Forecast people and tokens on one line
People cost is arithmetic. AI cost is behavior. Flowstate computes the first deterministically from allocations, salaries and working days, and builds the second from each person's own measured usage. Nothing in the forecast is a guess.
Same headcount, rising bill, no model that explains it
The trigger moment is always the same. You moved to an enterprise AI plan and the bill went up for the same headcount. The workforce model in your planning sheet never saw it coming, because tokens are labor too and the sheet only models salaries.
Every IC is a capital allocator, so the AI line moves weekly. A forecast that blends both kinds of labor is the only one that survives contact with the invoice.
Deterministic, down to the working day
Allocations times salary times working days, computed per geography with currency converted first. The model is held together by balance equations that must always hold, so a forecast that doesn't reconcile is a bug, not a footnote.
Built from each person's own measured usage
Each person's AI forecast comes from their own last 90 days of measured usage, multiplied by their allocation to the project. Not a team average, not an industry benchmark. Their behavior, priced.
Every projection carries a confidence grade, high, medium or low, based on how much of the allocated FTE is backed by real history.
The whole workforce, month by month
Every team, its people and its AI agents in one grid: FTEs, cost by month, and variance against the budget you actually committed to. The number your board asks about is the last column.
| Resource | FTE | May | Jun | Jul | Aug | FY26 | Vs budget |
|---|---|---|---|---|---|---|---|
| Payments | 38.5 | $712K | $748K | $791K | $823K | $9.2M | +$240K |
| People | 36.0 | $598K | $598K | $612K | $612K | $7.3M | On budget |
| AI agents & tools | 2.5 | $114K | $150K | $179K | $211K | $1.9M | +$240K |
| Platform | 29.0 | $531K | $540K | $538K | $546K | $6.4M | -$118K |
| People | 27.0 | $471K | $471K | $466K | $466K | $5.6M | -$92K |
| AI agents & tools | 2.0 | $60K | $69K | $72K | $80K | $840K | -$26K |
| Data & ML | 18.0 | $342K | $351K | $365K | $378K | $4.3M | +$36K |
| Total | 85.5 | $1.59M | $1.64M | $1.69M | $1.75M | $19.9M | +$158K |
No default rates. No default prices. No inventing.
There are deliberately no default token rates and no default prices anywhere in the model. If there are no measured inputs, there is no forecast for that line, and the product says so where a made-up number would otherwise sit.
That costs us impressive-looking charts in week one. It also means that when the forecast does show a number, you can take it to the CFO.
The chart and the alerts share one model
Forward projection applies a modeled month-over-month adoption growth rate, and the chart and the alert evaluator read the same one. An alert never fires on a number the chart never shows.
Forecast deviation
Actuals drift from the forecast line beyond your threshold. Caught while the month can still be steered.
Spend anomaly
A spike against rolling baselines. The new agent loop that burned a weekend of tokens shows up on Monday, not at close.
Budget exceeded
A cap is breached. The alert carries who, which project and how far over.
Every alert has a lifecycle, open, acknowledged, snoozed, resolved, and the state syncs across the app, email and Slack. Acknowledge it in Slack and it is acknowledged everywhere.
Forecast the whole workforce
Book a demo and see your people line and your AI line on one chart, with the confidence grades to back it.