

How do you turn customer interviews into a positioning decision?

Run it in three passes instead of one: extract, then cluster, then decide. Founders who paste ten transcripts into a chat and ask "what should our positioning be?" get a bland summary, because the model is doing three different jobs at once and none of them well.

Pass one: extract

Feed one transcript at a time and ask for the same fixed schema every time: the problem in the customer's own words, what they tried before you, what they'd have to give up to switch, and any direct quote about budget or urgency. Verbatim quotes only. Tell it to write "not mentioned" instead of filling a gap, and it mostly will.

Pass two: cluster

Paste the extracts together and ask it to group them by underlying problem rather than by feature request, with a count of how many interviews support each group. Then ask which clusters came from a single person. Those are the ones you'll over-weight later, usually because that person was articulate and you liked them.

Pass three: decide

Now bring in a role. "Act as a skeptical seed investor. Here are the clusters and their counts. Which one would you bet a company on, and what would have to be true for that bet to work?"

Argue back before you accept the verdict. Ask for the strongest case against whichever cluster you already favor, and notice how much of your answer is evidence versus preference.

The call is still yours. What you got back is the afternoon you'd have spent re-reading transcripts, plus a check on the quiet bias where the interview that agreed with you is the one you remember.

The AI habits that cost founders the most time

Bad prompts aren't the expensive part. It's the output that looked finished, got shipped, and came back.

• The one-shot mega-prompt. Asking for an entire pitch deck or business plan in one request gets you something structurally correct and specifically empty. Chain it: positioning first, then the narrative arc, then slide copy, feeding each result into the next step.
• Regenerating instead of correcting. Hitting retry five times is slower than one message that says what's wrong: "too long, cut the second paragraph, and the tone is pitchy where it should be matter-of-fact."
• Numbers you never checked. Ask about your category and a model will produce a competitor's headcount, funding total and churn rate in one confident paragraph, none of it sourced. In a deck, that's one investor question away from a bad meeting. Anything numeric that leaves your building gets verified against a real source.
• Automating a process you've run twice. If you can't write the steps down, you can't automate them. Do it by hand until it's boring, then build.
• Outsourcing the things only you can say. The investor update after a bad month, an apology to a customer you let down, a rejection to a candidate you liked. Speed was never the constraint on those.

The failure mode shifts as you grow. Solo, you're the only reviewer, so the risk is fatigue: at 11pm everything reads fine. Once there are three or four of you, unreviewed drafts start landing in shared docs and get treated as settled, because nobody remembers who actually wrote them. Label AI first drafts as first drafts and that problem mostly goes away.