Step 1: gather the people and last month's cases
Book an afternoon with the two or three people who do the function today and bring a sample of last month's real cases: a few dozen tickets, transactions, applications, or drafts, including the awkward ones. The remit is written from cases, not from a job description, and the people who handled them know which cases were routine and which needed judgement.
Open the first-run wizard for the Digital FTE so the remit is written straight into it; the wizard pre-fills what the platform already knows and leaves the three lists for you.
Step 2: write the owns list in outputs and tools
Write what the role owns as specific outputs inside specific tools: reconcile bank lines against invoices in the ledger and propose matches with evidence; answer billing-question and how-to tickets in the helpdesk from the knowledge base with a cited source; screen applications for the named role in the ATS against the scorecard. Avoid titles and verbs without objects; handle support is not a remit.
For each output, name the evidence it must carry, because the approval gate will show it: documents for a match, the article for a reply, the passage for a score.
Step 3: write the always-escalate list with a receiver per category
From the awkward cases, write the categories that reach a person every time regardless of confidence: for support, money, account access, complaints, legal language, distress; for recruiting, anything about a candidate's circumstances; for finance, anything above a threshold, write-offs, adjustments; for sales, replies, named accounts, pricing. Name who receives each category, because an escalation without a receiver is a queue nobody owns.
Add the confidence triggers: no source found, conflicting sources, unclear intent. Then mark the whole list as fixed; it will grow after incidents and never shrink with autonomy.
- Category → named receiver, for every line
- Confidence triggers: no source, conflict, unclear intent
- Summary standard: asked, found, why not resolved, next step
- Marked fixed
Step 4: write the never list and the tone rules
Write what the role never does beyond the platform's own never-automated list: never contacts a protected account, never issues a credit, never posts an entry, never comments on a named competitor, never replies to a complaint. Then write the tone rules and the prohibited claims for anything the role writes, which the Brand Brain and the fact gate will enforce: the register per channel, the words the brand avoids, the claims it never makes.
Keep each rule to a sentence the FTE can act on. A rule that needs a paragraph is two rules.
Step 5: test the three lists against the sample
Take the sample of last month's cases and, for each, ask which list it falls under: owned and answerable, escalated, or never. Every case that the room argues about is an ambiguity in the remit; resolve it now with a rule, or decide it genuinely needs a person and put it on the escalate list. A remit that sorts the whole sample cleanly is ready; one that sorts most of it is normal, and the rest will surface as send-backs.
Write the targets at the same time, three or four with counterweights, because the owns list has just told you what they are.
Step 6: let the send-backs finish it
Save the remit and let the Digital FTE start at Level 0. Over the first month, read the send-back reasons weekly; each repeated reason is a rule the afternoon missed, and adding it is a minute. By the first monthly review the remit is precise, the send-back rate has fallen, and the record it produced is the basis for any promotion decision.
Log every remit change with a reason. A successor manager reads the history of the remit to understand why it says what it says.
What does a good first remit look like?
A good first remit fits on a page: a short owns list in outputs and tools with the evidence each must carry, an escalate list with a receiver per category and the confidence triggers, a never list, a handful of tone and claim rules, and three or four targets with counterweights. It reads like instructions to a careful new colleague rather than a policy document, and the people who wrote it recognise last month's cases in it.
The remit templates in the ftes.ai library give a starting draft per role; the afternoon is spent making one yours.