I kept doing the same things around each task: loading the right skill, carrying information between conversations, remembering what was waiting and deciding what could begin next.
The work also crossed between people. I needed to see how my tasks related to a teammate's and which results the next task depended on.
Connecting our ticketing service made that easier. I could put what was in my head onto a board, tie related tickets together and start separate tasks at the same time. We had a shared place to return to.
Attention became the constraint.
There were more ideas we wanted to pursue than conversations we could comfortably keep in view. We needed to hand over a larger job, see where it stood and step in when a decision needed us.
The best latency is no latency. You wake up, and the work is already done, or it has stopped at exactly the point that needs your judgment.
01 Put the questions in one place.
One change from our own build captures the idea. The request was to put waiting questions on one page. The result was a Questions page with answers grouped by ticket.
We could make our choices, add a note and send the answers for each ready ticket. Questions we had not answered stayed in view. We no longer had to open every session to answer across the work.
02 Give the next person a starting point.
Your best AI work is trapped in one chat history when the next person cannot use what happened there. The useful part includes the choices you made, the context you supplied and the instructions you refined.
A team should be able to pick up that work and improve its shared method. In Constraint, a role defines an AI worker's job. A skill holds the instructions for doing it. Versions let the team see what changed and start from the same approach.
03 Why Constraint
Good delegation needs boundaries: the result you want, the information that matters and the choices that need a person. Those constraints give the work direction.
We built Constraint so more of our ideas could become work the team carries together. We direct it. AI workers take on the pieces. The decisions that need us have a place to land.