← All notes
Knowledge bot

The answer is on Basecamp.

What I built after The Gaffer: a Slack bot that finds everyday information in Basecamp and brings back an answer, the context and a link to its source.

A familiar question, another search

When is the next staff meeting? When is Jake’s birthday? Who should I contact about PR enquiries? They are simple questions, and the information is already in Basecamp. Finding the right item is the extra bit of work.

You might remember seeing an update without remembering where it was posted. You open a project, search for a phrase, try another one and eventually find the detail you needed. Or you ask a colleague who knows where to look.

After building The Gaffer, I wanted to apply what I had learned to that everyday problem. I built a knowledge-base bot that people can message in Slack. They ask in ordinary language, and it searches the Basecamp project for the relevant information.

Bring the answer to the conversation

The interaction is deliberately simple. Ask the question in Slack, where the conversation is already happening. The bot replies with an answer, the surrounding context and a link to the Basecamp item it came from.

That last part matters. Someone asking about a staff meeting may only need a quick answer. Someone asking about a policy may want to read the original wording. The reply gives them a starting point and a route back to the source.

I think that is where this becomes useful: it saves the person asking from having to work out where to begin their search. Basecamp still holds the information. Slack becomes another way to reach it.

THE KNOWLEDGE BOT / THE REPLY

An answer.
And a way to check it.

ASK IN SLACKWhen is the next staff meeting?
01 / THE ANSWER

The detail.

The information you asked for,
brought into the conversation.

02 / THE CONTEXT

The bigger picture.

The surrounding information
that helps you understand it.

03 / THE SOURCE

The original item.

A Basecamp link to follow
when you need to read more.

The reply saves a search. The link keeps the original information within reach.
A conceptual view of the reply: an answer and its context arrive in Slack, alongside a link to the original Basecamp item.

The lesson that carried over

The Gaffer had already shown me how convincing a bot can sound when it gets something wrong. In that project, a claimed player transfer became a fact in the reply even though the squad data did not support it.

The lesson I took into this app was to keep the answer connected to the information behind it. Including the Basecamp link gives people a way to inspect the original item rather than relying entirely on the summary.

When a search returns nothing, the bot points the person towards someone to ask. That is a useful outcome too. It gives them a next step without presenting an answer it could not find.

A source link does not make every summary correct. It does make checking easier. For policy questions in particular, I see the reply as a route into the original information, with the context there for anyone who needs it.

THE KNOWLEDGE BOT / WHEN NOTHING TURNS UP

A useful answer.
Or a useful next step.

SEARCH THE BASECAMP PROJECT

Is the information there?

INFORMATION FOUND

Bring it back.

Answer, context and source link.
Continue in Slack.

NOTHING RETURNED

Point to a person.

Suggest someone to ask.
Give the question a next step.

When the search finds nothing, the bot points the person towards someone to ask. The source stays visible whenever an answer is returned.

A small job worth doing well

In day-to-day use, it usually comes back with what people are looking for. The practical benefit is simple: people can ask a question instead of digging through the project themselves.

I like that the scope is easy to explain. The app finds information that already exists and makes it easier to get to. Staff dates, policy questions and the right person to contact are ordinary things, but they come up often enough to be worth making easier.

For me, this was a useful next step after The Gaffer. The same attention to evidence and how a bot responds became part of a quieter tool, built around a small task people already needed to do.

If your team has useful information that is difficult to find, get in touch with RKH, where I work. The team can talk through the problem and whether a tool like this could help.

Talk to RKH about your idea ↗
← Back to all notes