Endava has now shared its AI transformation story on the Canvas stage across multiple events worldwide.
At Canvas San Francisco, Endava’s Regional CTO for Europe and Global Head of Dava.X AI, Joe Dunleavy, walked through how the company built Dava.Flow™ and what three-plus years of transformation has actually taught them about change. At Canvas London, Endava’s Global CTO, Matt Cloke, picked up the same thread from his perspective across the organization, unpacking how a 25-year-old services company decided to reinvent itself for an AI-native world.
Their two talks, told from different altitudes of the same company, add up to one of the most complete accounts of enterprise AI transformation we’ve heard.
Challenge
When AI can generate code, what exactly is a software services company selling?
That is not a rhetorical question. Three years ago, it was the live strategic problem facing Endava’s leadership, and the answer they landed on has reshaped how the company delivers, prices, and sells its work.
Endava is a global technology and consulting services company with more than 11,000 people and around 60 delivery centers worldwide. For decades, they have operated on one conviction: technology is the how, and people are the why. When generative AI arrived, that conviction got tested fast.
Matt Cloke described the moment vividly at Canvas London. A few months into his role as CTO, nearly every conversation he had turned into a version of the same challenge: now that ChatGPT exists, do we even need IT services companies anymore? He framed the choice the leadership team faced as an evolutionary one.
“Did we want to be the dinosaurs, put our heads back down and carry on eating the grass? Or did we want to be the tiny little mammals that are going to evolve and plot our path as we move forward?”
Matt Cloke, Global CTO at Endava
Endava chose to evolve, taking a specific approach. As Joe Dunleavy spoke to this transformation in San Francisco, the scarce resource in AI-assisted delivery is not technical output. It is the structured understanding of what problem you are actually solving, for whom, and under what constraints. Context quality, not code generation, became the insight that would lead to Endava developing their engagement lifecycle methodology, Dava.Flow™.
Solution
Before there was a methodology, there was a reinvention program. Endava called their change management Keystone and used it to examine every part of the business through one lens. The guiding principle of the Keystone project was deliberately ambitious.
“We weren’t looking to use AI on top of an existing process for a 10 or 15 percent incremental improvement. We were trying to think about, if I use AI, how does this completely challenge the way that I work?”
Matt Cloke, Global CTO at Endava
Keystone ran across six streams, from transforming how the legal and finance teams work, to how roughly 8,500 engineers design, code, and test, to talent acquisition, sales, and Endava’s own internal IT systems. The team became what Matt called client zero, experimenting on themselves first. That experimentation surfaced a clear pattern in how AI adoption matures, in three stages:
- The first stage was assistive: simple code completions a developer accepted or rejected. Matt was refreshingly honest about how that went at first. Endava deployed 4,500 GitHub Copilot licenses to its engineers, then looked at the data and found that 70% of the AI suggestions were never accepted by a developer.
- The second stage was collaborative, where AI helped reason about code, check whether testing was adequate, and summarized documents.
- The third stage, where the real change happened, was agentic: a person directing an agent to do work, and agents coordinating with other agents.
That last stage broke the familiar rhythms of agile delivery, the standups and the sprints, and pushed the team to describe something on the other side of agile. They coined a word for it: flow.
Around that time, Miro invited the Endava team to Amsterdam to collaborate on how Miro could support Endava’s business transformation. Out of that work came Dava.Flow™, Endava’s proprietary AI engagement lifecycle methodology. Rather than layering AI onto existing delivery processes, it reimagines the entire engagement around one premise: get the context right first, then let agents work. It runs in four phases:
Signal: qualifying the right problem
Before any AI agent writes a line of code, the work is anchored in a clear understanding of the business challenge, competitive context, and desired outcomes. As Matt noted, a signal can come from anywhere: a meeting transcript, a photograph, even a scribbled piece of paper.
Explore: structuring and enriching that context
This is where information from Signal gets synthesized into the structured inputs that give AI agents the guardrails they need. Joe Dunleavy was direct about why this phase is critical: “By the time you get to the agents writing code, they are working on a focused context. It’s a very managed, controlled environment. That’s where you get the higher quality of the output that is important to enterprise customers.” The most important output of this phase, Matt added, is an agent-ready backlog of work.
Govern: human oversight during delivery
Teams stay in the loop as AI agents do the technical work, iterating toward production-ready output. Endava no longer calls this building. They call it humans governing the output produced by AI agents.
