A chatbot for agencies should help a prospective client understand what you do and whether a conversation with your team would be worthwhile. Agency websites often contain strong visuals and polished case studies, but practical questions remain: Do you work on projects of this size? Can you help with one part of a project? What happens after an enquiry? A well-prepared assistant can make those answers easier to find.
The aim is not to replace a discovery call with an automated sales pitch. It is to explain approved service information, ask useful clarifying questions and direct a visitor to the right next step. A public website assistant should not invent a proposal, promise delivery dates or claim that a team member has reviewed an opportunity.
Decide which enquiries the assistant should help with
List the services visitors most often misunderstand. A design agency may need to distinguish brand strategy from website delivery. A marketing agency may offer ongoing campaigns as well as one-off audits. A development studio may take on complete products but not small maintenance requests. The bot needs those distinctions in plain language.
Identify the audience for the first version. Existing clients asking about project access have different needs from new prospects comparing services. If the public bot is intended for new enquiries, say so in its role and provide the correct route for current clients. Do not upload private project communications merely to make the public assistant appear more informed.
Choose suggested questions that expose real decisions, such as “Can you help with a website redesign?” or “What does your discovery process involve?” A helpful starting question is more effective than asking every visitor to book a call before explaining anything.
Turn service pages into useful knowledge
For each service, explain the typical problem it addresses, the work included and the boundaries. “We deliver digital transformation” is too broad to answer many practical questions. A description of research, design, development and handover gives the assistant something concrete to work with.
Include prerequisites and client responsibilities. If the client normally supplies product copy or access to a system, state that. If discovery determines the final scope, explain why. The assistant can then help a prospect prepare for the process without pretending the scope has already been agreed.
Avoid repeating the same service under several slightly different names without explanation. Map common customer terms to the agency’s terminology. Someone who asks for a “new online shop” should be able to reach ecommerce development information even if the website calls it a commerce experience.
Use case studies as evidence, not promises
Case studies can show the kinds of work your agency has completed, but the source needs context. Name the problem, the work performed and any approved outcome. Distinguish the agency’s contribution from work completed by the client or another supplier.
Do not ask the bot to generalise a past result into a guaranteed future improvement. A successful campaign for one client does not establish that another client will receive the same conversion increase. If outcomes are illustrative, label them accordingly. If they are real, use only results the agency is authorised to publish.
Provide a short explanation of relevance. A prospect in a different industry may still benefit from a case study with a similar challenge. The assistant can explain that similarity without claiming direct sector experience that the agency does not have.
Handle budgets and timelines honestly
Many agency enquiries begin with “How much will it cost?” If you publish a starting range, include what that range applies to and what it excludes. If you do not publish pricing, explain the information needed to prepare an estimate. A vague answer that merely says “it depends” is less useful than a clear description of the scoping process.
A public assistant should not create a binding quote from a few chat messages. It can outline the factors that affect the work, such as project size, existing assets, integrations and review requirements. It should direct the prospect to the approved enquiry route for a specific proposal.
Treat timelines similarly. A typical process duration is not a promise of available capacity. The bot can explain published phases while making clear that the team confirms dates. This prevents a conversational estimate from becoming an expectation the delivery team never agreed to meet.
Ask a few questions that improve the enquiry
A useful enquiry might identify the service, the business goal and a rough timeframe. Avoid turning the assistant into a long qualification form before it provides any information. Answer the question the visitor asked, then request a missing detail only if it helps determine the next step.
Use questions that a prospect can reasonably answer. Someone exploring a redesign may know what is not working but not know their preferred technical architecture. Asking for business context is often more useful than demanding implementation choices too early.
Keep the assistant’s tone aligned with the agency. Confident and friendly does not mean relentlessly enthusiastic. The bot should be able to say that a request appears outside the published service scope and explain how to ask the team about it without inventing a capability to keep the conversation moving.
Build the agency assistant in Chaat
Use the guided creator to describe the agency and the enquiries you want to help with, or configure the bot manually. Define the role in plain language: explain services, use approved case studies, avoid promising prices or availability and guide interested prospects to the right contact route.
Add the relevant public URLs, service documents and approved FAQs. Review extracted knowledge for old portfolio pages, obsolete packages or claims that are no longer accurate. Each source remains visible and removable, so the collection can follow the agency’s current positioning.
Choose a concise greeting and reply style. Configure the name, photo, colours and launcher to fit the agency website. Design-conscious visitors may notice an inconsistent widget immediately, but readable text and clear controls matter more than decorative effects.
Keep client information out of public knowledge
An agency often holds detailed proposals, performance reports, credentials and internal discussions. Those materials may be useful for delivery work but unsuitable for a public assistant. Prepare a deliberately public knowledge pack rather than uploading a shared project folder wholesale.
If a prospect asks about a confidential project, the bot should only use information the agency has approved for publication. The absence of a case study is not permission to improvise one. Similarly, an existing client asking about their project status should be directed to the established client channel.
Review every sample quote or testimonial before adding it. Confirm that the wording is authorised and still accurate. The assistant can make public information easier to access, which also makes errors or unauthorised claims more visible.
Test the enquiry journey from beginning to end
Create a scenario with a realistic prospect. Ask about a service, add a constraint, compare it with another service and request a next step. Check whether the assistant explains the distinction and avoids jumping to an unsupported recommendation.
Include a request that is outside scope, a deadline the agency has not confirmed and a question about a private client outcome. These cases show whether the bot maintains its boundaries while remaining helpful. Review important replies with Chaat’s Test source option rather than judging only the tone.
If contact requests are enabled, send a clearly identified test request and inspect the staff experience. Decide who receives notifications and who reviews the dashboard. A lead-generation assistant is only useful if the agency can follow up on the information it collects.
Install where prospects need context
Place the chat on service and enquiry pages where it can answer practical questions. Check it alongside existing contact, booking and portfolio controls. On mobile, the launcher should not cover a call-to-action or compete with sticky navigation.
Keep direct ways to contact the agency available. The assistant should be an additional option for visitors who want a quick explanation, not a gatekeeper that blocks everyone from reaching the team. Test the shared link if staff plan to use it during early conversations.
Review the first few weeks for repeated gaps. Questions about a service you no longer offer may indicate old pages still being discovered. Questions about pricing may reveal that your enquiry page needs clearer expectations. Improve the website and knowledge together.
Maintain a useful commercial voice
Schedule reviews when services, team capacity or positioning change. Assign a content owner who can update the sources and retest the key enquiry scenarios. Do not leave the assistant describing an old package simply because the main website redesign is complete.
Measure whether prospects understand the agency’s fit and whether enquiries arrive with useful context. More messages are not inherently better, and fewer enquiries may not mean better qualification. Review the quality of the next step rather than treating conversation volume as the outcome.
The most effective agency chatbot sounds like a well-informed first conversation: clear about services, honest about limits and helpful about what happens next. It gives prospects enough confidence to speak with the team without manufacturing a proposal or promising work the agency has not agreed to deliver.