An agency has eleven client assistants. Each was built slightly differently, by whoever was available, for whatever the client asked for that month.
Nobody can say which ones are current. Two were configured by somebody who has left. One client's answers are eighteen months out of date and the client has not noticed yet.
The agency is earning a retainer from all eleven and is one complaint away from discovering that it is running a support operation it never designed.
Step 1: standardise the build before the sixth client
The move from a handful to a dozen is where improvisation stops working. A standard structure introduced early costs a fortnight and saves a year.
Step 2: use one topic skeleton for every client
The same categories in the same order, with client specific answers inside them. This means anybody on the team can open any client and know where things are.
Step 3: document each configuration in the same format
Integrations, escalation rules, owner at the client, review date and anything unusual. One page each, in one place, in the same shape.
Step 4: name an internal owner per client
With a deputy. Ownership by whoever is free produces the situation where nobody knows the account and everything is rediscovered each time.
Step 5: schedule reviews rather than reacting
A cadence per client, in the diary, with a checklist. Reactive maintenance means you only touch an account when something has already gone wrong.
Step 6: batch the same work across clients
Seasonal updates, opening hours changes and policy reviews affect many clients at once. Doing them together takes a fraction of the time of doing them individually.
Step 7: keep a shared library of reusable answers
Generic material that most clients need, adapted rather than written from scratch, which is the same efficiency as handover and training the client team.
Step 8: standardise the reporting
The same numbers in the same format for every client, generated the same way. Bespoke reporting per client is where agency margin disappears quietly.
Step 9: track what each client actually costs to serve
Hours by client, monthly. Agencies routinely discover that two clients consume the margin from six, and only a record shows it.
Step 10: define what the retainer covers
Per client, written, with a change request route for anything beyond, which is the model priced in pricing assistant builds for clients.
Step 11: keep a single change log across the estate
What changed, for which client, by whom and when. Without it, a client asking why an answer is different from last month produces an afternoon of archaeology, and a staff departure leaves nobody able to explain how any of it reached its current state.
Step 11: keep client access properly separated
Credentials, permissions and data. This is both a professional obligation and the thing that becomes urgent the moment somebody leaves.
Step 12: build an offboarding process
What happens when a client leaves, what they take with them and what you remove. Having this ready is what makes a departure amicable.
Step 13: standardise how clients ask for changes
One route, one format, one queue, rather than requests arriving as messages to whoever a client happens to know. Scattered change requests are how work gets done twice, missed entirely, or delivered without anybody recording that it was chargeable.
Step 13: onboard new clients from a checklist
Same steps, same order, same documents, which is the structured start described in onboarding a new bookkeeping client without chaos.
Step 14: treat it as a service line, not a series of projects
Priced, staffed, reviewed and reported as one operation, which is the model set out in how marketing agencies use chatbots for their clients. See features, start at try, and the sector view sits under agencies.