An agency looks at offering assistants to its clients. The instinct is to treat it as an installation, quote a setup fee and move on.
That is the smallest and least valuable part of it. The thing clients cannot do themselves is write the answers properly, keep them current as prices and services change, and read what visitors actually asked.
Positioned as a project it is worth a few hundred once. Positioned as a service it is a monthly fee across an entire client base, doing work the agency is already better at than anybody else.
Step 1: decide which clients this suits
Ones with enquiry volume, repetitive questions and something to lose from slow replies. A client receiving four enquiries a month does not need this, and one losing evening enquiries every week does.
Step 2: position it as a service, not an install
The setup is an afternoon. The value is the answers, the maintenance and the reporting, and the pricing should reflect that.
Step 3: start with one client properly
A single reference case with real numbers is worth more in selling this than any amount of explanation to the rest of the portfolio.
Step 4: build the answer set from the client's own material
Website, sales emails, support replies and the questions the client complains about answering, which is the brief described in briefing an assistant the way you brief an agency.
Step 5: agree the boundaries with the client
What must never be answered, what must escalate and where a wrong answer would cause real harm, agreed in writing before launch.
Step 6: decide the handover route
Who at the client picks up a conversation, how quickly, and what happens outside their working hours.
Step 7: price the maintenance explicitly
A monthly fee covering answer updates, new questions and reporting, which is the part that makes this a business rather than a project.
Step 8: set up the reporting from day one
Volume, questions asked, what could not be answered and what converted, which is what the client will judge it on.
Step 9: use the unanswered questions as an upsell
Each one is a page the client's website is missing, which is content work the agency is already selling.
Step 10: clarify the ownership position
Who owns the account, the answers and the data, which clients ask about and agencies are frequently vague about.
Step 11: standardise the onboarding
The same process every time, so the fifth client takes a fraction of the time of the first and the margin improves with volume.
Step 12: manage the portfolio centrally
Rather than separately per client. One place to see every account, update answers and pull reporting is what makes twenty clients viable, and per client administration is what makes five feel like too many.
Step 13: report monthly without being asked
The single thing that keeps a recurring fee, since a service nobody sees becomes a line item somebody questions.
Step 14: decide what happens when a client leaves
Whether the answers, the account and the history go with them. Agreeing it at the start costs one paragraph and prevents the most acrimonious conversation available in agency work.
Step 15: sell the outcome rather than the technology
An agency that can show a client the enquiries answered at midnight, the questions their website failed to answer and the leads captured is selling a result, not a widget. Set it up through install the widget and train your bot. The commercial questions sit in white label questions agencies ask first and pricing assistant builds for clients, and the build decision in chatbot builder vs ready made assistant. The sector view sits under agencies.