Manager
AI data team Manager: how it plans and delegates work
Your team lead and the front door.
Last updated Sunday, Aug 2, 2026
Every small team eventually needs someone to take "can you get me X" and turn it into an actual sequence of work: who does what, in what order, and how the pieces connect. OptimaFlo's Manager is that role. It's the one you talk to first, and the one that hands work to the other six.
What the Manager actually does
It runs as a loop: reason about the goal you gave it, decide whether the next step is calling a specialist or reporting back, act, and repeat. Each of the other six roles, Ingestion Engineer, Data Engineer, Analytics Engineer, Analyst, BI Developer, Quality Engineer, is available to it as a callable capability, so "connect my new source and build a monthly revenue dashboard from it" becomes: call the Ingestion Engineer, wait for the connection, call the Data Engineer to build the pipeline, call the BI Developer to build the dashboard, then report back with what happened at each step.
That loop is not unbounded. It runs under a hard cap on how many iterations a single turn can take, a token budget on what it's allowed to reason over, and a per-workspace token limit over a rolling 24-hour window, rechecked every 60 seconds or so rather than before every single call. The point of all three is the same: an agent that can call other agents needs a hard stop built in, not an assumption that it will politely stop on its own.
What you approve
The Manager pauses at the points where guessing wrong is expensive: when a request is ambiguous enough that two different plans are plausible, or when a step it's about to hand off to a specialist needs your confirmation first. It reports back after each specialist finishes its part, so you're seeing progress in stages, not a single opaque result at the end with no visibility into how it got there.
Honest limits
It sequences and delegates; it does not substitute for someone owning what the numbers mean. If "done" for a request depends on a business judgment, whether a metric definition is right, whether a dashboard is ready to show a customer, that call stays with a person. And it can only hand off to a specialist that's capable of the task: a source with no connector, a metric with no definition anywhere, those limits live with the specialist, not the Manager, and the Manager will surface them rather than paper over them.
Frequently asked questions
See the whole team in action on the AI data team overview or browse every role.
Now in early beta. One flat plan, no per-query tax. Runs in your cloud. Your data never leaves.