Part 3 of 12 – Conversation Intelligence Platform series Back to Part 1
The thing about infrastructure decisions is that by the time the rest of the industry makes them, you’ve already forgotten why they were hard. The difficulty was in doing them when the path wasn’t obvious, the tooling wasn’t mature, and most of your peers thought you were being eccentric.
We’ve been here before. A few times, actually.
Our first API dates to 1996. They weren’t called APIs then – they were called “web services,” and they were largely a theoretical future vision. I recall Microsoft writing whitepapers about what they might eventually become. I was writing ours. Some of our competitors still don’t have usable APIs, thirty years later. That gap, measured in decades rather than product cycles, is either a damning indictment of the carrier industry or a useful reminder that innovation requires someone to go first.
The L3 switch story is more instructive. At the time we replaced traditional routers with L3 switches running selective routing patterns, we were aware of exactly one other organisation doing anything similar at scale, and they were Spotify – not a company most telcos were looking to for inspiration. The conventional wisdom was that you used proper routers for serious routing. L3 switches were for enterprise LAN work. We disagreed, we did it, and it’s why a fibre going hard down in Slough doesn’t produce a single dropped call.
Then virtualisation. We moved to virtual infrastructure before most of our competitors had heard the word. More to the point, we left virtualisation behind for containerisation before most of them had even adopted virtualisation. As far as I’m aware, nobody else was doing what we were doing with containerisation in 2015. That’s not a ten-month lead. That’s a ten-plus-year lead, and many of those others are still catching up on virtualisation. It is, I’ll admit, a slightly unglamorous position to be so far ahead that you’re effectively alone – but I’d rather be alone and right than in good company and wrong.
The documentation story deserves honesty here. In most of these moves, we built the thing and documented it… eventually. Or not. We’ve been genuinely dire at documentation over the years, and we’ve survived primarily because our customers are unusually smart and patient. The Conversation Intelligence platform is different – Part 2 covers why – but I’m not going to pretend we earned the right to claim documentation-first as a Simwood principle. We earned the right to claim it as a Simwood achievement, for this platform, right now.
The Conversation Intelligence platform is the same move, made again – but this time with a product dimension that didn’t exist before.
Carrier-grade AI applied to conversations in real time is not, right now, normal. Nobody else has built what we’ve built. The market has point solutions: third-party transcription services bolted on after the fact, AI features that sit outside the call path and require you to route recordings to a cloud service and wait. What nobody has is a carrier doing this at the carrier layer, in real time, with the full architectural separation of Agents, Operators, and Conversation Memory that we’ve built.
This isn’t just telephone calls either. Via the Simwood Potato, any voice passing through our platform is in scope – Teams calls, WhatsApp Calling, any channel. Customers whose service is locked with carriers who’ve invested in nothing for years can access Conversation Intelligence via BYoC without porting a single number. The same platform capability that enterprise emergency services teams can deploy is available to anyone whose carrier can deliver the call to Simwood – we process it and hand it back as if it came from a Simwood number. No porting required.
How long before the rest of the industry catches up? Based on the history above, I’d estimate somewhere around fifteen years. Which sounds like a joke until you remember that I genuinely don’t know of a carrier who containerised in 2015 who isn’t us. The industry moves slowly when it moves at all.
What matters is that we built it now. Our customers get access to it now, while the platform is young and the relationship between Simwood and the teams building on it is collaborative rather than transactional. By the time this becomes industry standard, we’ll have a decade of production learning that no late adopter can buy. I do wonder who the next startup our competitors will acquire for £46m in the hope of catching up. It’ll be a startup built on Simwood.
I’m genuinely proud of this. Not in the way that press releases talk about being proud to announce things (and if you ever see that phrase in a Simwood post, someone has failed the editorial process). I mean I find it quietly satisfying to look at what Charles and his team have built and recognise the same instinct we’ve always had, better executed than we’ve ever managed before
We’ve been the only carrier doing X for long enough that it no longer surprises me. This is the latest X. It won’t be the last.
Previous: Part 2 – The AI-accelerated build
