AI chatbot development for agencies

A chatbot on the agency's own site

An agency chatbot can explain services, show relevant work and help prospects book a call. It should only reference client work the agency has permission to show, avoid promising results and pass serious enquiries to a person. Many agencies find a well-designed contact form does the job; a chatbot is worth it if prospects arrive with many varied questions.

Client chatbot responsibilities

Accounts in the client's name

Set up AI provider and chatbot platform accounts in the client's name, with the agency granted access. This keeps billing transparent, lets the client keep the chatbot if the agency relationship ends and makes data protection responsibilities clearer.

Client approval

The client's business is responsible for what its chatbot says. Have the client approve the scope, the sources, the instructions and the handover design in writing, and review test results before launch. Record approvals so there is no ambiguity later.

Delivering a client chatbot

1. 01

   Scope with client

   Topics, handover, success measures.

   Output: Approved scope

2. 02

   Accounts

   Created in the client's name.

   Output: Access arranged

3. 03

   Sources and instructions

   Drafted, then client-approved.

   Output: Approved configuration

4. 04

   Testing

   Real client questions.

   Output: Shared test results

5. 05

   Launch and handover

   Review routine, documentation.

   Output: Handover pack

Example: an agency adding chatbots to its services

Agency chatbot checks

White-label partners

If a partner builds chatbots under the agency's name, agree confidentiality, who talks to the client, how feedback flows and who supports the chatbot after launch. The partner should follow the agency's standards for accounts and approvals. Keep the client relationship with the agency throughout.

The handover pack

Include the scope, sources and their owners, instructions, handover design, test results, account details and running cost estimates. If the client moves to another agency, the pack lets them continue without starting again. Update it whenever sources or instructions change.

Ongoing review

Chatbots need regular review. Agree whether the agency offers this as a service and what it includes, or whether the client's team takes it on with guidance. Put the arrangement in the contract so nothing falls between the two sides.

Data protection

Clients are usually the controller of their chatbot's personal data; the agency may be a processor. Agree data processing terms, and support the client's data protection impact assessment, which ICO guidance says AI use usually requires.

Setting client expectations

Clients may expect a chatbot to answer anything. Explain early that a reliable chatbot has a defined scope, depends on accurate sources the client provides and maintains, and needs regular review. Agreeing realistic success measures, such as questions answered from approved sources and smooth handovers, avoids disappointment and protects the agency's reputation.

Running costs for clients

Give clients an estimate of monthly running costs at their expected volume, covering AI usage, platform fees and review time, before they commit. Surprise bills damage relationships more than an honest estimate.