Evolve: monitoring in production
Telemetry and performance data feed back into Signal, creating a continuous loop rather than a one-time handoff, and opening the door to predictive analytics and fixes.
Crucially, Dava.Flow™ keeps humans firmly in the loop, which matters enormously when your clients are banks, payment systems, and insurers who cannot risk, in Matt’s words, a “hallucinated insurance return.”
Endava built their entire Dava.Flow™ knowledge base and part of the delivery environment on the Miro canvas, from onboarding materials to the live working boards where client context gets captured and refined. The Canvas is the place where distributed teams collaborate on the information that makes the whole methodology work.
Miro Sidekicks support the Signal and Explore phases specifically, helping teams surface insights from client conversations and structure them into reliable inputs. Endava has gone further and built its own Sidekicks on top of that foundation, connecting through an MCP connector into its context warehouse, enriched with the firm’s prior experience with each client. Matt’s favorite client demonstration shows the payoff: creating virtual personas, running virtual workshops, and updating diagrams in real time as live feedback comes in.
“The entire criteria, the roadmap, all of what makes up Dava.Flow™ — we presented all of that information in Miro, in a way that people can actually use.”
Joe Dunleavy, Regional CTO for Europe and Global Head of Dava.X AI at Endava
Impact
Dava.Flow™ is live and already winning work. Joe cited active programs with large enterprises, where the methodology is being used to scope, deliver, and evolve AI-enabled solutions.
The strongest proof point of Endava’s success with their new, AI-native methodology came from Matt, and it speaks directly to the quality concern enterprise buyers often raise. Using independent measures and external reference points, Endava found that the quality of the output the company now produces is better than the output it produced when it only had people delivering work. The reason is not that AI runs unchecked. It is that augmenting people with AI frees them to spend far more of their time thinking about the problem, which produces better results in the end.
As AI handles more execution, Endava’s teams invest more time in problem qualification, context-building, and client partnership; the work that creates new value rather than automating existing processes. Both leaders, on both stages, drew the same sharp line between the two modes. Here’s how Joe put it:
“The world, at the moment, is obsessed with value capture — how do we get our ten, fifteen percent of automation? But that’s losing the point. The opportunity in AI is really strong collaboration and the creation of new things. Value creation is what we’re here for and excited to help deliver with our clients.”
Joe Dunleavy, Regional CTO for Europe and Global Head of Dava.X AI at Endava
Matt, citing Stanford d.school’s Jeremy Utley, reinforced this point in London: value capture is a defensive improvement, while value creation means inverting the problem and using AI to reach entirely different places. That distinction is also changing how Endava sells, moving toward outcome-based models that reflect the new economics of AI-assisted delivery.
The bigger picture
The most candid moments from both leaders were not about what’s working. They were about what is hard.
“The technology is ready. It’s the worst it’s going to be today — it’s only going to improve. The challenge of doing an AI transformation like this isn’t waiting for the technology to be perfect. It’s bringing people along on the journey with you.”
Joe Dunleavy, Regional CTO for Europe and Global Head of Dava.X AI at Endava
Matt reached the same conclusion from the engineering side. Making Dava.Flow™ work across the organization, he said, was never about handing people training material and walking away. It was about shifting mindset, which is far harder. The work is to make experimentation feel normal and to get people’s hands on the keyboard. For a company that sells to regulated industries, it is also about reassuring clients on the unglamorous but essential ground of trust, governance, control, and auditability.
As Endava continues to navigate this journey, here’s what they’ve found actually moves the needle:
- Tiered training works better than broad rollouts. Endava built a structured program, from apprentice-level onboarding to team-level practice, so that AI fluency could grow at different rates across different roles without leaving anyone behind.
- Champions matter more than mandates. Getting more than 11,000 people to genuinely change how they work requires people inside each team who believe in the direction and model the new behavior day to day.
- Groups like Legal, HR, and leadership have to move together. Adoption stalls when operational functions are not brought along with delivery teams. The organizational infrastructure is as important as the technical one.
- Strong partnerships accelerate everything. “We don’t have to do this alone,” Joe said. “Where you’ve got strong partners, that’s the play.”
Running underneath all of it is something simpler that Matt kept returning to: visual context. The ability to bring people together to collaborate in the place where they are most comfortable, he said, really makes a difference. The methodology is the visible output. The training programs, the champions, the cross-functional coordination, and the shared canvas where people actually do the thinking are what made it real, and what any organization attempting a similar shift will need to invest in first.d what any organization attempting a similar shift will need to invest in first.
