Q: Can you tell us about your background and how it led you to start Pair?
I started off with probably the least sexy AI career in history — looking at neural nets in ferrets’ brains, and then training neural nets to diagnose trench foot in soldiers. Rewind 15 years ago, I was probably applying machine learning to one of the least sexy areas in the world.
Fast forward 10 years, the tech had got a lot better and I was fortunate enough to get to apply AI to some of the country’s biggest decisions, building the data science team in 10 Downing Street. It was the time of COVID and dashboards and data were really coming to the fore of the public mind. But this was the first moment I thought really carefully about the adoption problem. We built these amazing dashboards for the PM cabinet and did some really cool predictive analytics, but we were surfacing those insights to cabinet members of which every single one except one person had a qualitative background. Making sense of those insights and actually making use of the power of AI was an adoption problem.
Then just before I started Pair, I was the CTO of a specialist military regiment. It was very clear, very quickly, that GPT-4 was going to be a very powerful technology. But again, it came down to how people actually used it. The only time results came was when that incredible technology was actually pointed at the right problems to solve.
That brings me to now. I founded Pair a couple of years ago with my old lead software engineer from the specialist military unit. We really wanted to help others adopt AI effectively because we could see that the adoption problem was going to be the real issue.
Q: What are Pair’s values, and what does success look like for the company?
Our number one value as a company is that humans remain the main character. I suspect a lot of people might think that’s a bit weird for an AI company focused on AI adoption. But it’s really important to us. I’m on team human, I’ve got kids, and I want them to have good jobs and feel fulfilled. I think work is a really important part of what humans are here to do day to day and part of a really meaningful life.
We see the future being one where humans are still in the workforce doing meaningful things and AI is just helping us do things we thought were unimaginable previously. AI is undoubtedly getting more powerful, and that’s just bringing more and more exciting opportunities. It doesn’t have to just mean mass displacement and us all chilling on a beach, cashing out our UBI. I don’t really subscribe to that future.
Q: What is Pair?
At a very simple level at the moment, we’re a software platform that provides personalised AI training. Most AI training just shows you how to write in the box: it shows you a couple of features and gives you some generic examples, and unfortunately that doesn’t really transform how people work. It’s really hard to reimagine how you work if you are working as a project manager in an aerospace company, or a bid writer in a manufacturing company.
What our platform does is individually personalise each bit of training to you so that you learn how to apply AI effectively to your role. The most common feedback we get is: “The platform gets exactly what I do in my job and completing your assignments was directly translatable to ROI.”
Where we’re going is much richer than that. Capability building is just one part of AI adoption. As we’ve had more and more users on our platform (we’re in the tens of thousands now!) we see our customers have very similar challenges. We see them building thousands of agents through our platform. One of our customers built a thousand agents in a month just through completing Pair certifications. When you suddenly have teams that three months ago were just humans and now are teams of humans and hundreds of agents, there are really interesting performance review implications, governance implications, and implications for how you organise a team.
We also see that organisations have got their one subscription – perhaps Microsoft Copilot or ChatGPT – and they’re asking, am I getting ROI on these licences? Generally without effective training, they’re not. So we can help them make ROI decisions on which licences they go for, and in the future how they start to broaden out their AI subscriptions into more agentic tools.
Currently, we are a market leading personalised AI training platform. But we want to become the AI adoption layer for enterprise.
Q: What are organisations getting wrong?
Senior leaders are often just thinking about this wrong. They understandably want a top-down solution: the five agents we’re going to build to transform what we do, allocate cash to them and project managers, and then report to the board in a year. But what we see in reality is that GenAI tools are best used by those with domain expertise, and they tend to be those at the coalface.
If you give knowledge workers access to good gen AI tools and really uplevel their capability, the productivity gains are amazing. You start to get these seed corns of projects that can become really big, well-funded, top-down initiatives. So I think that’s the thing most leaders get wrong: they think this is a top-down problem, which they can allocate a few million pounds to and come back to in 12 months, when actually it’s quite bottom up.
The best organisations we work with have leaders who are familiar with the tools. They take time to reimagine their own work and go, “Oh my God, the AI training we did where we told everyone about the risks and not to put personal data in these tools – that was actually a net negative on our AI adoption.” Instead, we should take it very seriously how we empower people bottom up to transform how they work, because there’s just massive ROI there.
"I think that's the thing most leaders get wrong: they think this is a top-down problem, which they can allocate a few million pounds to and come back to in 12 months, when actually it's quite bottom up."James Kuht MBE ~ Pair
Q: Can you walk us through what the product actually does?
People can try it out – the first module is free.
Essentially, you log in, put in your role and where you work, and we use that to give you a light-touch diagnostic of where we think AI is going to help you. We use that data model of you – your role, your company, your focus areas – to personalise your AI training.
There are three levels: fundamentals, proficiency, and mastery.
Fundamentals is all about writing more effectively with AI – prompt engineering and things like that. The average knowledge worker spends 19.8 hours a week writing, so it’s a good place to start.
Proficiency covers data analysis, prototyping, image analysis and generation, generating presentations. By the end of that, most people have built their first agent and done their first AI-generated PowerPoint deck and data analysis.
