Elevator pitch

I went through all eight project folders on your Desktop, and here's the pitch built from them.

The pitch (about 30 seconds out loud)

Every small business runs on promises: an appointment, a pickup, a callback, a follow-up. I build the systems that make sure none of them slip.

Right now I run the online booking and client records for an independent hair stylist. I moved her client history, nearly 1,900 people, over from her old salon software and checked that every one arrived. Every morning the system checks her calendar against her booking system and fixes anything that didn't make it across. I also built an AI assistant for an account manager at a recycling company. It reads their email and meetings, turns them into a to-do list, and gives them a briefing for the week ahead.

I build fast, with AI doing the heavy lifting, and I have other AI systems double-check the work before it ever reaches the client. Where I'm headed: turning this into something any appointment-based business can switch on, starting with independent stylists.

The 10-second version:

I build AI-powered back offices for small businesses: booking, client records, and assistants that turn your inbox into a to-do list. Nothing your business promised gets dropped.

What you do now, and where you're heading

Now: live and in real use

  • A solo stylist's booking and client records (b, c, s/client). The booking site has been live since July 25. Every morning at 6am it checks her calendar against the booking system. It holds about 2,000 client records, about 1,880 of them carried over from her old salon software.
  • An AI assistant for a recycling company's account manager (gwcli), running for months. It turns their email into a to-do list and writes a weekly briefing, which AI from four different companies double-checks first. You chose which AI models to use by testing them on that client's own 54 real questions.
  • Fast one-off tools. In about two days (mgmwork) you built a secure page so a colleague could cover that manager's week away.
  • Turning messy data into something trustworthy. You moved 18 months of salon history over, and analysed 9,573 tip records to choose her tip buttons.

Next: built pieces, not launched

  • An AI receptionist that books appointments by voice or text (s/a). It's built but has never been launched.
  • A booking platform for independent stylists (s/5/shape), planned at $29–49 a month. The engine is built, but it has no real screens yet.
  • A list of 107,289 Michigan license records, and a ranked list of salon suites to approach.
  • A test of 19 AI models from 6 companies, ranked on cost, speed and accuracy (lab/arena).

Heading toward: designed, not built

  • The "commitment layer" (lab/manifesto/commitment-layer.md). It's an AI that keeps track of every promise a service business makes: appointments, deposits, memberships, follow-ups. It then proposes the next step for the owner to approve. Stylists come first, then other service trades. The pitch's opening line about promises comes from this document.
  • Voice-first assistants that aren't tied to any one AI company (ix/los, ix/f, ix/chassis).

Gaps to be honest about

  • Nothing in your files shows payment. If these jobs were paid, say "clients". If not, say "businesses I work with".
  • Each of the two live businesses got its own one-off build. Neither is yet a package you could hand to the next customer.
  • The product side is mostly design so far. Your own research found that Square already gives stylists a free AI agent, so the stylist product needs a narrower angle.
  • A booking outage on Aug 28 lost 3–4 bookings. That's why the pitch says "checks and fixes" rather than "never misses".
  • Your files don't say whether you're a stylist yourself, so I left it out.

Two decisions only you can make

  1. Were those two jobs paid? That decides the wording above.
  2. Is there a second buyer? You also have real depth in choosing AI models and checking their work, which you could sell to developers and tech teams instead of business owners. Nothing in your files says you want to, so it isn't in the pitch. Say if you'd like a version for that audience.

An independent reviewer checked 15 claims against the files and passed it with fixes. It caught a wrong count of AI companies and some overstated wording, and all of that is fixed above. No client, customer or business names appear in the pitch.