Engineering Beyond Agile: AI and the Rise of Agentile teams
AI maturity is forcing a structural rethink of engineering organizations
If you’re leading or building in tech, this is the moment to rethink everything you know about engineering organizations. While I don’t have a crystal ball, it’s time we have this conversation – openly, boldly, and with a sense of urgency. Over the past months, In my new role as Chief Dev Advisor, I’ve found myself in countless discussions about how Agentic DevOps, AI-assisted coding, and agentic experiences are fundamentally reshaping software engineering.
Here are a few examples I’ve seen firsthand, working with customers who are pushing boundaries:
With agents coding, the pressure mounts on reviewing massive PRs – the scale and speed are unlike anything we’ve seen before.
Specification-based software development makes it harder for teams to run multiple workstreams in parallel – alignment is everything.
Traditional ceremonies can become a burden. We’re moving faster, and both the information we share, and our collaboration models, are evolving. Microsoft’s Work Trend Index reports that 41% of leaders expect to redesign business processes from the ground up with AI in the next 5 years, moving past rigid schedules. Agile coaches have begun discussing “AI-augmented Scrum” where some ceremonies (backlog grooming, even code reviews) are partially handled by AI, freeing humans to focus on higher-level coordination.
Feedback loops are changing – Traditional Agile relies on human communication (stand-ups, planning meetings) to share status and feedback. Teams adopting GitHub Copilot or similar have noted that waiting a full day for stand-up or a two-week review cycle feels antiquated. Anecdotal evidence suggests some teams are dropping daily stand-ups in favor of AI-generated progress reports combined with ad-hoc syncs. The risk, however, is losing the human touch – e.g., engineers might feel out of the loop if AI handles too much communication.
A practical look at Agentic Software Engineering
An agentic powered flow could look like the graphic below. But let’s unpack what this looks like in practice for a traditional Agile team.
Imagine a dev team working on an MVP. After agents help the engineers and the product owner with brainstorming and designing a solution, this needs to be validated.
Agents grounded in user research, business requirements, and other relevant data help with validation.
The outcomes are used with an agent, or in this case Spec-Kit, to guide the team to create the initial specifications and constraints. Now the team starts to form three workstreams: backend, data, and frontend.
Each of these teams, as SMEs, want to define further the specifications and constraints for their workstream. Out of this we have adjusted specifications, and constraints. This needs to be reviewed before starting the work.
When reviewing, the workstreams notice the coherency is already deteriorating. The approaches are not well aligned, but still, it’s possible to make small adjustments. The documentation is very verbose, so this takes time.
Planning of the work items start, since all workstreams have branched out, and the agent is not aware, the agent can only start planning work for its own workstream, using the branch as context.
Code is being generated in each of the workstreams. Going through the task list at hyper speed. The assistant, being helpful, solves challenges at the go.
Progress, after the initial friction with all the specs and reviews, is suddenly super-fast, though when examining the first output of one of the workstreams, it’s obvious the agent made several choices, without updating the plan, docs, or specifications. Now we need to align these for all of the workstreams before we’re able to merge the branches.
The agent implemented the data model different across the three branches. Now the choice needs to be made about how to align this.
This summary comes from a real experiment with one of our customers. Like any experiment, we learned what works, what doesn’t, and where the products need to evolve. But most importantly, it highlighted how much our ways of working as humans need to adapt. This isn’t an isolated case – many teams are wrestling with how to rebalance as we become more agentic.
It’s clear: agents can be real enablers at every level of product engineering. However, it will require us to adjust our beliefs, structure and habits. Think of Agile, DevOps, Product Management, etc. These new best practices will form in the next 1-2 years.
The question is, are we ready to embrace this shift?
The Product Manager: Evolving at Lightning Speed
Product Management has evolved into a strategic powerhouse. Today’s PM isn’t just a backlog manager or feature shepherd, they’re also data analysts, coordinators, and researchers. This shift is driven by rapid AI adoption, digital transformation mandates, and increasing cross-functional complexity. PMs are now at the center of change, orchestrating not just features, but the transformation of their organizations.
