Short answer: A chatbot service line works for an agency when the setup is templated, the client bots are managed from one place, and the price covers ongoing maintenance rather than just installation. The failure mode is not the technology. It is selling a one-off build and then absorbing months of unbilled tweaks.
Why do agencies add chat as a service line?
Three reasons, in ascending order of importance.
It is a visible deliverable. Clients can see it working on their own site within days, which is a pleasant change from technical work that takes a quarter to show results.
It generates evidence. Every conversation is a record of what that client's customers actually want to know. That is raw material for content briefs, landing page copy and ad angles you would otherwise have to guess at.
It is recurring. A build-once-charge-once service fills your calendar and empties it again. A managed service line compounds.
The catch is that the third benefit only materialises if you price and package for it. Otherwise you have sold a build and inherited a support obligation.
What should you standardise before your first client?
Build the template first, sell second. Standardise five things.
The intake. A fixed questionnaire that gets you the client's services, service area, hours, pricing policy, what they will and will not answer publicly, and where leads should go. If you improvise this per client, every build takes twice as long.
The content prerequisite. Decide your minimum bar for a client site before you will install anything. If their pricing page has no prices and their service page has no service area, the bot will be vague and you will get blamed. Make site content part of the scope or part of the qualification.
The guardrails. Standard rules about what the bot never answers: legal advice, medical specifics, custom quotes, anything with liability attached. Write it once, apply it everywhere, adjust per client only where needed.
The theme. A base configuration in Widget Studio that matches typical brand needs, so per-client theming is a colour and a logo rather than a design project.
The handoff and notification pattern. Where captured leads go and who gets told. Standardise on one route, usually email plus a webhook into whatever the client already uses, and only deviate when a client insists.
With those five fixed, a build becomes a repeatable few-hour job rather than a bespoke project. Setup itself is fast, roughly ten minutes to live, and there is done-for-you setup available if you would rather not run the first build yourself. How it works covers the mechanics.
How do you manage many client bots without losing track?
This is the operational question that decides whether the service line scales past five clients.
The Agency plan supports fifteen agents, which is what makes multi-client management practical: one account runs up to fifteen client bots, each with its own website, knowledge files, styling, and leads inbox, so client work stays cleanly separated even though it is managed from a single login. There are no per-user team seats or separate client logins, so if several people at your agency will work in the dashboard, plan around a shared account. See pricing for what the tier includes.
Beyond capacity, three habits keep it manageable.
A fixed review cadence. Once a month per client, half an hour: read the thumbs-down conversations, correct the wrong answers, note the content gaps. Put it in the calendar or it will not happen.
A shared correction log. When you learn that a certain phrasing produces a bad answer for one client, it is probably true for the others. Write it down once.
Client-visible reporting. The leads inbox pipeline, New through Contacted, Qualified and Won or Lost, is the report. Do not build a separate deck. Show them the enquiries and what happened to them, because that is the only thing they care about.
How should you price it?
Price the ongoing work, not the install. Three models that work.
Setup fee plus monthly retainer. The most common. Setup covers the build, intake and first month of tuning. The retainer covers monitoring, corrections, and a defined amount of content updating. This is the model most agencies should default to.
Bundled into an existing retainer. If you already run a client's site or ads, adding chat as an included capability is a straightforward way to raise the retainer and increase switching costs. The risk is that it becomes invisible and unvalued, so report on it explicitly.
Performance-adjacent. Charging partly against captured leads. Attractive to clients, harder to administer, and it exposes you to the client's failure to follow up. If you go here, make follow-up obligations explicit in the contract.
Whatever the model, do the arithmetic on your own costs first. Your underlying platform cost is flat monthly with no per-conversation billing, which is a genuine advantage when reselling: a client having a great month does not silently eat your margin. Your real variable cost is your team's time, so price against hours of maintenance, not against message volume. How much does an AI receptionist cost is useful context when a client asks you to justify the line item, and the honest answer is always to run their own numbers rather than quote an industry figure.
What should be in the scope document?
Be specific, because ambiguity here is where margin goes to die.
Included: initial build and training, a set number of answer corrections per month, monthly review, theme setup, notification routing, and a monthly leads summary.
Excluded, or billed separately: rewriting the client's website content, integrating with systems that need custom development, adding new service lines mid-contract, and unlimited on-demand tweaks.
Client responsibilities: keeping their website content current, following up on captured leads within an agreed time, and telling you when prices, hours or services change.
That last one matters more than it sounds. The bot trains on the client's site. If they change their pricing and do not tell you, the bot will be wrong and you will hear about it.
How do you sell it without overpromising?
Sell it as enquiry capture and question deflection, not as a replacement for staff. Show the client their own website, ask them what their top ten customer questions are, and point out that a visitor at 10pm currently gets nothing.
Be honest about the cases where it does not work. A client with almost no website traffic or no repeat enquiry pattern will not see much, and selling it anyway buys you a churn in four months. Do I need a chatbot for my business is a useful qualifying conversation to have before you quote.
Compliance is worth mentioning to larger clients: Clerkzo is SOC 2 Type II examined, GDPR compliant and HIPAA assessed, which removes a common procurement objection before it is raised.
Frequently asked questions
Can I put my own branding on the widget?
Widget Studio theming lets you match the client's brand, which is what most agency work actually requires, and the "Powered by Clerkzo" line can be switched off in the widget settings. Discuss any further brand-presentation specifics with the vendor before you promise a particular arrangement to a client.
How many clients can one person manage?
With standardised intake and a monthly review cadence, a competent operator handles a substantial book. The constraint is the review discipline, not the tooling. Budget roughly half an hour per client per month and staff accordingly.
What happens when a client cancels?
Agree in advance who owns the configuration and the captured lead data. Write it into the contract. Sorting this out during an offboarding is unpleasant for everyone.
Should I build my own instead?
Only if chat is your product. If it is a service line, the maintenance burden of a custom build will exceed the licence cost within the first year, and you will be doing platform engineering instead of client work.