Case study: Telegram bot for a logistics company that replaced a $40k SaaS

Most case studies are marketing. This one is mostly the parts that didn't go well, because that's where the lessons are.

The setup

A 30-driver regional logistics company in eastern Europe. They were paying ~$40k/year for an industry SaaS used in dispatch / proof-of-delivery / fleet tracking. The dispatcher loved it. The drivers refused to use it. Drivers were still sending WhatsApp messages and photos to the dispatcher, who manually entered them into the SaaS.

The owner called me asking for "an app that the drivers will actually use." After two hours of calls with the dispatcher and three drivers, the answer was obvious: don't build an app, build a Telegram bot. Every driver was already in Telegram all day.

Scope

After a one-week discovery:

• Telegram bot for drivers: log pickups, scan delivery proof, request route changes, message dispatcher.
• Small web dashboard for dispatcher: see live driver status, assign jobs, export reports.
• Integration with their existing ERP (one-way push: jobs from ERP → bot, status updates from bot → ERP).

Hard cut: anything the existing SaaS was good at but drivers refused to use. Custom routing optimization, custom maps, custom analytics. The SaaS had all of these and they were unused — we explicitly didn't rebuild them.

Build (4 weeks)

Week 1: Bot spine. Login flow, list of today's jobs, mark pickup, mark delivery with photo. Tested with 3 drivers on day 5.

Week 2: Dispatcher dashboard. List of jobs, drag-and-drop assignment, live driver status. Boring but functional.

Week 3: ERP integration. The hardest week. The ERP's API was poorly documented and we had to reverse-engineer two endpoints. Lost 2 days to a webhook that fired duplicates.

Week 4: Rollout. Trained the dispatcher (45 minutes), trained the drivers (a 90-second voice message in their group chat). Done.

Adoption

Within a week, 28 of 30 drivers were using the bot for every job. Two holdouts were near retirement and we let them keep using WhatsApp; the dispatcher transcribed manually.

Within a month, the dispatcher requested the SaaS contract not be renewed. They were processing more jobs with less data-entry time.

The numbers

• Build cost: ~$18k for 4 weeks of work.
• Hosting: $50/mo for a VPS + database.
• SaaS savings: $40k/year.
• ROI: positive in month 5.
• Driver adoption: 93% in week 1 (vs ~10% for the SaaS over 2 years).

What didn't go well

• ERP integration took twice the estimate. Always estimate integrations at 2x your gut.
• The dispatcher initially missed the SaaS analytics. We added two basic reports in a follow-up week, paid separately.
• One driver lost his Telegram session on a phone replacement. We added an "I'm a returning driver, here's my phone number" recovery flow.

What I'd do differently

• Demo the bot to one driver on day 3 of week 1, not day 5. The earlier feedback would have caught two UX issues we shipped.
• Build the ERP integration last, not in week 3. If it had failed we'd have shipped without it and added it later.
• Charge a 4-week + 1-month-of-support contract, not just 4 weeks. Inevitable polish always happens.

The bigger lesson

The right product surface for your users is the one they already have open. For office workers it's a web app. For drivers, field workers, and most blue-collar service work, it's a messaging app. Building a "better app" they refuse to install is a slow expensive way to lose to a Telegram bot you can ship in a month.