Mastery is about agents. By the end, people have usually built their second or third agent and had it marked, and they’re getting to the stage where they will be the best person at using AI in their workplace. These people have passed 10 AI-marked assignments where they’ve applied AI to almost every element of their job. By that point, if they’re not saving half a day a week and using AI 10 or 15 times a day, we’ve categorically failed.
Q: What’s happening in the organisations that have built thousands of agents through Pair?
There’s one in particular covered in the FT that I’m able to speak about. Mimecast, a cybersecurity company of about two and a half thousand people. They have Pair for their whole workforce, and 99% of their workforce are AI certified. And these aren’t multiple choice questions; these are hard assignments that are AI marked.
They’ve built thousands of agents just through their Pair certifications. You can look at their financials for last year and their growth, they’ve had a record year and nearly a billion dollars a year of revenue.
I can’t isolate Pair as the causal factor, but I’d like to think we had some contribution. They’re a shining light of what great looks like in terms of widespread AI adoption: very deliberate, bottom up, with senior leaders who really buy in and care about the outcome — which should be real business growth, not just cool experiments you can tell your board about.
Q: How do you handle resistance, from employees fearing their roles, or sceptical managers?
The best leaders help us do this themselves. They role model the behaviour they want to see. One example: I worked with a fantastic CEO of a manufacturing company called Graham Haw. He came on stage at an SLT meeting with an AI-generated song from Suno as his entrance music, which he’ll admit was a little bit cringe. But it just disarmed everyone in the room. He then had a HeyGen-generated AI avatar of himself where he poked fun at himself. And then he said, “I’m already certified on Pair and this is how it’s helping me.”
Senior leaders can really help disarm their staff by role modelling the behaviour they want to see. And then for us it becomes about building momentum. In most organisations we work with, we get to about 70% of the workforce AI certified within the first quarter. At that point, when seven out of ten of your mates have got AI certified and think it’s great, the other three quickly follow, it just becomes socially unacceptable not to. We’re social creatures and we want to fit in. If everyone else is using AI effectively and looks like they might get promoted, you don’t want to be the odd one out.
"In most organisations we work with, we get to about 70% of the workforce AI certified within the first quarter. At that point, when seven out of ten of your mates have got AI certified and think it's great, the other three quickly follow, it just becomes socially unacceptable not to."James Kuht MBE ~ Pair
Q: Looking five years ahead, what does enterprise AI use and adoption look like?
A few high-level observations. First, the hybrid team piece is going to come to life in a big way. 2025 was the year agents were overhyped. In 2026, they’re very real. Leading organisations will genuinely have hybrid teams of humans and agents, and team leaders are going to need to ask who is accountable for each agent: making sure it does the right job, delivers the right value, is governed appropriately, sunsetted when it doesn’t work, and replaced by an upgrade when one comes out.
Secondly, we’re going to move from transactional use of AI to more delegation. Right now most AI usage is pretty transactional: you go to a chatbot, ask a question, get an answer, go back to your workflow. Going forward, we’re going to delegate tasks of greater and greater size. Where I expect to get in a year or two is a tool that can take a whole task end to end. For example, it can proactively research the latest AI releases, come up with a compelling script, get Synthesia to build the video, design a personalised assignment, and by the time I wake up in the morning, another agent has quality checked all that work.
Thirdly, knowledge sharing around this is going to get pretty fascinating. At the moment all these organisations are out there having a go at building a customer service agent, trying AI for bid writing. A thousand flowers are blooming and some are beautiful roses and some look like trampled, discoloured daisies.
Hopefully Pair can become a platform where if you’re a bit behind in the AI adoption race, we can help you mimic the best practice of the beautiful roses rather than having to go through 15 sets of trampled daisies first!
Q: How do you see education and the psychology around work shifting as a result of all this?
This is an extremely important problem to solve. 18-year-olds are not leaving school equipped for the AI age. That’s actually where we spent our first year, selling AI training to sixth forms. Unfortunately that market just wasn’t ready, because schools don’t have any money and aren’t incentivised to drive this.
An 18-year-old who’s relatively smart with AI skills can achieve what five Ivy League graduates could combined just a couple of years ago. The country that manages to get their education system to grasp that is going to have a phenomenal workforce. At the moment most young people are scared of AI because they’ve been told not to cheat with it, not to put their data in it, not to use it. They use it a little bit – AI filters on Snapchat, whatever – but they’re just not harnessing the possibilities. That’s a crying shame and a big opportunity.
Q: Ten years from now, is the human still in the loop?
The human’s still leading, but they’re out of the loop on a lot of everyday tasks that we’d think they need to be in today. Humans are going to be doing less but managing more, and I think that’s very exciting. I want to be able to achieve 10, 20, 30x more. I want to see more of my ideas come to life. I already build 10 times as many prototypes now as I used to, and I’m probably about 50% more efficient at knowledge work. In the future, these systems are going to get so good that I can just delegate more and more to them. So increasingly I’m going to be the manager of humans, but also the manager of some incredibly powerful AI. I can’t wait for that future – and to bring it to a lot of our customers too.
Pair are hiring across all roles.