Moving Away from Agile: What’s Next

In this presentation, the McKinsey speakers focus on the organizational, people, and operating model transformations required to fully capture AI’s potential in software development—beyond simply layering AI coding tools onto existing Agile processes.

They argue that most enterprises see only modest gains (5–15% productivity uplift) because they treat AI as a point solution rather than rethinking how teams are structured, roles defined, work is allocated, and decisions are made.

McKinsey highlights how traditional Agile practices limit AI’s impact in software development. Enterprises must adopt AI-native operating models with smaller autonomous teams, continuous workflows, and evolved roles—shifting developers to AI orchestrators—for 5-6x productivity gains.

Key Points:

Agentic and adaptive structures

Flat networks of human-AI teams (“agentic organizations”) with real-time feedback loops, modular designs, and governance built around outcomes, risk, and learning rather than activities or traditional metrics like story points.

The Value Gap

While individual developers can achieve dramatic speed-ups (turning days of work into minutes with AI assistants and agents), enterprise-level impact remains limited. Traditional Agile practices—sprints, story points, heavy code review bottlenecks, stage gates, and fixed team configurations—create friction that prevents scaling AI benefits. Tech debt often increases as AI generates more code volume without corresponding oversight.

Why Operating Models Must Change

Core aspects of software development haven’t evolved much in 10+ years. AI changes the economics of coding, testing, documentation, and even planning, requiring a shift from human-centric, sequential workflows to AI-augmented, continuous, and parallel ones. The talk emphasizes the human and organizational side over pure technology.

Characteristics of Top Performers (5–6x Gains)

• • AI-native workflows: Move to continuous planning (vs. quarterly), spec-driven rather than story-driven development, and AI usage across the full software development lifecycle (SDLC), not just coding.
  • Smaller, empowered teams: “One-pizza pods” or 3–5 person full-stack “product builder” teams where individuals orchestrate AI agents instead of writing every line.
  • Evolving roles: Product managers prototype directly; developers become orchestrators and reviewers of AI output; new emphasis on AI coaching and upskilling.
  • New collaboration patterns: Separate flows for bug fixes vs. greenfield work, proactive AI-driven impact analysis, real-time feedback loops, and upfront enforcement of security/observability.

Enablers for Success

• • Hands-on training (e.g., “bring your own code” workshops).
  • Updated incentives and measurement systems focused on business outcomes (time-to-revenue, bug rates, developer satisfaction) rather than just tool adoption or activity metrics.
  • Change management to address varying AI proficiency across teams.

Evidence and Examples

Drawing from McKinsey’s survey of ~300 enterprises and client cases (e.g., an international bank achieving 51% more code merges and 60x AI agent usage), top performers are 6–7x more likely to redesign team structures and apply AI broadly across SDLC stages.

Overall Message:

The talk positions AI as a catalyst for a post-Agile paradigm—AI-native development—where smaller, more autonomous teams deliver faster with higher quality when supported by reimagined processes and roles.

It’s not about abandoning discipline but evolving it: developers shift from coders to conductors of AI agents, and organizations must invest in the “soft” side (culture, structure, skills) to turn individual productivity into enterprise transformation.