In 2016, I stood in front of a room of executives and told them that chatbots were about to change everything. I had a 16-slide presentation. I had examples. I had conviction.

I was right. I was also wrong. And the gap between those two things is where the interesting lessons live.

The pitch

At the time, I was running Vertical Leap, a digital marketing agency I'd later sell. We'd been experimenting with conversational interfaces, and I'd put together a presentation called "Conversational UI and Chatbots" that I delivered internally and to a series of clients.

The thesis was straightforward: conversational interfaces were about to become the third major platform, after websites and mobile apps. I laid out three categories — voice assistants like Siri, conversational web interfaces, and chatbots on messaging platforms. I showed KLM's Facebook Messenger bot as the future of customer service. I mocked up what Google Analytics might look like if you could just ask it questions instead of navigating dashboards.

The benefits slide read like a manifesto:

  • Nothing to download
  • Cheaper than mobile apps
  • Cross-platform
  • One interface, unlimited capability
  • No menus or UI to navigate — just ask
  • Eliminates friction between people and computers

And then the bold prediction: "Will be ubiquitous in less than two years."

2016 slide, Conversational UI: what are the benefits? Nothing to download, cross platform, no menus to navigate, just ask

What I got right

The core thesis has aged remarkably well. "One interface, unlimited capability" could be Claude's tagline today. "No menus to navigate — just ask" is precisely why ChatGPT hit 100 million users faster than any application in history.

The insight about eliminating friction turned out to be the key to everything. When I showed that Google Analytics mockup — where you simply ask "what were my top traffic sources last month?" instead of clicking through menus — I was describing exactly what's now being built across enterprise software. Every major platform is racing to add conversational interfaces.

The "third platform" framing was directionally correct. Conversational AI hasn't replaced websites or apps, but it has emerged as a genuinely new computing paradigm. People now start their day asking Claude or ChatGPT questions they'd previously have Googled. That's a platform shift.

I was even right about the generational angle. I'd positioned this as particularly relevant to "Generation Z." Today, younger workers are the fastest adopters of AI tools, using them for everything from coding to writing to research.

2016 slide, Conversational UI: why should I care? Chatbots as the third platform, ubiquitous in less than two years

What I got wrong

The timing, obviously. "Ubiquitous in less than two years" was about six years early. That's not a small miss.

But the bigger error was the technology path. Every example in my presentation — KLM on Messenger, branded chatbots, platform-specific integrations — largely failed. The 2016 Messenger bot boom collapsed by 2018. Those rule-based systems hit a ceiling almost immediately. They could handle "track my flight" but fell apart the moment someone asked anything unexpected.

I was imagining incremental improvements to existing NLP. What actually happened was a fundamental breakthrough: transformer architectures, massive language models, and emergent capabilities that nobody fully predicted. The revolution came from a direction I wasn't watching.

Here's the irony: I was right about the destination but wrong about the route. The Facebook Messenger bots I showcased? Mostly forgotten. The underlying vision of talking to computers naturally? Vindicated by technology that didn't exist when I gave that presentation.

The lesson

I'll be honest: I'm proud of that presentation. Not because I got the timing right — I didn't — but because the fundamental insight was correct when most people thought chatbots were a gimmick.

In 2016, the prevailing wisdom was that conversational interfaces were a novelty, a marketing stunt, maybe useful for checking flight times but nothing more. I argued they represented a genuine platform shift, that "one interface, unlimited capability" would change how we interact with computers. That was a minority view at the time.

The technology took a different path than I expected. Transformer architectures and large language models weren't on my radar. But the destination I described — talking to computers naturally, eliminating the friction of menus and interfaces, having a single conversational layer that could do almost anything — that's exactly where we've arrived.

Sometimes the value of a prediction isn't in getting the implementation details right. It's in seeing where things are heading before the route becomes obvious. The executives in those rooms heard something that sounded ambitious in 2016. Now it sounds like a description of their Tuesday morning.

The takeaway

If you're making bets on technology trends, focus on the underlying human need rather than the specific implementation. "People want to eliminate friction with computers" was a durable insight. "Facebook Messenger bots are the future" was not.

The need doesn't change. The technology does.

I was six years early and wrong about the mechanism. But I was right about where we were going — and I've got the slides to prove it.