Your AI team

Each piece of work has an AI team that takes care of it. The team is the unit that handles the work, and each member is an assigned AI. A team of one is still an AI team. The team name is a short phrase based on what you handed over, such as “Inventory checks” or “Weekly sales report”.

It works right away

The team starts at once with what UpServe can do today, without waiting for anything to be prepared. As results you accepted build up, it gets tuned to your work. This is called specialization. The team, its conversations, memory, files and connections stay the same.

On the work screen of a newly created AI team, open Manage work (⋯) → View setup to see the current stage. The item appears only for teams that have a setup to show.

Stage	Meaning
Learning your work	It already works right away. If you confirm results with “Looks good”, UpServe verifies a setup that fits this work based on them
Checking the custom setup	A setup tuned for this work is being verified. Until then the team keeps working with the current setup
Custom setup applied	From the next task on, the team works with the setup tuned for this work
Keeping current setup	The team keeps working with the current setup

UpServe doesn’t show progress percentages or target dates for these stages, so it never promises something it can’t back up.

When it needs your OK

When a verified new setup would widen what the team can do, a confirmation card appears under the header. It tells you what changes.

It will use a tool it doesn’t use now.
It will be able to send things outside. It will still ask you before sending, as it does now.
Results will be more thorough, so the SU per task may go up.

Apply switches to the new setup. Keep current means this work won’t be specialized again. While this is waiting, the chip in the header says Setup needs your check.

When specialization doesn’t happen

This kind of work may not support specialization yet. You can keep handing it over by chat as you do now.
Work that needs a ledger or an outside connection does better once a work screen is attached.
If the verified setup falls short of the bar, the current setup stays. Work in progress is unaffected.
Preparing a specialization has a limit on count and cost per user over a period. If you reach it, specialization may be put off until the limit frees up, and the work keeps going with the current setup in the meantime.

Specialization and the work screen are separate. Specialization goes on without a work screen, and you can attach a work screen before specialization is done.

Connecting a work screen

When a ready-made work fits your task, you see “a work screen that fits this work” and a Take a look button. In the Connect a work screen window, review what changes and press Confirm and connect.

The team’s conversations, memory, files, connections, tasks and skills stay in place.
New scheduled runs are prepared turned off. Scheduled runs you set up before stay paused when connected, and you can turn them back on from the work screen.
Connecting does not run anything by itself. Tasks you hand over afterwards use SU as usual.
If your existing data needs a mapping to the new work’s tables and none exists yet, nothing is connected. No data is moved and your current team stays as it is.

Only a small number of ready-made work types can be connected today. Check Browse work to see what is available.

Assigned AI settings

Choose the skills your team uses in Manage work (⋯) → Assigned AI settings. You can also open it from the card menu in My work.

Skills are chosen per team. Every assigned AI in that work shares the skills you pick.
Changes apply from the next task that starts. Tasks that already started finish as they are.

Section	What it holds
Built-in skills	Skills that come with the work. You can’t remove them
Skills you added	Use Add skill to pick from Official or My skills, then Add to this work. Use Remove to take one out

When a skill has a new version, an Update to new version button appears. If what the skill runs or the access it needs changed significantly, you are asked to review before it applies. A skill that needs a connection shows Connect or a field to enter a value, and a saved value is never shown again. Some work also shows extra settings that are specific to it.

Going back, and pinning the version

In the View setup window (shown only for teams that have a setup) you can:

Go back to the previous setup: results already made and anything already sent stay as they are. Tasks in progress finish with the setup they started with, and new tasks use the previous setup. What it can do and what it costs stay the same.
Pin version: when on, the team’s setup won’t adapt or go back on its own. The exception is that it goes back to the previous setup for safety if the current one can no longer be used. To add work to a specialized setup, unpin it first.
Setup history: when and how the setup changed. Labels are Auto-improvement, You went back, Went back automatically and Added work.

If you don’t like a result, don’t go back. Press Needs changes under the result and tell it what to fix.

How this relates to scheduled runs

A newly created AI team can’t take recurring scheduled runs until it is specialized. See Scheduled runs.

Advanced

You don’t need this section for everyday use. Read it only if you wonder when specialization starts.

Stages show only values the server decided. The server starts preparing specialization automatically for a team once enough accepted results have built up.
The cost of preparing a specialization is separate from the SU used when you run tasks. Running a task uses SU as usual.
A specialized setup changes only that team’s setup. Its name, conversations and memory stay as they are.
Settings addresses are /work/teams/team-id/settings or /work/work-id/settings.