What is the visible cost?
The visible cost of an AI employee on ftes.ai is the plan: a block of work metered in tokens for the organization, shared by every Digital FTE, with usage reported per role. It is the number on the pricing page and the ceiling on what can be consumed in a cycle.
Total cost of ownership starts there and adds the human time the worker needs to be managed well. That time is real, it is yours, and it belongs in the comparison.
What is the approval cost?
Approval is the largest recurring human cost. Every proposal a Digital FTE makes is reviewed by a person until that class of action earns autonomy, and each review takes minutes. For an active role in its first months, that adds up to hours per week for the manager.
It falls over time. As the approval rate rises and routine classes are promoted on the autonomy ladder, the manager reviews fewer proposals and spends less time on each. The trajectory is visible in the approval record, which is the honest way to forecast it.
What is the setup cost?
Setup is front-loaded in the first month: connecting the tools the role needs with scoped credentials, writing the remit and escalation rules, confirming what the FTE has learned about the business in its Memory tab, and reviewing the first proposals closely. The first-run wizard pre-fills what the platform already knows, but the remit is a human decision.
Budget setup as human hours in month one. The implementation timeline guide on ftes.ai describes what each week involves, and the hours are mostly the manager's.
- Connections: scoped credentials for each tool, once per tool
- Remit: what the role owns, what it escalates, what it never does
- Memory confirmation: what the FTE knows, corrected where wrong
- Early review: close reading of the first weeks of proposals
What are the ongoing small costs?
Integration upkeep is small but real: when a tool changes its access model or a credential rotates, someone reconnects it. Governance reviews take a little time each cycle: reviewing the budget, deciding autonomy promotions from the approval record, and reading the monthly review the FTE writes.
Neither is large, but both are recurring, and a total cost that omits them is incomplete. An hour or two a month per role is a reasonable placeholder until you have your own record.
What do send-backs cost?
Send-backs are the honest measure of editing time. A proposal approved unchanged costs the minutes of reading it; a proposal sent back costs the reading, the reason, and the review of the revision. A high send-back rate in a category is a cost signal as well as a quality signal, and it usually points to a remit that needs tightening.
Track the send-back rate per category alongside usage per role, and the total cost of ownership becomes something you can state from the record: plan, plus approval time at the measured rate, plus the small recurring items, compared with the outcomes in Targets.
How does the total change over the first year?
Over the first year the total cost of ownership changes shape. Setup is front-loaded and then gone. Approval time is highest in the first two cycles, when every class of action is gated and the manager reads proposals closely, and falls as routine classes are promoted on the autonomy ladder and approvals are routed to team members close to the work. The plan cost follows usage, which typically settles once the remit is tuned and rises only when a second role is added, which draws from the same wallet.
The small recurring items, integration upkeep and governance reviews, stay roughly flat. Send-back time falls as the Brand Brain and remit are corrected, and a role whose send-back rate is not falling by the third cycle is telling the manager something specific. The honest way to see the shape is to log the human hours from the first week and read them alongside usage per role at each monthly review; the total is then a measured curve rather than an estimate, and its downward slope is one of the results the organization actually bought.