At the start of this journey, I wasn’t setting out to solve an industry-wide problem. I was trying to build a support bot for my team — a bot that would make my life easier so I didnt spend time answering questions which either I have repeated, or something they would have found out anyway. We used a bunch of no-code chat bot tools to solve this.

The results were… underwhelming.

The bot struggled to deliver accurate answers. Frustration grew as I realized the core issue wasn’t the bot itself but the knowledge it relied on. Very few that needed support had questions directly about the product. They instead had problems. Problems that needed troubleshooting. And AI sucks at troubleshooting. The troubleshooting scenarios weren’t documented explicitly and almost never is. The bot was only as good as the knowledge it had access to, which, in this case, wasn’t what customers needed.

This problem wasn’t unique to my team. Many companies are rushing to implement AI systems, tweaking models, and throwing more data at the problem, hoping for better answers. But this approach misses the mark. AI can’t perform magic on incomplete or static information.

Why We’re Using AI Wrong

It’s tempting to think AI is the problem. Maybe the model isn’t smart enough, maybe it needs more training. And some of this is true. Chain of reasoning prompting that ChatGPT’s Io model uses actually does a fair job. But more often than not we’re asking AI to do incredible things without giving it the foundational knowledge it needs.

Current methods of knowledge management haven’t evolved to meet the demands of AI. Documenting real-world scenarios is hard, and even when we manage to do it, the information quickly becomes stale or irrelevant. So, what do we do instead? We rely on AI to somehow “figure it out.” And then we act surprised when it doesn’t.

A Shift in Thinking

This is where Knowledge Keeper was born: out of the realization that instead of waiting for AI to “get better,” we need to rethink how we approach knowledge itself. What if, instead of static documentation, we captured knowledge as it happened — every troubleshooting scenario, every insight, every update?

That’s the idea behind Knowledge Keeper. We’re building a tool that collects knowledge continuously and dynamically updates it as new information becomes available. No more outdated manuals or forgotten processes. With the help of large language models (LLMs), we’ve turned messy, real-time workflows into something structured and accessible.

Here’s How It Works

Knowledge Keeper is simple in concept:

1. Absorb: It connects to tools like Slack, Teams, or even email, pulling in knowledge from where it’s naturally created.

2. Update: As teams interact, the system captures changes and keeps everything fresh and relevant.

3. Build: With an updated knowledge base, teams can build AI tools that actually work — whether it’s a support bot, an HR assistant, or something entirely unique.

What makes it exciting is that this isn’t just about making AI smarter. It’s about capturing knowledge that would otherwise be lost, making it easier for humans and machines to work together effectively.

Where This Matters Most

The potential applications are endless, but a we’ve been speaking to managers and here are some of the top usecases we’ve found:

• Employee Voice Bot: Knowledge Keeper records employee one-on-one meetings in a centralized knowledge base for easy access. A custom AI chatbot helps users retrieve information from these records, answer questions about past discussions, and generate summaries.

• AI Product Manager: Product managers use Knowledge Keeper to maintain all product-related documentation in a centralized knowledge base. This updated repository powers a custom AI bot, enabling the creation of test cases, user stories, and other product artifacts with ease.

• HR Policy Bot: HR teams use KnowledgeKeeper to centralize policies, forms, and FAQs, creating a self-service knowledge base for employees. A custom AI bot leverages this repository to answer questions about benefits, leave, and payroll, or guide employees to the correct forms. This approach streamlines HR operations, ensures consistency, and improves employee satisfaction.

• AI Documentation Bot: Engineers use Knowledge Keeper to centralize and maintain all API-related documentation, including endpoints, authentication methods, and error codes. This powers a custom API AI chat bot, enabling developers to access real-time API details, troubleshoot issues, and generate example requests effortlessly.

When we started this journey, we didn’t set out to solve a universal problem. But in wrestling with that issue, we realized the root cause is something almost every organization struggles with: our knowledge systems aren’t built for the way we work today.

At Knowledge Keeper, we’re not just trying to improve AI. We’re trying to help people and teams rediscover their own knowledge, making it actionable and alive in ways it’s never been before.

This isn’t about selling you a product. It’s about sharing a way of thinking. If you’re frustrated with the state of knowledge in your own work, you’re not alone — and I believe we can do something about it.