Better first meetings.
Suggests the next question while you meet a new client, then hands you a structured intake summary. You stay in the conversation; the notes write themselves.
AI assistants for expert-led firms
More clients per expert, without a new hire.
Your best people wear too many hats. They bring in the clients, then write the notes, send the follow-ups and chase the documents. I find the one task that eats the most hours and build an AI assistant for it, inside your email and CRM.
Built and run by Roman Bova, AI systems engineer in Winnipeg.
Your morning, prepared by AI.
- Follow-up: intake with D. MoreauAI draft · readySummary of Monday's call, next steps, two documents to request
- Reminder: bank statementscase 1042Second reminder, same polite tone as the first
- New client at 2:00 pmprepIntake Copilot has the first questions ready
- Check-in: Patel fileAI draft · readyThree weeks since last contact, short note drafted
AIDrafted by your assistant from Monday's notes, in your voice. Nothing goes out until you say ok.
Hi Daniel,
Thanks for your time on Monday. As we discussed, the next step is the pre-approval. Could you send your two most recent pay stubs and the signed consent form by Friday? I'll take it from there and come back to you with options next week.
Best,
Anna
Ok, sendEdit
Who it's for
Your experts' time is your revenue. Find where your firm sits.
It fits
Firms of 5–30 whose experts close deals or bill hours, and still work every client file themselves.
Mortgage brokers
Applicants wait while you look up what they still owe.
Immigration consultants
What each client still owes lives in someone's head.
Law practices
A first meeting ends with a page of scribbles to type up.
Insurance brokers
Renewal reminders go out the day they're due, or the day after.
Not the right fit
An operations team already owns this work.
Not the right fit
You want a chatbot for your website.
How I work
AI prepares, your people decide.
Your data stays in your accounts.
I don't take any of your data out. Servers are yours, and the AI models run on your firm's own accounts. You choose the providers, and we go through their terms together. Business APIs such as Anthropic's are contractually barred from training on your data, and zero data retention can be arranged where you need it. Speech is transcribed by AssemblyAI or OpenAI, or on your own computer. Recordings aren't stored unless you want them.
Prefer everything on your own hardware? That's an option too.
Nothing goes out without your ok.
The assistant prepares drafts, summaries and reminders. A person on your team approves anything that reaches a client.
You own it.
Code, servers and AI accounts are set up in your firm's name. No subscription to me, no lock-in.
No new system to learn.
The assistant works in the email, spreadsheets and CRM your team already uses.
Assistants
Most firms start with one of these. Each one is built around a firm's own workflow.
Follow-ups that actually get sent.
Every morning: what needs you today, with follow-up emails already drafted in your own voice.
Documents without the chase.
Knows which documents each client still owes, prepares polite reminders and keeps the case status current, so nobody has to chase by hand.
Something else eats your team's time? That's what the first call is for.
What you get
One workflow. One number. Evolution, not revolution.
- 01
A first call.
You tell me where the hours go. I show what an assistant could take over.
- 02
Diagnosis.
We pick one workflow, agree on the number that should move (for example, follow-ups sent the same day) and measure where you are today.
- 03
Build.
I build the assistant in your accounts, on your email, your spreadsheets, your CRM.
- 04
Before and after.
We measure the same number the same way, and you see both side by side.
Results
Results from real work.
Each number is measured the same way before and after, with the caveat next to it.
Client follow-ups sent within a week
A career advisor's practice, with a CRM whose assistant drafts every follow-up. Before: 55% went out within a week and 29% never went out. After: the first two weeks, 10 follow-ups. The full-month figure comes in November.
First client meeting, intake to written summary
With Intake Copilot suggesting the next question and live transcription instead of typed notes.
Time to finish a task
The same method in a software team: three engineers, before and after adopting my AI process.
I use Intake Copilot in my own client meetings.
About
Who's behind AI Gear.
AI Gear is run by Roman Bova, an AI systems engineer in Winnipeg. I design and build AI systems that take routine work off people, and I measure whether they actually do.
Roman on LinkedIn
Where do your team's hours go?
Tell me a little about your firm.
- 01I read it myself and reply with a few questions.
- 02A short call to see what an assistant could take over.
- 03We pick one workflow and its number, or agree it isn't a fit.
or email hello@aigear.ca
Start with one workflow.