Who they are

They teach people to drive across the US and Canada: in-car lessons and online driver education, at scale, in a lot of places at once. And driving rules aren't national. The permit requirements, the licensing steps, the age you can start, the lesson rules, all of it changes from state to state and province to province.

Every region they operate in adds another rulebook the support team is answerable to. The more they grow, the more sets of rules any one support rep has to be right about.

The work, as it actually happened

A customer asks the support team a question, and the correct answer depends on where that customer is. What's the driving age here? Do I need a permit first? Can I take two lessons in a day? The same question has a different answer in California than it does two states over.

No one can hold that many rulebooks in their head, and no one should have to. In practice a support rep takes a question, realises it's state-specific, and goes looking for the right document among many before they can answer. On email that's dead time between tasks. On live chat it's worse: the customer is sitting there, watching the typing indicator, while the rep hunts for the rule.

Why it hadn't been fixed already

The knowledge wasn't missing. They had it all written down: a document for each issue, for each state. That is why it was hard. The right answer always existed somewhere; the problem was finding the one correct document, fast, in the middle of a live conversation. Adding more documents doesn't make that moment any quicker.

What we did

We put the answer one question away, in Slack. We connected Rivulo to their knowledge base and gave the support team a knowledge bot right where they already worked. A support rep types the question in plain language (“what's the driving age in California?”) and gets the answer back from the relevant state's document in a couple of seconds. They don't leave the conversation and they don't go hunting.

Every answer carries its source. The bot shows the document each answer came from. If a rep wants to read the full detail or double-check before they reply to a customer, the source is right there to open. On regulatory questions, being able to see where an answer came from matters more than the answer arriving quickly.

We embedded it, we didn't just hand it over. We sat in a shared Slack channel with their team and treated the bot as part of their operation rather than a thing we'd shipped and left. Getting it right at the start meant making sure the bot was drawing on their knowledge exactly as written; in the first days we found some of their documentation hadn't been brought in cleanly, corrected how it was ingested, and it ran well from there. That is what the embed stage is for.

Then we managed it. Support knowledge is never finished, and theirs changed as rules and offerings changed. When they flagged something or updated their documentation, it was handled quickly, usually the same day. What they kept coming back to wasn't a feature. It was that we stayed close to the work and moved fast when it mattered.

How it landed

We're not putting a number on this one. What their team valued was how we worked with them, the closeness and the speed, and they said as much.