How Endava Is Scaling AI-Native Delivery with Devin

Endava is a publicly traded technology services company headquartered in London. Its engineers design and build custom software for enterprise clients across payments, banking, healthcare, retail, and other industries, spanning AI solutions, cloud infrastructure, and digital platforms.

Recognizing early that AI would reshape the technology services market, Endava launched Keystone under the sponsorship of CTO Matt Cloke. Keystone is the company-wide transformation through which Endava is proactively redesigning its capabilities, ways of working, and commercial model for an AI-native future. Out of Keystone came Dava.Flow, Endava's AI native methodology for delivering client work, spanning the full lifecycle of change from ideation through production release and continuous evolution. Cognition's Devin is a key enabling platform within Dava.Flow, supporting agentic engineering within Endava's broader, tool-agnostic, and human-accountable delivery lifecycle.

As Rob Purcell, who co-leads the initiative within Endava's Dava.X AI group, puts it:

"Dava.Flow wasn't created and then we went looking for a tool. It was created with these products from the start."

— Rob Purcell, Dava.X AI Group, Endava

Today, more than 2,000 Endava employees are provisioned on the Devin platform, and adoption reaches well beyond engineering into business analysis, product, and Endava's growth organization. Devin is already at work on more than 20 active client engagements, plus eight internal projects that help jumpstart future client work.

Endava's journey with Cognition

Endava started with 200 seats in late 2025. Six months later, 2,000 people are on the Cognition platform. Dava.Flow provides the framework through which Endava builds AI-native delivery capability, with teams learning the methodology, governance practices, and enabling tools together. Access to Devin is embedded in onboarding for the methodology, so every engineer trained on Endava's new way of working is trained on Cognition's tools concurrently.

The deployment spans the platform. Devin Desktop is the primary tool, and Devin Cloud, Devin Review, and DeepWiki are in use across a growing set of engagements. The Cognition tools do the heavy lifting across each phase of Dava.Flow: Signal, Explore, Govern, and Evolve. In Signal, teams capture and shape the right problems to solve. In Explore, that intent becomes clear, agent-ready specifications. In Govern, agents handle the build and test while engineers review changes and enforce standards. In Evolve, the product is monitored with AI-enabled analysis of metrics that can surface new Signals. Endava has certified Devin within Dava.Flow through its emerging certification practice, which validates the agent and model combinations known to work well with the methodology and guides how much preparatory work a given client engagement will need.

Deep adoption on client work required structural change. Endava reviewed its contractual arrangements across its customer base to make AI-enabled tooling part of its standard terms of service. Not every engagement is client-facing: internal projects split between Endava's own IT work and accelerators, starter builds the company invests in and later reuses on client engagements.

What's changing inside Endava

The most striking shift is in how Endava's people work. Code creation has accelerated dramatically, and much of the toil that bogged down engineers has been removed. Despite initial skepticism, sentiment among engineers has turned sharply positive as teams grow familiar with Devin. Engineers are rebalancing toward solution design, thinking more deliberately about customer and business needs rather than keystrokes.

The shift is measurable. Since rolling out Dava.Flow with Devin, Endava has seen delivery productivity and efficiency increase by roughly 30 percent across selected workflows, the clearest measure of throughput per team. Problem-to-backlog conversion is 40 to 60 percent faster, shortening the path from business need to delivery-ready work.

Requirements-led change requests and clarification loops are down 20 to 40 percent, and unresolved critical assumptions before build have fallen 50 to 70 percent, letting teams start with clearer scope and fewer avoidable delays. Endava measures these results on selected workflows, baselining productivity and end-to-end cycle time before and after adoption.

Devin usage also extends well beyond software engineers. Business analysts, product managers, and members of Endava's growth organization all use Devin to accelerate their part of the delivery lifecycle. Dava.Flow pairs this acceleration with rigor: human review remains a core facet of every engagement.

Enablement has matched the scale of the rollout. A single round of Cognition-led workshops reached 1,100 Endava engineers across APAC and EMEA, with more than 500 joining hands-on labs. That momentum now sustains itself: recurring office hours let a growing community of internal evangelists share learnings and best practices across the rest of the organization.

Why Cognition

Endava's evaluation of Devin was effectively a head-to-head against the incumbent AI coding assistant already prevalent across the organization. Two things stood out then and remain true today. The first is the pace and scale of product improvement, including the near-immediate availability of newly released frontier models. Feedback raised by Endava's team shows up in the product on short turnaround. "It feels like we're being listened to," Rob says, "and the product is much more capable than it was even a year and a half ago."

The second is enterprise readiness. Endava works on behalf of clients for whom data privacy is paramount. Cognition's model-agnostic platform, zero data retention guarantees, and consistent security baseline across every available model provider made it viable to handle highly regulated client work from the first conversation. Cognition already had enterprise proof points across Financial Services and Federal, two of the most heavily regulated industries. That track record built trust from the outset and let Cognition's experience speak for itself.

What's next

Endava sees Dava.Flow as its answer to the question now facing the entire services industry: what does a firm that builds software for clients look like when autonomous agents like Devin can build and ship the software itself? The company's insight is that value is shifting from writing code to architecting everything around it: shaping the right problems, specifying work agents can execute, and governing quality at speed. The firms that master that operating model first will win the next decade of client work.

The near-term plan follows from that. Endava is bringing Dava.Flow into dozens of conversations with new and existing clients, meeting sophisticated enterprises deep into AI-native development as peers and bringing earlier-stage organizations up the curve. Client pilots have moved past slideware: demonstrations now walk through working software built with Devin against the client's own business needs. From there, the ambition is deeper client adoption of Devin, Devin Review, and DeepWiki beyond the initial teams.

We're excited to keep working with Endava as it scales AI-native delivery across its client base. Organizations interested in deploying Devin at scale can contact our enterprise team to learn more.