In practice, a PM might refine an AI model’s behavior in the morning, rally diverse teams around a shared vision at lunch, and craft a narrative to influence executives by afternoon. The role is evolving faster than ever, demanding fluency in data, automation, and cross-functional leadership. (See the flow diagram below for how agentic-powered teams operate.)
Tomorrow’s PMs will be AI-native leaders – fluent in data, automation, and cross-functional influence. They’ll orchestrate not just features, but the very transformation of their organizations. The connective tissue between vision, technology, and customer value.
AI and automation have redefined the PM role: routine tasks are increasingly handled by AI, freeing PMs to focus on strategy, vision, and stakeholder alignment. With over 96% of product teams now using AI tools daily and research showing productivity gains of 40%, the expectation is clear—PMs are not just feature owners, but strategic orchestrators who bridge technology, business, and customer needs. Their influence now spans workflow optimization and cross-functional leadership, making them essential to modern product teams.
The Engineering Team: Time to Rethink Tradition
This might be a hot take, but if we want to remove friction and bias for fast experimentation, the ‘traditional’ agile engineering team is on borrowed time. It’s simply too complex to organize, maintain ceremonies, and integrate work at today’s speed. With agentic workflows and teams, and the relentless pace of technology, the old model is just that — old.
If AI tools continue to mature, we could see teams structured to be lean, agile, and multidisciplinary, with overhead reduced to a minimum. Why have a three-week sprint, while we steer for outcome-based software engineering and optimize for hyper velocity collaboration? As stated in recent Microsoft research: AI may free time for complex, creative problem-solving and human-facing coordination, but this shift is not automatic; it requires intentional job crafting (redesigning roles, rituals, recognition) so higher-order work is visible/rewarded; and support for horizontal skill expansion (product sense, data/AI literacy, operations).
Agentile = Agentic + Agile
We’re evolving from traditional Agile teams to lean, AI-powered groups that blend human creativity with agentic automation. I would like to introduce this concept as the Agentile Team – a lean, AI-powered, agentic evolution of Agile designed for hyper-velocity, outcome-driven collaboration. This team might consist of someone placed within a business team, building an agentic flow (Please, let’s get rid of “citizen developer” and call them builders), a PM, a junior, and a senior engineer — each empowered by AI and agents to do more, faster and better. Optimized for outcome, not checkboxes.
While these teams can tackle more tasks with agents that augment their work, supporting specifications, skills, and guardrails, they won’t always be able to be the SME on everything. That’s why SMEs will possibly start to become part of enablement teams – at the project or organizational level. Think of Team Topologies but even more optimized for hyper velocity and rapid iteration.
This model truly enables smaller, outcome-driven changes – rather than rigid, acceptance-driven work items. Changes are easier to control, reviews are creating less pressure, and ownership is a default. The toolsets, platforms and tests should be based on templates, specifications, constraints, and have proper guardrails in place. This does mean Platform Engineering principles are now, more broadly, becoming crucial for organizations to modernize their approach to AI, Apps, and Agents.
Hiring and roles – Is the junior Engineer being replaced by AI?
So, what does this mean for hiring and team structure? Junior engineers may actually soon become in hot demand as the backbone of feature teams. Mid-level engineers could support projects or feature teams, while Seniors – often the SMEs – should focus on enabling others: educating, maturing specs driven development, writing skills, driving transformation, and setting guardrails. As AI and software engineering roles boom, scaling domain and organizational knowledge will be essential.
Piecing it all together
With the rise of the next generation Product team and technological capabilities, we can now truly break the barriers across the whole software development process. AI Agents and agentic processes help with unlocking data that can be used to create or clarify initial specifications but also improve validation and measuring outcomes. A Product manager can rapidly prototype and validate ideas that can result in code, better specifications or fast learnings. Creating good tests will take minutes instead of hours. A prototype or mockup can be created during a meeting.
Let’s spark a conversation
How is your team adapting to the rise of AI and agentic workflows? What bold moves are you making – or resisting? Are you changing your hiring and talent strategy? Does the concept of Agentile teams resonate? Drop your thoughts, experiments, and even failures in the comments. Let’s learn from each other and shape the future of engineering together.
This is the first article in the series “Engineering Beyond Agile”, Read part two: