All posts

A rental shop run by three people: what we automated at Rundle Bike

How a bike rental shop in Canmore handles peak season with a small team, which jobs software took over, and which ones stayed human on purpose.

Rundle Bike is a bike rental shop in Canmore. In peak summer, the shop handles a steady stream of bookings, walk-ins, hotel referrals, returns and questions about trails. It runs with at most three people working at any one time.

That's possible because almost everything that isn't face to face is handled by software I built for the shop. Here's what that covers, what it doesn't, and what I'd tell another small operator thinking about the same thing.

The principle: one record, many doors

The most important decision wasn't about AI at all. It was making every part of the business read and write the same reservations.

The online booking page, the chat assistant, the staff counter app, the admin dashboard, the hotel referral portal and the scheduled jobs all work from one database. A booking made online at midnight shows up on the counter screen in the morning. A bike marked as damaged at check-in disappears from online availability immediately. Nothing gets typed twice, so nothing gets typed wrong twice.

If you take one idea from this post, take that one. AI on top of three disconnected systems just makes the disconnection faster.

What the software does

Bookings and payment

Customers book and pay online, with checkout through Stripe. Waivers are signed digitally per reservation, so nobody fills in paper forms at the counter while a line builds behind them.

A chat assistant that answers from live data

The website's booking assistant answers questions like "do you have two e-bikes free Saturday morning?" by checking real availability, not by guessing. It has eight tools it can use, including looking up availability and prices. When a question needs a person, such as a damage claim or an unusual group request, it hands off to staff instead of improvising.

Check-in at the counter

Staff use a counter app to check renters in. A vision model reads the renter's name off their ID, which saves typing and avoids misspelled names. Because it reads images supplied by customers, it has a guard against prompt injection: text on a document can't instruct the system to do anything.

Pricing and planning

Prices adjust for weather, since a rainy forecast changes demand. A 14-day demand forecast helps decide staffing and fleet prep.

The jobs nobody has to remember

Scheduled jobs send reminders, issue lock codes and reset the fleet status every night, so each morning starts from a clean, accurate picture of what's available.

Partners

Hotels have their own portal to refer guests, so a front desk in town can send customers without phone tag.

What stayed human, on purpose

The software is deliberately kept away from a few things.

Fitting and safety. Adjusting a bike to the rider, explaining the e-bike controls and making sure someone is comfortable before they head onto a trail is a person's job.

Advice. "Which trail should we do with a nine-year-old?" deserves a conversation with someone who's ridden it.

Judgment calls. Damage, refunds, an unhappy customer, a group that shows up late: staff decide. The software gives them the information to decide quickly.

Repairs. Software can log a problem. It can't true a wheel.

Those are the moments customers remember, and the reason a small team works is that the team spends its time on them instead of on paperwork.

What I'd tell another small operator

Fix the data first. Get bookings, inventory and customers into one system before adding AI anywhere.

Automate the boring, frequent jobs. Reminders, confirmations, availability questions, waivers and nightly resets are high volume and low judgment. That's where the hours are.

Give AI tools, not freedom. The assistant can look things up and suggest a booking. It can't change prices or approve refunds.

Keep the handoff easy. Every automated part has a clear point where a person takes over, with the context already filled in.

Watch it in peak season. Problems appear under load. Review what the assistant said and what the jobs did during the busiest weeks, and fix it while it's fresh.

The result

A small team that spends its day with customers, not with spreadsheets. The software isn't the product. Good rentals and good advice are. The software just clears the way for them.

Have work that needs doing?

Send me the role, the problem, the process or the team you want trained. I'll come back with what I'd deliver, how I'd run it and a fixed fee.