Missions, not tickets: rethinking how teams work in the age of AI
Agile gave us a way to coordinate slow, expensive human execution. Put capable AI in the loop and that assumption quietly breaks. Most of our rituals break with it.
A working note, written between shipping things.
Scrum is twenty-five this year, and the Agile Manifesto isn't far behind. For most of my career, some flavour of agile (sprints, standups, a groomed backlog, story points) has been the unquestioned default for how software and data teams organise themselves. And for good reason: it worked. But every operating model encodes an assumption about what's scarce, and agile's is starting to look shaky. It was built for a world where the slow, expensive, rate-limiting thing was humans doing the work. That world is ending.
I lead an enterprise agentic-AI program, and the pattern I keep seeing is this: once a capable agent is genuinely in the loop, the cost of producing a first version of almost anything collapses. A draft, an analysis, a slice of code, a campaign: minutes, not days. And the moment execution stops being the bottleneck, a process built entirely around scheduling and protecting execution starts to feel like ceremony for its own sake.
What agile was actually solving for
Strip away the rituals and agile is a coordination system for scarce human throughput. Standups exist because people can't see each other's progress and drift out of sync. Backlog grooming and estimation exist because we have to choose what to spend our limited hours on, and guess how many hours it'll take. Two-week sprints batch that expensive coordination into a rhythm humans can sustain. It's a genuinely elegant machine, built for a constraint that's now evaporating.
When the manifesto was written, that was exactly the right bet. It's worth re-reading with fresh eyes. Notice how much of it is about people, conversations and sustainable pace, all premised on humans being the ones who build:
Here's the part that should stop every delivery lead cold. Writing the ticket, sizing it, agreeing a definition of done, and slotting it into a sprint used to be a rounding error against the cost of building the thing. Now it's often the reverse. When an agent can take a one-line intent and hand back a working draft in the time it takes to run a refinement session, the ceremony of describing and queuing the work costs more than the work itself. You can spend an hour grooming something you could have just shipped.
The rituals that stop making sense
If execution is no longer the constraint, much of the daily apparatus inverts, not because the goals behind it were wrong, but because the mechanism assumed slow hands. Same intent, different machine:
The swaps I keep coming back to:
- Standups → ambient transparency. If every human and agent action is already legible in the system, you don't gather people to narrate yesterday. The status is the work.
- Backlog grooming → decision intelligence. When options are cheap to generate, the scarce act is choosing which to pursue. That deserves real tooling, not a weekly ritual of dragging cards.
- Story points → outcomes. Estimating human effort matters far less when the effort is near-zero. What matters is whether the outcome actually moved.
- Tickets → missions. A ticket is a unit of execution to be assigned. A mission is an outcome to be owned, with humans and agents collaborating to reach it however the work demands.
"Missions, not tickets"
That last swap is the centre of gravity, so let me make it concrete. A ticket says do this specific thing; it's a hand-off, where someone decides and someone else executes. In an AI-native team the decision is the expensive part and execution is cheap, so making execution the primary unit of work is backwards. A mission flips it: you state the intent and the outcome that matters, and a small pod of humans plus agents self-organises to get there, generating options, killing the weak ones, and escalating the genuinely hard calls to a human with the context and the taste to make them.
The team's job stops being "work the queue" and becomes "own the outcome and out-judge the alternatives."
This isn't "no process"
To be clear, I'm not advocating for anarchy, or for delivery leads having no role to play. The opposite, actually. When execution is cheap and judgment is the scarce thing, you need more discipline about direction, not less: clearer intent, a sharper definition of "good", and tighter feedback on whether the outcome landed.
You also need a lot more of that judgment, from a lot more people. The number of decisions a team has to make goes up, not down, and you can't funnel every one through a single lead and still move at pace. So decision-making decentralises. Everyone on the mission is making calls, continuously, to keep things moving and get to a result.
Which is why our fundamental grasp of business and technology becomes more important than before, not less. To decide well, and to guide an AI toward the right execution pathway, you have to genuinely understand the problem you're solving and the context around it. The scarce skill is no longer producing the work. It's staying laser-focused on the mission, holding the business problem clearly in mind, and steering the AI toward the outcome that actually matters. Hand that to someone without the judgment and they'll confidently ship the wrong thing, faster than ever.
So the roles evolve. The product manager, the delivery lead and the developer don't disappear; they grow out of their agile versions into something broader. Less grooming queues and assigning tasks, more owning outcomes. The people who deliver value this way need far greater autonomy and end-to-end ownership of the mission than a typical agile team grants them, with enough room to carry it from intent all the way to a delivered result and let AI do the heavy lifting in between.
The process doesn't disappear, then. It moves up a level, from coordinating typing to coordinating judgment. Agile optimised the cost of building. Whatever comes next has to optimise the quality of deciding.
- Agile is a coordination system for scarce human execution, the very assumption AI is dissolving.
- As execution gets cheap the rituals invert: transparency over standups, decision-intelligence over grooming, outcomes over points, missions over tickets.
- That means more discipline about direction and judgment, not less process for its own sake.
This is the thesis behind what I'm building: an AI-collaboration-first way for teams to work, where the AI co-worker runs on the same model as the humans and participates in the mission rather than just autocompleting tickets. If it resonates, or you think I've got it wrong, I'd genuinely like to hear from you.