Building AI at scale is easy. Getting people to use it is hard. Reimagine the workflow or settle for a rounding error.

We call ourselves an agentic customer experience partner. That means delivering content, marketing, data and commerce solutions using agentic AI workflows. The opportunity is to deliver what we have always delivered, digital experiences, and to do it agentically. Doing that at scale is the challenge, and no one has done it before. A few companies are at our heels, and a lot are waiting to see where the market goes. It is inspiring to be in the lead and it is terrifying, because there is no blueprint, so you pivot very frequently and test what works.

Agentic customer experience is twofold. Our content, commerce, marketing and data solutions are delivered with agentic workflows, so they are dynamic, they are not static and set in stone. And our delivery teams are a matrix of humans and agents working together, where humans set the agenda and the direction and agents scale and execute.

A lot of people look at AI as another change, and that is true. We saw it with mobile, with social, with digital. But it is more than that, because AI changes the business of services. Historically services have been about people: large integrators staffing accounts with hundreds of people, billing time and materials. AI makes it about outcomes, outputs and business value. That democratises size. A company like Apply with 700 people can now go neck to neck with the very largest firms, because we can deliver those solutions at scale.

Business impact is the only measure that counts, and anything we measure is rudimentary if the client does not buy into it and it does not tie to their goals and KPIs. Often the client does not have that information, so our job is to sit down and work out how we truly move the business. Once you have the KPIs, you can engineer the engagement around them. When you do this you have to do it with fixed price, and fixed price is scary for agencies and clients alike. You have to do it now. When the client sees the value and the savings from an outcome and output based engagement, it is crazy not to go with it.

We are still very early stage, because of training and adoption, and because there is a lot of hype. When it is not done right, AI falls on its face. The brands that fail to deliver value take AI and overlay it on their existing workflows.

You get 20 to 30% efficiency savings, and once you count the cost of building it and the usage cost, that is negligible. A CFO looks at that and asks what they are paying for. Where we have seen success is when you take AI and reimagine your workflows. That is when 20 to 30% turns into 20 to 30 times. People get it wrong when they use AI as an accelerator for existing work, because they think of AI as ChatGPT: you sign up for a chatbot, adapt it internally, build a bespoke solution, and now you are stuck with it, expecting employees to work out where they fit.

Most people think it is a technology gap, but we all have access to the same models and the same solutions, and everybody can afford it now. It comes down to training, enablement and governance: training your staff, enabling them with the latest technology, then putting the right governance around it so you are not putting your brand at risk. Those three things become a rocket ship for adoption. You still need the tools and clarity on what you are building, but that is the separation.

The first thing we did was build our own custom Apply training programme. The content comes from partners like Anthropic and Google, but the framework is ours: the streams, the training tracks for different roles. We built a badging system and we expect everyone to reach a certain level, because that is how we work now. It has worked really well, and since there is no blueprint we keep testing and adapting. Beyond that, the companies that get past the first phase invest in the right partners, people who have done this before and will sit with you to reach your goals.

The economics rely on the people. We are in the people business, we do not have products, our products are our people. That is our moat. AI amplifies everything: a good team becomes a great team, a bad team becomes a really bad team. So the job is making sure talented people who are masters of their craft have these tools. Historically the service industry was about bodies and seats. Now it is about top talent, because that is where the multiplier comes from.

With finance leaders, what works best is sharing our own experience, because we went through exactly what we ask clients to go through: the challenges, the training programmes, the enablement, the tools and the governance. When I meet clients to talk about AI in project work, both sides need to be incentivised to innovate. When they are, we go with an outcome based engagement and everybody benefits: the client gets the AI discount, we get our margins, our people get the tools, and the outcomes lead. When it is one sided, it is not a healthy relationship.

Building AI at scale is easy. Getting people to use it is hard, and that was my bigger challenge at Omnicom, where I built and deployed Omni AI, our agentic platform. Seeing the excitement in the hands of a 100,000 person organisation was very promising, but the harder task is getting people trained and enabled. What I wanted was to take that level of solution to every client, which is what I am doing at Apply. We deploy our people alongside our solutions, and we build purposeful point solutions on Google Cloud, which in our mind is the only cloud platform spanning models, agents, cloud and data. Clients gravitate towards what works; when they see it running, it is easier to picture themselves using it.

The hardest part is the people. There is understandable fear around job security, and you cannot help someone adopt from a place of fear and panic. I see three categories. Evangelists are ahead of you and just want access. Detractors, you can do almost nothing to change their mind, and some will find themselves out of a job while the enabled will be in very high demand. It is the people in the middle, the overwhelming majority, who are open but scared, busy, tired and unsure, and they are worth the time and effort. I always tell people AI will not replace your job, someone with AI will.

You always have to have a human in the middle, because AI is unpredictable and indeterministic. It uses probability and sometimes produces things that are wrong, unsafe and risky for your brand, so a human authenticates and authorises before publication. Put that aside and AI is very good at concepting, ideation and scaling content. On the governance side we are working towards ISO 42001, which we understand to be the first certification focused on AI governance, safety and security. It is focused on AI usage, the nuances of the models and their connectivity to tools, where data access partitioning is suddenly multiplied. That partition has to happen for every single account. It is not about the badge; the process is teaching us to be better at governance.

We are seeing a graveyard of applications being built across the industry. Building an application used to take developers, product managers, creative strategists, time, effort and money, so you thought hard about it. Now anyone on a whim can pull up cloud code and build one, and they are built with no purpose. We wanted to avoid that and avoid redundancy: everyone will create a skill to connect to Gmail, everyone will create one for Slack. So we are building an internal marketplace where all the skills, tools and applications are published, and people can reuse them, improve them and discover what else they can do.

It is simple: EQ will become the separator, because with AI much of the nuance of coding and rolling out development becomes easy for everyone. What matters is EQ and mastery of your craft, the forethought to make a decision and understand its nuance. We label our interviews: a skills interview, an EQ interview and a demonstrate your abilities interview, each run for that purpose. An EQ conversation looks for empathy, recognition of a situation and the ability to identify with the person across from you. Those things are hard to come by, though not impossible to train.

It applies across the board. AI will elevate everyone's role towards high EQ, thoughtful tasks, because it will eat the automatable, simple, mundane, rudimentary work at the bottom. Everyone has to raise their game. With clients, we share our experience, because we know their anxieties, and in some cases we have shared our training framework. The best way to teach an organisation is to show it how another organisation has done it.

What triggered my founder mindset was my doctorate, in video game psychology and educational theory. It taught me to think critically about every conversation and it taught me the EQ I keep describing. Even in a larger organisation like Omnicom I wanted to build something rather than be part of the machine, because there is much more satisfaction in founder pride. You have to be really good at what you do, and you also have to get lucky. I got lucky with Apply: a really good team, the breadth of clients and the work we have done have propelled us to where we are. For years I wondered what I would have done in 2000 knowing what I know now. I do not have to wonder anymore, because I am living it, and living something even bigger.

I will not dare predict three years out. The rate of change with AI is unlike anything we have seen and it is only going to accelerate. Six months I might manage: we will start seeing organisations that really buy into 20 to 30 times instead of 20 to 30%, truly reimagining the way they work. At 20 to 30 times, usage, cost and investment are negligible. At 20 to 30%, all of it matters. The advice I would give my younger self is do not settle, always challenge yourself. I have rarely stayed in one company beyond two or three years, because I get bored. My wife says the twinkle in my eye goes away after two to three years, so I need a new challenge.