What does AI adoption look like in reality? You know about the hype and the obvious potential, but what about the everyday impact, measurable value and long-term sustainability?
Answers came at this year's AI Academy at SBC. Kanda Kumar, Director of AI at evoke plc, gave us one of the most concrete talks on enterprise AI adoption in gambling this event has seen. For anyone going through the process of their own AI transformation, it was both a reality check and an inspiration.
Kanda joined evoke in 2024 to build Enterprise AI from scratch, after 18 years leading enterprise AI and automation transformation at Ericsson, a business operating in more than 180 countries. For Kanda, Enterprise AI isn't a set of isolated use cases. It means redesigning end-to-end customer, employee and vendor journeys across every function.
Everything was grounded in numbers. His starting point was to agree with Finance up front how value would be measured, and report EBITDA impact to the CEO and Board rather than adoption metrics or pilot counts.
That discipline shaped everything that followed. By the end of 2024, four AI solutions were live in production with signed-off EBITDA impact. In 2025, evoke assessed more than 1,500 business processes and built what Kanda called an 'AI Factory'. This was a repeatable route from business problem to production solution.
By the end of 2025, four solutions had become sixty, all with EBITDA signed off by Finance.Today there are more than 120, and evoke has built a shared platform for running AI agents safely across the business, almost four times the investment. That’s measurable value.
The most striking part of the talk wasn't the scale but what evoke learned from embedding AI into day-to-day work. Anonymised, aggregated analysis of how around thousands of colleagues used evoke's enterprise AI tools, covering more than 600,000 AI assisted activities, surfaced eighty-two recurring work patterns.
Those patterns now feed evoke's agentic pipeline, alongside the transformation opportunities that come from business strategy.
On agents specifically, Kanda described a two-stage approach shaped by how much a regulated business can trust them:
- Human-triggered agents, where someone starts the agent, observes what it does and reviews the output. These are live at scale today, automating repeatable tasks such as pulling data together across multiple systems.
- System-triggered agents, where an event such as an AML alert starts the agent to execute the process automatically, but still passes to a human for judgement. These are being industrialised next, under governed, risk-based autonomy, where humans intervene for exceptions.
Kanda was clear that the hardest constraint evoke faced wasn't the technology but organisational alignment. Every function had formed its own reasonable but different view of what AI meant for them. Agreeing on a shared, Finance-backed measure of value gave everyone a common language to work from.
In a regulated industry, governance was designed in from day one. Human accountability for customer-impacting decisions, transparency, explainability and contestability are built into every workflow. Because small errors compound across multi-step agentic processes, controls sit at each step rather than only at the end.
Kanda's advice for anyone starting this journey
- Agree how value will be measured before you build anything
- Redesign work into agentic workflows with human accountability
- Embed controls by design rather than after the fact
- Build for reuse rather than isolated pockets of AI
- Focus less on what the technology can do and more on what your organisation must look like to use it at scale
As Kanda put it, evoke didn't get there alone. CreateFuture, AWS and other partners played a key role in helping build and scale that capability. We'd be happy to talk to you about our role in evoke’s story and the lessons that could be applied to your organisation.