Scheduling and reading the Analyst
Analyst answers a governed library of revenue questions. How to choose which run on schedule, what each answer state means, and who sees which questions.
A governed library, not a query box
Analyst is a library of company-level revenue questions, each resolving to a prepared answer with its charts, reasoning and evidence. The set is defined rather than open-ended, and that is the point: a bounded set of questions asked the same way every time produces answers you can compare across quarters and defend when somebody disputes one.
The library is also server-owned. The questions, their wording, their categories and their availability all come from the platform, so a question your organization gains or loses appears or disappears without anybody shipping anything. The same list drives the Analyst page, global search, the assistant’s Ask Analyst shortcuts, the role editor and the Analyst Agent page, which is why they can never disagree about what exists.

Choosing what runs on schedule
Scheduling lives on the Analyst Agent page, under AI Agents. It lists every question the analyst runs for your company with its category, cadence, a Scheduled switch, when it was last checked, which result is current and its status. A switch in the page header, Scheduled analyses, turns scheduled runs on or off for the whole company. The Scheduled switch on each row decides whether that question runs.
Changing either switch needs the Edit AI Agents permission. Anyone else sees the switches but cannot move them, and the page says that only people who can manage AI agents can change scheduling. A question the schedule cannot include reads "Not schedulable" rather than offering a switch that would fail. Turning scheduled analyses off stops every scheduled run until it is turned back on, and existing results stay available.
That is the governance point. RevTech runs the analyst, and your team decides which questions it spends its runs on. Treat the list the way you treat SLA rules: a short, deliberate set that someone owns.

What cadence is telling you
Every question carries a cadence — weekly, monthly or quarterly — and the Analyst service owns it. You decide whether a question runs on schedule; you do not decide how often. The surface states the recurrence and never computes a next-run date from it.
So cadence is a bound on waiting rather than an appointment. A question with no result yet says it is generated weekly, which tells you the first answer is due inside a week. That is the honest version of the question everyone actually asks, which is "when does the next one arrive".
Some things are still not in the product. You cannot trigger a run, change a cadence, share a result or compare it with an earlier one. If your operating cadence needs an analysis to land before a specific meeting, that is a conversation with your RevTech team, not a setting.
- Cadence is displayed on the empty state, on a card with no result yet, and beside the result timestamps
- An unavailable question carries no cadence, and the surface says nothing rather than guessing
- Nothing in the product lets a user start a run — there is no button withheld by permission, there is no button
The states worth recognizing
Most of the judgement in using Analyst well is reading the state of an answer before reading the answer. The surface is deliberate about this: rather than hiding a run that failed or a result that is thin, it labels what you are looking at and keeps it usable.
- Needs setup — RevTech is missing data this question needs; the question opens on the steps that fix it, and any saved answer stays as it is until the next scheduled run
- Inputs ready · awaiting run — the data is now in place, and only the scheduled run is left
- Not yet generated or Preparing analysis — no result yet, or the first run is working; the cadence bounds the wait
- Update in progress — a newer run is working, and the current answer stays readable meanwhile
- Last result available or Generation failed — a run failed; where an earlier result exists it is kept and labelled
- Update blocked — a run could not go ahead on credits; the earlier result stays, and the next scheduled run tries again
- Insufficient data — the limited result is preserved and marked as limited rather than suppressed
- Unavailable in this version — the question exists in the catalog but has no answer behind it here
- Temporarily unavailable — a service failure says so; it never falls back to sample content
Reading what comes back
Every analysis opens with a quick answer: a concise summary that directly answers the question, with a short analysis under it. Full Analysis opens the structured write-up — executive summary, root cause, recommendations — and below the quick answer sit the KPI cards, charts and breakdowns the answer rests on.
Two details decide whether a number survives being repeated. "Source snapshot collected" is the moment the source data for that run was gathered; it is not a claim about how current your CRM is, and quoting it as one is the most common way to mislead a room with an accurate page. "Last checked …, no change" means a run looked again and found nothing new, so the result stands. And a link on an entity means the platform resolved it to a record you are authorized to open — a plain label is a label, never a broken link.
The habit worth building is opening the supporting data once for any number you are about to carry into a decision.

New answers on Home
You do not have to visit Analyst to know something changed. Home shows the questions you can open that have an answer, newest first, and marks each one you have not opened with a red dot. The dot stays until you open that question, from Home or from the Analyst page, and for now it is remembered per browser.
Who sees which questions
Analyst access is granted per role, not per person, and it is granted question by question. On the Admin page, in Roles & permissions, giving a role the Use Analyst permission opens the question picker, and the set chosen there is what that role can open — in the library, in global search, on Home and in the assistant’s shortcuts alike.
This is the answer to the support ticket that always arrives in week two: a rep who cannot find a question a manager can see is looking at a working permission model, not a bug. It is also worth reviewing on the same rhythm as the rest of your access, since a question that reveals company-level performance is a reporting decision as much as a data one.
What to take away
- You choose which questions run on schedule; the Analyst service decides how often. There is still no manual run.
- Scheduling is a governance decision: it needs Edit AI Agents, and one switch turns scheduled analyses off for the whole company.
- Read the state before the number. "Needs setup" tells you what data is missing and how to add it.
- "Source snapshot collected" is when the source data was gathered for that run. It is not a statement about how fresh your CRM is.
- A question you cannot see is a grant you do not have. Question access is assigned per role, in the role editor.
Put it to work
Jobs this page is behind
See how the work gets done.
Explore how agents prepare work across your revenue teams, and how people review it, guide it and decide what happens next. Fictional data. Real product.
Explore the demo