Sitting in the corner of the press lounge at SBC Lisbon next to the rattling espresso machine feels somewhat luddite compared to the words coming out of Kanda Kumar’s mouth. The director of AI at evoke, the parent company of William Hill, 888 and Mr Green, has just finished speaking on a panel with the likes of AWS and OpenAI. He tells me he used AI to find our meeting point.
Kumar joined evoke in 2024 after an 18-year stint at telecommunications firm Ericsson and was handed the task of building out the operator’s AI capabilities The London-listed operator’s 2025 annual report lays out its ambition to be an “AI-first organisation”. Last year, more than 4.4 million “operational tasks and process steps” were executed by the tech.
“In addition to the execution of our AI strategy, we will prioritise the transformation of all our workflows into agentic workflows, democratise AI across the whole of evoke to upskill employees in safe usage and to create an AI-first culture,” the report added. So, no small remit for Kumar, but one he appears to be approaching with genuine excitement and fervour. A brave new world, indeed.
EGR: When did you recognise AI could be a powerful tool for change?
Kanda Kumar (KK): AI is not something I’ve only recently started working on. I spent 18 years at Ericsson before joining evoke in 2024. During my time there, I led company-wide AI transformation. That was a company with around 120,000 employees operating across 180+ countries. We were one of the early adopters. I was then brought into evoke to help build the same kind of capability I had helped build at Ericsson. I recognised AI’s potential a long time ago, but the pace of change has been much faster than I expected.
EGR: How long did you think it was going to take?
KK: AI, in one shape or form, has been around since the 1950s or 1960s. We’ve been working a lot on machine learning, and then unstructured data was a big thing. Machine learning continued to improve through developments in deep learning and neural networks. I think it’s the generative AI that was the tipping point, especially when OpenAI introduced ChatGPT to the whole world. They democratised AI at a much larger scale. I think that’s when I knew it was going to be a big thing.
EGR: Are we beyond the point of no return? Are we firmly now in an “AI world”?
KK: This train is not going to stop. You need to get on. AI is playing a key role in everybody’s personal life. That has taken off much quicker than in the enterprise world because the models understand how you book a restaurant, how you plan a trip.
But in an enterprise, the model needs the relevant business context, access to appropriate information and controls around its use. That has taken a bit longer, but it is getting better and better at a much faster pace, and that is why the industry needs to distinguish between ChatGPT and enterprise AI.
EGR: How do companies differentiate when everyone reaches the same base level of AI use and engagement?
KK: There are different levels of using AI. AI use has become much more widespread over the past two years. Everybody is using AI, but the important questions are: Are you scaling AI? Are you focusing on getting value out of AI? Is it transforming your business? That’s how you measure impact. Using AI is very easy, but what are you getting out of it? How are you scaling it? Are you transforming the way work gets done? Because you need to.
AI is fundamentally changing how work gets done, which means organisations need to change their processes and ways of working. And that doesn’t happen a lot across companies. That means you need to unlearn and redesign and reinvent work. There is a huge gap between the companies starting to redesign work and those that are not.
Many companies are using AI without yet fully understanding how it could fundamentally change work or how they need to prepare for it. It means fundamentally rethinking and reinventing work. That hasn’t happened. Very few have done it.
EGR: What does it actually mean to be an “AI-first” business?
KK: Being an AI-first company involves a number of things Firstly; it’s a culture you need to create. AI is another technology. Yes, it’s very capable, but it’s another technology. It is there to help us. Like Microsoft Word or PowerPoint, it is a tool people need to learn to use effectively.
The second thing is that while it’s very capable, it can also put us at risk. You need to know how to use AI in a safe and responsible way. Then we are talking about rethinking and redesigning processes and ways of working because AI is disrupting traditional ways of doing things. People may have firmly established views shaped by 20 or 30 years of experience. They have a very clear view of what things need to be done or how things need to be done. They need to unlearn and rethink how work gets done in some areas. At evoke, we have made progress but there is still more to do but we are on the right track.
EGR: How do you bring along those people who are reticent to engage? You’ve just mentioned the individuals with “firmly established views”.
KK: The only way for us to do that is to democratise AI; give people access and help them understand how to use it. That’s one piece. The other piece is to demonstrate things AI has actually changed. For example, it could be customer services or it could be the registration process, for instance. You can show how introducing AI has changed the process compared with how it worked before.
That’s the only way they can see it in real life. Otherwise, it’s too abstract. It’s just theories and slides, so we have to show it. We have to show them the impact it has in the business, whether it’s efficiencies or whether it is customer retention or satisfaction or even revenue. Depending on who it is, whether it’s a C-level executive or someone building the solutions, you present the evidence in a way that is relevant to them. Evidence matters.
