Endava Restructures Engineering Teams Around AI Agents, Compressing Project Cycles

Endava Shifts to Agent-Guided Delivery Model

Endava, a software contractor with engineering teams across Europe, the Americas, and Asia, is restructuring around AI-assisted workflows. According to the OpenAI Blog, the firm has adopted a model it calls an “agentic organization,” where OpenAI’s Codex agent encodes senior architectural knowledge and makes it available to guide junior and mid-level engineers across client engagements. The shift is producing measurable changes in project velocity and team composition—particularly in how expertise flows from experienced architects to less-seasoned developers.

How Senior Expertise Gets Codified Into Agents

Joe Dunleavy, Endava’s regional CTO for Europe, told OpenAI that the firm has “gone from producing a lot of the code ourselves to now overseeing the work that Codex can produce,” with output quality rising substantially. Mike Krolnik, Endava’s Global SVP of Agentic Architecture, describes a concrete workflow: senior architects articulate architectural constraints and design principles; Codex then embeds those constraints into guidance that junior developers receive in real time as they work. Krolnik explains that Codex acts as a persistent teaching layer—junior developers can ask questions about architectural decisions, and the agent contextualizes senior judgment without requiring synchronous pairing or code review cycles.

This flattens the traditional mentorship bottleneck. Instead of one senior engineer mentoring one or two junior engineers through review cycles, a single senior’s codified perspective can guide multiple less-seasoned teams in parallel.

Compression of Sequential Delivery Stages

According to Krolnik, Endava historically treated requirements analysis, design, and build as separate sequential phases, each consuming days or weeks. With Codex integrated across the entire engagement lifecycle—from intake through operations—those phases now collapse into a unified workflow. Krolnik reports that a recent cross-team engagement was compressed from weeks of work into days, though Endava did not disclose specific project scope or team size.

The consolidation works because Codex operates across the full delivery stack: requirements analysis, design, specifications, development, and operational deployment. This eliminates handoffs between specialists and reduces the context loss that traditionally occurs when one phase concludes and another begins.

Why This Matters

Endava’s restructuring signals a shift in how software contractors may approach staffing and project economics. If smaller teams can deliver client work at senior-quality levels by leveraging AI agents that encode expertise, the constraint moves from headcount to the quality of architectural codification. For contractors, this could improve margin economics—fewer bodies on a project, higher output per engineer-month. For junior developers, it creates a different career pathway: skill development shifts from learning-by-doing through code review to learning-by-collaboration with an AI agent that explains decision rationales in real time.

The real test is reproducibility. Endava’s results depend on whether senior architects can successfully encode their judgment into agent prompts and constraints, and whether those constraints remain coherent as junior engineers apply them to novel client contexts. If the pattern scales across Endava’s portfolio, it will validate a model other contractors may adopt—and conversely, if it fails on larger or more ambiguous projects, it will reveal where agentic guidance breaks down.