Endava Treats Codex as a Lifecycle Agent, Not a Coding Assistant
Endava executives Joe Dunleavy and Mike Krolnik argue that Codex is changing software delivery less by speeding up individual coding than by shifting teams toward supervising generated work across the lifecycle. Dunleavy says small teams can deliver more value in compressed time as their role moves from producing code to overseeing Codex’s output. Krolnik says the tool also helps senior architects turn intent into usable artifacts and enables junior staff to produce more mature work, extending Codex’s role into planning, documentation, diagrams, and client-facing explanation.
Codex changes the unit of delivery from individual output to small-team leverage
Joe Dunleavy describes Codex’s value at Endava in terms of delivery compression: “small teams of people” are able to “deliver massive value” in a “very, very condensed timeframe.” His framing is not about faster typing or narrow code generation. The change he points to is organizational: smaller groups can take on more consequential work when Codex is part of the delivery process.
Dunleavy, Endava’s regional CTO, treats the shift as a before-and-after change in how work gets done. Endava’s role has moved away from producing much of the code directly and toward overseeing what Codex can produce. In that model, human work is centered less on first-draft production and more on oversight and review.
The shift has gone from us producing a lot of the code ourselves to us now overseeing the work that Codex can produce, and the quality of that has just kind of gone up exponentially.
Dunleavy’s emphasis on oversight matters because it locates the value not only in what Codex generates, but in how the team’s work is reorganized around that generation. Codex is presented as leverage inside the delivery process: people specify and supervise more of the work, while the tool produces material the team can evaluate and refine.
The tool narrows the gap between senior intent and junior execution
Mike Krolnik says Codex unlocks value at “two ends of the spectrum.” On one end are senior architects working in complex environments. On the other are more junior team members who need to absorb, reuse, and act on senior-level thinking. Krolnik says Codex helps convert expert intent into more accessible material, while also helping junior staff produce outputs that look more senior and mature.
A dark-themed editor shows a markdown document titled “Endava Agentic Contracting Strategy v1.0,” with sections including an executive summary, problem statement and objectives, and scope. The artifact is not simply source code. It is strategic planning content structured as a working document, which aligns with Krolnik’s point that Codex is being used to shape the communication and packaging of technical or delivery thinking.
For senior architects, Krolnik says the important act is articulation: they can describe what they want, and Codex can make that intent more accessible to junior people on the team. Codex helps translate high-context architectural judgment into documents, diagrams, or other working material that colleagues can use.
For junior team members, Krolnik says adoption of the tool can help them create “senior and mature level” outputs. The difference he describes is not merely speed. It is a change in the structure, presentation, and completeness of work when Codex is part of the workflow.
Endava is treating Codex as a lifecycle agent, not only a coding assistant
Mike Krolnik explicitly resists the idea that Codex should be understood only as a coding tool. He says that as Codex has matured, Endava sees it as “more than just a coding tool” and instead as a “general kind of desktop agent” used across the whole lifecycle.
We really see it as more of a general kind of desktop agent to use across our whole life cycle rather than just specifically targeting the coding capability.
The working examples extend beyond writing code. One screen shows a code editor generating architecture diagrams from a markdown document, including references to an “ELIZA Protocol Implementation Layer,” “Core Request and Task Lifecycle,” and “Long-Running Tasks and Push Notifications.” Krolnik’s prompt is not to write a function or fix a bug. It is: “Hey Codex, draw a diagram of what’s in there so it’s easier to understand for our clients.”
That example puts the value in client-facing comprehension as much as internal engineering work. Codex is being used to turn existing technical material into a diagram that makes the system easier to explain. The work around software delivery — planning documents, architecture explanations, diagrams, and artifacts that help teams and clients understand what has been built or proposed — is part of the value Krolnik describes.
Krolnik’s “desktop agent” phrase also signals a broader working pattern. Codex appears inside the ordinary environment of knowledge work: documents, editors, plans, diagrams. The boundary he draws is the software delivery lifecycle as a whole, rather than a coding-only task queue.
The operating shift is supervision across the lifecycle
Dunleavy’s and Krolnik’s claims converge on a changed division of labor. Dunleavy emphasizes that small teams can deliver greater value in compressed time because Codex is producing work under human oversight. Krolnik emphasizes that senior architects can express intent and make it more accessible, while junior team members can use the tool to create more mature outputs.
The screens shown alongside their remarks reinforce that model. Codex is shown in the context of a strategy document and architecture-diagram generation, not only code completion. Endava’s stated use extends beyond coding into documents and explanatory artifacts that support delivery work.
The result is a workflow in which senior intent is easier to distribute, junior output can be raised in maturity, and human effort moves toward orchestrating and reviewing generated work. Endava’s working model for Codex is compressed delivery time, broader lifecycle use, and a shift from direct authorship toward supervision.