If you run an agency long enough, you eventually realize you’re answering the same handful of questions over and over again. Not just with clients, but with your team, too.

And this week, I got fed up.

Not with them. With me.

So I built something to to fix it. The results?

• Made my team way more autonomous
• Cut our internal back-and-forth
• Made onboarding new hires way easier
• Standardized how we answer client questions
• And saved us a stupid amount of time

You should definitely do the same...

Solution v1: The Primitive FAQ Document

At first, I handled it like most people do.

An FAQ document.

Whenever a team member or client asked something that required a long explanation, I’d add the answer to a Google Doc. Over time, that doc became this giant, messy FAQ sheet I’d send to new hires or reference when I forgot how we solved a specific problem.

And honestly, this alone was extremely helpful and if you don't have something like this already you're missing out on a valuable resource.

But the more Paperboy has grown, the more I’ve wanted something better than a glorified text file.

So I graduated our document to something slightly more sophisticated.

Solution v2: The Custom Trained AI Support Agent

Client or team support is one of the most obvious use cases for AI. Im shocked it took me as long as it did to build this despite it being so damn easy.

Here's how.

Step 1: Gather the raw material

I started by looking back through the last month or two of communication, specifically:

• Emails where I sent long responses explaining a solution
• Slack threads where the team brought me tricky situations or edge cases
• Client questions that required me to dig in and give more context
• Internal handoff conversations between team members
• Notes from onboarding calls where the same concerns kept popping up

Any time I saw something that required thought — not a generic “yes/no” answer — I copied the whole thread into a document.

No formatting. No editing. Just a giant dump of real scenarios and real answers.

(If you wanted to go deeper, you could also pull data from: Loom videos, Notion docs, Asana comments, old proposals, client kickoff notes etc)

Don't be too precious about what you're capturing, we'll fix that in the next step.

Step 2: Turn it into an actual FAQ

Next, I fed this massive doc into ChatGPT/Claude and told it:

“Organize this into a question-and-answer style internal FAQ. Remove duplicates. Group related questions. Clean up the language but keep the intent.”

Within a few minutes, I had a clean, organized, highly detailed knowledge base — the same one I had been trying to manually build for months.

From there, I exported it as a single doc and uploaded it into a project so the AI could access it as the knowledge base behind the bot.

And honestly, even if you stop here, this alone becomes an incredibly valuable asset.

Sometimes you get a client question and think, “I’ve answered this before… where did I say that?” Now you can ask the AI and get the answer instantly — formatted, clarified, and ready to send.

Pretty awesome.

Step 3: Make it accessible where the team actually works

This is where things got fun.

I didn’t want the team to jump out of their flow, open ChatGPT, load the project, and ask it something.

Too many steps.

So I built a Slack bot.

Using the Slack API and a bit of automation, I set up a simple system. Here's how it roughly works:

1. Tag it in any Slack channel
2. Ask your question
3. It pulls from the Paperboy FAQ and responds directly within Slack

No extra tabs. No friction. No “hey Shane, quick question”—which, as you know, is never actually a quick question.

I also connected it to Slack’s channel history so the bot can do things like:

• recap the last 7 days of activity
• summarize client interactions
• find old answers buried in 4,000 messages
• highlight potential issues that were mentioned but not resolved

Now the bot isn’t just a knowledge base — it’s an internal project assistant.

--

Here's why I’m so excited about this.

Beyond the cool factor, this solves real problems:

1. Makes my team more autonomous: The fewer interruptions they need from me, the more scalable the agency becomes.
2. It shortens the learning curve for new hires: Instead of me spending weeks training them, they have a searchable mentor available 24/7.
3. It preserves institutional knowledge: Nothing disappears when a team member leaves. Nothing relies on someone’s memory.
4. It makes us more consistent: Clients get the same high-quality answer every time — no matter who responds.
5. It saves real time: Some answers take 15–20 minutes to rewrite from scratch. Now they take five seconds.

Anyway — bit of a nerdy build, but I’m genuinely excited about it. If you’re running a service business, I highly recommend creating your own knowledge base.

Even if you don’t take it as far as building a Slack bot, the first two steps alone will save you hours and keep your team aligned.

Happy building!

— Shane