EGR: That evidence of efficiencies can also lead to job losses. What do you say to critics who argue AI will lead to a job apocalypse?
KK: If you go back in time, these questions came up, going all the way back to horses and cars. When the Cloud came in, all the data centres were disappearing, and people thought, ‘Oh my God, people will lose jobs’. Now you see the data centres are exploding. It’s the same here. There will be a transition phase. There will be new jobs created. Some traditional roles will change or disappear, for sure. Some of the roles across the industry such as data entry, have changed or disappeared. AI will come in and disrupt a number of roles, but also new roles will be created. A prompt engineer was never there before. Context engineering never existed. They are new roles within AI. The key point is the sooner you get onto the AI train, the sooner you will understand how your current job is transitioning or transforming into something else. The best thing to do is prepare for that transition and the opportunities it creates.
EGR: Could AI do your job?
KK: Yeah, definitely. It is capable of that, in theory.
EGR: Does it concern you?
KK: In theory it can do my job. It has the capability. I don’t think an individual understands how capable it is. AI is solving biological and mathematical problems that I will never understand. From that point of view, it is capable. It can do my job, but how do you take AI and make it do my job? That’s the challenge, and that is actually my job, turning those capabilities into something that works effectively within a business.
EGR: Can you see a world in which operators’ reliance on third-party supply collapses? Could an evoke build out a CRM platform or casino content internally?
KK: You mentioned CRMs, they’re also introducing AI capabilities that enable things that couldn’t be done before. From an operator point of view, suppliers have a lot more capabilities than before. Every CRM or any other platform you take, they all have their own AI capabilities. It’s very powerful. In an enterprise, they are only a piece of the puzzle. Multiple systems need to work together which creates integration challenges.
But, at this point in time, I don’t think we are going in a direction where the operators will build everything in-house. I think the operators will probably rely on fewer, larger, more powerful platforms, which need to be integrated.
EGR: What are some of the major challenges you are seeing with AI?
KK: There is value in AI, but the challenge is how to scale AI. In our industry, customer protection is at the heart of everything we do. Then you have the regulatory requirements that set clear boundaries. Human accountability and oversight remain essential for every customer decision. Then you have the models, and they are very capable. But if you take two similar cases, the AI models can produce different outputs.
With agentic workflows, if there’s a small error at one point in time, that compounds. Then you have the classic challenges of data availability and quality of the data. For companies that have legacy systems, integrating them with AI can be very challenging. You have these enterprise challenges and those are the things we need to fix. Some are quick fixes; some will take longer time.
There is a limitation as well. Even if the model is capable, there are limitations depending on which industry you are applying AI to. It’s very risky to let AI take more ownership of customer decisions. That is not the right thing to do. Which means, even where AI is capable, human judgement remains essential for high-risk customer decisions.
It’s extremely important we embed agent controls and decide what level of agency to give to the AI based on risk classes for every workflow. Something very simple that is internal, with no customer, employee or financial impact, you can give higher autonomy. Customer and employee cases and workflows are classified as very high risk, which means AI gets lower autonomy. I would advise anyone scaling agents to embed agent controls and risk-based autonomy by design.
EGR: Will there be another Big Bang moment in AI? Or will we just see a natural path which we will look back on and see the progress that was made?
KK: I’ll answer in two parts. AI is advancing at a very fast pace. The time to disrupt something is getting shorter. We are now talking about recursive self-improvement, which means the next version of the AI model can be built by AI itself and with less human involvement. It’s scary, and there are signs that it is starting to happen. That’s why there are a lot of discussions ongoing to slow it down.
The other part is about how quickly we can bring AI into the enterprise world or someone’s personal life. For me, if I look at my banking app, that’s still a bit traditional. A bank is still a bank, which means there is no switch that is going to come in and just change everything. But the disruptions will be much quicker, and whoever jumps on the disruption and rides the wave will also be successful.
In some scenarios and some areas, it will be very quick. In some areas, it will be slow. My instinct and experience tell me in an enterprise world it will be a little slower.
The post Q&A: Evoke director of AI outlines the need to “unlearn and redesign and reinvent work” first appeared on EGR Intel.
Speaking to EGR, Kanda Kumar reflects on AI’s impact on workflows, agentic autonomy, and how, in theory, the tech could do his job
The post Q&A: Evoke director of AI outlines the need to “unlearn and redesign and reinvent work” first appeared on EGR Intel.