Building an Internal AI Sales Assistant

This did not start as an AI project. It started as a problem that AI tools for sales teams turned out to be exactly right for solving.

It started as a sales and marketing strategy project. I was asked to put the strategy together, which raised a fair question: shouldn’t the strategy already exist before someone asks for the polished version of it?

At first, the discussion was about format. Should this be a webpage, a brochure, a presentation, a book? The annoying answer was yes. It needed to be all of those things, because the real need was bigger than a single deliverable.

At the same time, I was exploring how AI could help me build faster and think bigger. That is when the project changed. Instead of creating one finished artifact, I built a tool that could give people real answers in real time, based on actual company knowledge, and help them turn that knowledge into something useful.

That shift turned a documentation project into something much more practical: an internal AI assistant built to support the work as it was happening.

The Problem

The original problem was bigger than documentation.

There was a large sales and marketing organization, a lot of knowledge, and too much inconsistency in how messaging, strategy, and objections were being understood and communicated. The information existed, but it was scattered across years of presentations, case studies, internal notes, survey data, leadership input, and lived experience. Some of it was current. Some of it was outdated. Some of it was useful but buried. Some of it only existed because the right person had been around long enough to remember it.

That creates a predictable mess. People guess. Teams interpret things differently. Messaging drifts. New hires take longer to ramp up. Sales teams spend too much time hunting through folders instead of getting to the point.

What was needed was not another file. It was a better way to create alignment and make the strategy usable in the middle of real work.

The Groundwork

This project worked because I did not start with the chatbot.

I started with the mess.

I had years of context working in the business, which meant I knew where the information lived, how reliable it was, what had aged badly, and what was still worth building from. I could track down the source material, evaluate it, update what was old, identify what was missing, and ask a better question than “how do we package this?” The better question was, “what would make this useful?”

That meant building more than content. I had to construct the narrative, shape the outline, organize the knowledge, define the logic behind the tool, and create the structure behind it so the system had substance instead of novelty.

The source material ran deep: strategy, messaging, objections, case studies, survey input, books, presentations, leadership perspective, failures, and the accumulated knowledge that builds up over more than a decade of doing the work. The point was not just to collect it. The point was to turn it into something coherent, teachable, and usable.

Building AI Tools for Sales Teams

I built the entire thing.

What started as strategy documentation became a knowledge system, then became an internal AI assistant trained on the information people actually needed: strategy, objections, messaging, and the broader context behind how the business communicates and sells.

I gathered the information, cleaned it up, updated old material, filled in missing pieces, structured the content, built the narrative, documented the process, and turned it into a framework people could actually use. Then I built the chatbot itself, distributed it, and trained the team on what it could do and how to use it well.

That mattered because this was never supposed to be an AI experiment. It was built to solve a practical problem. People needed a faster way to get real answers without guessing or digging through disconnected folders. Leadership needed better alignment. The team needed something more durable than institutional memory.

This project also pushed me outside my usual lane. My background is marketing, but to build this well I had to think like sales, operations, process design, and training. I was not just making content. I was building a tool that had to teach.

The biggest shift was this: the final form of the project was never going to be one thing. So I built something that could help create many things, and not just through me.

Tools Used

I used a mix of design, documentation, and AI tools to build this project. The tools mattered, but only because they helped me turn scattered information into something structured, usable, and teachable.

Tool	Purpose
Adobe Illustrator	For diagrams, visual structure, and shaping information into something clearer.
Google Drive	For gathering source material, organizing documentation, and keeping the working system accessible.
ChatGPT	For drafting, synthesis, prompt development, and pressure-testing how the assistant should respond.
Google NotebookLM	For working through source material, finding patterns, and grounding outputs in real documentation.
Claude	For long-form synthesis, restructuring content, and comparing different ways to shape the knowledge system.

The Result

The result was a strategy system people could actually use.

Instead of leaving the work trapped inside a static format, I built something that made the thinking behind the strategy accessible in real time. It improved consistency in messaging, created stronger alignment across the organization, and gave people a faster path to useful answers.

It also reduced the dependency on one person having to hold all the context. That matters more than people think. Once useful knowledge becomes easier to access, teams move faster, onboard better, and make fewer avoidable mistakes.

Customers feel that difference even if they never see the tool itself. A sales conversation grounded in accurate, current strategy means fewer wrong answers, fewer “let me get back to you,” and a more consistent experience no matter who they end up talking to.

Since launch, the tool has been adopted by leadership and used as a practical internal resource. That tells me it did the job it was built to do. Not because AI is trendy. Because the tool was useful.

What This Project Proved

This project proved that sometimes the original brief is too small for the real problem.

I was asked to build a sales and marketing strategy. That would have been fine. A document, a presentation, maybe a polished internal resource. Useful enough.

But the real problem was not the lack of a deliverable. The real problem was the lack of a system people could use. As the project developed, and as the technology became more practical, it became obvious that the better answer was not just to package the information. The better answer was to build structure, create clarity, and turn scattered knowledge into a tool that could actually support the business.

That is the kind of work I like most. Start with something messy. Find the real point. Build the thing that survives.