Skip to content

ROI

The ROI tab prices a workflow against the manual work it replaces, so you can check whether a workflow is worth building before you build it.

Click Set up ROI economics to open the baseline editor.

The ROI baseline editor

Field Captures
Hands-on time per run The actual doing, per run — ~45m
Meeting time Coordination meetings this retires, hours/week
Status reporting Writing and reading updates, hours/week
Duplicated work The same answer retyped, hours/week
SaaS this retires Tool, seats, and $/seat/month
Who did this by hand The people whose rates price the hours

The three coordination fields are easy to forget. A 45-minute report can also cost a 30-minute status meeting and three people re-reading it — include that.

Rates cascade, most specific first:

  1. The person’s own rate — see People
  2. Their group’s rate — see Groups
  3. The workspace default — see Workspace
  4. A built-in $90/hr

Rates are visible only inside the Console.

You supply the baseline. Alma measures:

  • Runs per week — actual executions over the trailing 7 days
  • Cost to run — the metered cost of those runs: model and tool calls
  • The work, by hand — the measured baseline this replaces, per week
  • Cost to run — model + tool calls
  • Net return — the difference, per week, with an annualized figure
Ratio (baseline ÷ cost) Verdict
3× or more Clears the feasibility gate — worth building
1.5×–3× Thin margin — worth building if time saved matters more than dollars
Under 1.5× Doesn’t clear the gate — don’t build this yet

A workflow that barely pays for itself in dollars can still be worth building if it removes an interruption or a task everyone hates. The ratio informs the decision; it doesn’t make it.

Hands-on time is drawn from the manual work this workflow was proposed from; meeting, reporting, and duplicate hours are measured from calendars and connected tools. Run volume comes from source history. Hours are priced at loaded rates from the directory — never shown outside the Console. Cost to run is metered from every run’s model and tool calls.