The AI-first playbook: finding your idea
A conversation with Paul Taylor, Founder and CEO of Thought Machine
Paul Taylor, is the founder and CEO of Thought Machine, which has built a modern cloud native core banking system that’s now trusted by everyone from Tier 1 multinationals to regional banks and fast-growing fintech companies worldwide. They recently became Mastercard’s first end-to-end partner in core banking.
We became friends after he sold Phonetic Arts - which built AI software to generate natural expressive speech in video games - to Google. We got to know each other over many coffees and lunches, discussing progress in AI research and its potential applications. After initially exploring how AI could overhaul financial services, Paul decided to go in a different direction - fixing broken core banking systems.
At Air Street, we see a lot of gifted proto-entrepreneurs rushing to start the latest ‘AI for X’ without doing their homework on the market or potential customers. Paul stands out as someone who took the time to scope out the opportunity properly, so his team could build something the customer couldn’t live without. We’re grateful he’s agreed to answer a few questions on his journey.
Before starting Thought Machine, you’d co-founded and sold two companies. Why did you feel the need to start out on the journey again?
When I left Google I didn’t have a set plan, but rather I wanted to engage in the startup community again. Even in the three years I was out (2010 to 2013) I noticed a ton of progress - London was really moving ahead. There were so many good opportunities and the investor base had grown significantly.
Like many, I considered an angel or VC role, but it’s really not my kind of thing. As VCs like to say, I am an operator not an investor…
Phonetic Arts had an obvious connection to your prior academic work on speech, but Thought Machine was something completely different. How did you narrow down the problem space you were interested in?
While everyone is of course very excited about AI, few realise how mind-bendingly difficult it all is. Only a few people or groups can really build these models and that has always been the case. At Google, I was the only one who had a deep understanding of why the algorithms worked and how to make a change to get a desired outcome. It is lonely and frustrating and each piece of work can take months or years.
However Google showed me you can build a successful company or product line by tackling a problem with many known solutions, but by doing it far, far better. Both email and online maps existed before Google, but Gmail and Google Maps were fast and reliable.
So this led me to ask myself - what’s a known problem, with an existing but poor solution, where excellent technology could add real value?
At that time, enterprise software was going through an on-prem to SaaS/cloud transformation, so I instinctively felt there was something there. After talking to many banks, we arrived at the problem to end all problems in financial services: core banking.
Core banking was perfect for us. Most banks use legacy technology, which ranges from bad to catastrophic. It often runs on mainframes and is reliant on software written for a different era. It is expensive, inflexible, hard to understand, hard to change. It isn’t in the cloud, isn’t real time and has a ton of other disadvantages. We thought that if we could bring the best practices of modern cloud computing to bear, we could build a far better platform and unlock value for banks.
Did you have a process for identifying people to meet on the customer side? Did you find they were willing?
We didn’t do any interviews. We just asked people who worked in banks. We found it remarkably easy to get intros - banking is very networked and many consultancies provided intros.
Banking is drowning in PowerPoint, so we quickly saw the potential of live demos. We did as many of these as we could and found they really had an impact.
How did you go from concluding a problem existed in core banking to working out the shape of a commercially viable solution?
I set two engineers a simple task. You have 6 months to build a credible demo that we can show to a bank. It didn’t have to be a product, just a demo. I didn’t insist on any further requirements.
Our two best people (Will Montgomery, now CTO, and Peter Dudbridge) did this intensely for 6 months. In the end we had an excellent demo with two made-up banks (NatWast and the Royal Bank of Scotland), where we could send money between the banks, see a basic bank app, ledger and interest rate calculation engine. It was connected to the card rails. The banks we demoed this to were blown away.
Banking can be a very conservative industry - what led you to believe there would be sufficient interest in an outsider overhauling their core systems?
In my mind, the insider/outsider thing is just an excuse that failed start-ups use. The world is far more meritocratic than that. The pertinent questions are:
Have you found a deep and real point of pain? Yes, enormous.
What do the banks think of the incumbent suppliers? They hate them.
Are you a credible answer? Our demo was great, we came from Google, we know how the cloud works.
Is there anything you would do differently? What advice would you give someone setting out on this journey for the first time?
Even we thought revolutionizing all of banking infrastructure was very bold. In the early years, we hedged by doing other pieces of work too. We shouldn't have done this - our analysis was correct and we should have gone all in quicker.
Finally - and I think this is the most important lesson - be aware and observant. Don’t get lost in tunnel vision on the exact nature of your first idea. But when you do see a real opportunity, a massive gap in the market - go all in and commit everything you have, these are rare and you have to really go for it.