Skip to main content
Home About Services Blog Get in touch
← Back to blog
[[Strategy]]

What 30 Years of Lotus Notes Taught Me About AI

Illustration for What 30 Years of Lotus Notes Taught Me About AI

I started working with Lotus Notes in the mid-1990s. For those who don’t remember it, Notes was IBM’s enterprise collaboration platform: email, databases, workflow automation, document management, all in one. It was, genuinely, ahead of its time.

Over three decades I’ve watched Notes go from the future of enterprise computing to something organisations are still trying to migrate away from. I’ve led those migrations. I’ve archived those databases. I’ve sat with teams who couldn’t explain what half their Notes applications actually did, because the person who built them left in 2011.

That experience is more relevant to AI adoption than any whitepaper you’ll read this year.

Every platform starts with a promise

Notes promised to make organisational knowledge accessible. It would connect people, automate workflows, eliminate paper. And in the right hands, it did all of those things. Some of the workflow applications built in Notes in the early 2000s were genuinely sophisticated: approvals, routing, document lifecycle management, audit trails.

AI is making the same promises now. Connect your data, automate your processes, make your knowledge searchable. And like Notes, it can deliver on those promises. In the right hands.

The question is whether your organisation is the right hands.

The tool is never the problem

Here’s what I learned from three decades of enterprise platforms: the technology is rarely what fails. What fails is the gap between what the platform can do and how the organisation actually uses it.

Notes databases would start clean, well-structured, with clear ownership. Within two years, half of them had become dumping grounds. Nobody knew what was current. Nobody knew who owned what. The search function technically worked, but the results were useless because the underlying data was a mess.

I see the same pattern starting with AI. Organisations feed their documents into a retrieval system and expect useful answers. But if the documents are contradictory, outdated, or scattered across six different platforms, the AI will confidently synthesise garbage. It doesn’t know that the 2019 process guide was replaced. It just pattern-matches across everything it’s given.

Tribal knowledge outlasts every platform

The most valuable information in any organisation has never lived in a system. It lives in the heads of the people who’ve been there longest. The person who knows why client X has a special arrangement. The person who remembers the workaround for the billing system. The person who keeps the production environment running with undocumented fixes.

In every Notes decommission / migration I’ve worked on, the hardest part was never the technical extraction. It was unpicking what the databases meant. What was still in use. What the workflows were actually doing. Why certain fields existed. That knowledge existed in people, not in the system, and when those people had already left, the knowledge was gone.

AI doesn’t fix this problem. If anything, it can make it worse, because it creates the illusion that the knowledge has been captured when it hasn’t. You can point an AI at a thousand documents and it will generate answers. Whether those answers reflect how your business actually operates is a completely separate question.

Automation without understanding is technical debt

One of the great features of Lotus Notes was how easy it was to build workflow applications. A competent developer could have a working approval system in a day. This led to an explosion of custom applications: hundreds, sometimes thousands, per organisation.

The problem came later. When the original developers left, nobody understood the applications well enough to maintain them. When business processes changed, the applications didn’t. When it came time to migrate to a new platform, organisations discovered they had critical business processes running on applications that nobody could explain.

I see the same risk with AI automation. It’s now remarkably easy to build an automated workflow: take this input, process it, route it, generate this output. The temptation is to automate everything quickly. But if you automate a process you don’t fully understand, you’ve just made it harder to fix when it breaks. And it will break, because business processes change.

The lesson from Notes: understand the process first, automate second. Document what you build. Make sure more than one person knows how it works.

Lock-in is always about data, not about features

Every few years, a new platform promises to replace the old one. SharePoint will replace Notes. Teams will replace SharePoint. AI will replace everything. The features change. The lock-in pattern doesn’t.

Lock-in in Notes wasn’t because the features were irreplaceable. It was because the data was embedded in a proprietary database, and in a structure that doesn’t map cleanly to anything else. Extracting it is technically possible but practically expensive.

AI tools create the same kind of lock-in, just faster. If your prompts, your workflows, and your organisational knowledge are all encoded in one vendor’s system, switching costs accumulate quickly. The data itself might be portable. The way it’s been structured, connected, and automated around a specific platform is not.

My advice: keep your source of truth in formats you control. Markdown, plain text, structured documents. Let AI tools work with your data, not own it.

What actually works

After thirty years, the organisations that got the most value from Notes, and from every platform since, shared a few traits:

  1. They understood their processes before they automated them. They didn’t build an application and hope it would create order. They mapped the workflow first, then built the tool to support it.

  2. They assigned ownership. Every database, every application, every document collection had a named person responsible for keeping it current and coherent. Not a team, a person.

  3. They maintained their systems. As well as keeping them healthy. Ongoing investment in the digital assets also retains understanding in how they are put together.

  4. They cleaned up regularly. They didn’t let dead applications and outdated documents accumulate until the system was unusable. They treated information hygiene as ongoing work, not a one-off project.

  5. They planned for migration from day one. They knew that no platform is forever. They kept their data portable and their processes documented well enough that the next platform transition wouldn’t require archaeology.

These aren’t technology principles. They’re organisational principles. And they apply to AI adoption exactly as they applied to Notes adoption twenty-five years ago.

The opportunity is real, but so are the patterns

AI is more capable than any enterprise platform I’ve worked with. The potential is genuine. But the patterns of how organisations adopt, misuse, and eventually struggle with technology platforms haven’t changed in thirty years.

If you’re a New Zealand business looking at AI adoption, the best thing you can do isn’t to buy a tool. It’s to get your house in order first. Understand your processes. Clean up your data. Document your knowledge. Then bring in AI with clear, specific jobs to do.

The businesses that do this will get genuine value. The ones that skip straight to the technology will end up, a few years from now, wondering why their expensive AI system is confidently giving them wrong answers drawn from documents nobody has looked at since 2024.

I’ve seen this movie before. It runs about thirty years long. The ending is always the same.

Very Useful AI helps businesses get their knowledge and workflows in order before, during, and after AI adoption. Based in Horowhenua, working across New Zealand. Get in touch.