What does setup involve, and who needs to be there?
The manager of the function and, for a few hours, whoever owns access to the tools. The manager hires the role from the workspace, and the first-run wizard opens pre-filled with the company, site, and profile from onboarding, each field with a provenance chip. It asks for what is new to the function: the remit in three lists, the always-escalate list with receivers, and the connections. The access owner creates a dedicated account per tool with the narrowest permissions the remit needs, and the manager enters it once; it is encrypted and never shown again. The manager reads and corrects the pre-filled Brand Brain, approves the proposed targets and the Day-One plan, and the role starts.
No engineer is needed unless the role publishes to a site built as code, in which case one agrees the pull request convention. The first-day story in the library shows the afternoon.
How do I write the remit?
In three lists, from the role's remit template, with the people who do the work today, in about an hour. Owns: the task classes the role proposes with evidence. Asks: what always goes to a named person with a summary, money, complaints, anything about a person, anything you want to see. Never: what the role does not do at any level, on top of the platform's never-automated list. Every task the role could meet sits on exactly one list; if you cannot place one, the remit is not finished. Read it to the team in five minutes. It is revised from the send-back reasons at each monthly review.
The playbook on writing your first remit walks the hour; the Three-List Remit framework explains the lists.
How much of my time does it take each day?
Minutes. The morning brief takes five: what the role did, what is waiting, what it detected, spend against budget. Approvals take seconds each because every proposal carries its evidence: the match with its records, the reply with its citation, the shortlist with its passages. Send-backs take a sentence, because a reason is required and the reason is what tunes the role. The first two weeks take longer because every proposal is read closely; by the third week the send-backs are about tone and the approvals are routine. Reading the send-back reasons weekly, sorted, takes half an hour and is the most useful half hour of the week.
The myth piece on more autonomy meaning more value explains why approval minutes were never the cost.
What happens when I am away?
The role keeps working and proposing; nothing happens until a person decides. Proposals wait in the Approvals tab, escalations reach their named receivers, detections appear in the brief, and any class promoted under the Promotion Standard continues within its bounds, logged. Name a deputy approver in the routing before you go, briefed on the remit and the recurring reasons, so the function does not pause for two weeks; the workforce view's waiting count shows whether the deputy is keeping up. That is the safe failure mode, and it is the opposite of an agent acting on its own because nobody was watching.
The Q4 planning playbook describes deputies and holiday coverage in detail.
When are the first results honest?
At the second monthly review. The first cycle includes setup, close reading, and calibration, so its approval rate, cost per outcome, and send-backs are not representative and are labelled first-cycle. The first review at thirty days sets targets, tunes the remit, and decides whether one routine class has earned promotion; the second review, after a full cycle with a tuned remit and a corrected memory, is the comparison against the baseline you recorded before the pilot. The Two-Cycle Go/No-Go framework gives the five criteria to read then.
Record the baseline before you start, with the baseline time template, or there will be nothing to compare.
What if it is not working?
Read the send-back reasons, sorted. Facts point at the memory: correct the Brand Brain with provenance. Rules point at the remit: add the line. Edge cases point at the class bounds: narrow or split the class. Tone points at the voice guidance. Substantive send-backs that fit none of these usually mean judgement work was put on the Owns list, and the Three-List Remit moves it to Asks. If a promoted class starts collecting send-backs, step it back the same day. If after two cycles the outcomes and cost are clearly short with no named fix, stop: pause the role, export the trail and memory, revoke the credentials, and record why.
The myth piece on working perfectly from day one describes the calibration curve and why pilots end early.
What should I read before starting?
The glossary entry on what a Digital FTE is, for the model. The how-it-works piece for the role you are hiring, for the mechanics. The role's thirty-, sixty-, or ninety-day playbook, for the sequence. The remit template for the role and the escalation list template, for the lists. The go-live checklist, for the day before. And the persona piece for your seat, for what you own, ask, measure, and refuse. An hour of reading, and the afternoon of setup goes as the playbook says.
The library is organised as hub-and-spoke clusters; each of those pieces links to its neighbours.