Skip to content

Web agencies: offering an AI assistant to your clients and reselling it

Why an AI assistant fits naturally into a web agency offer, how to set one up for a client step by step, and how to set your own resale price with confidence.

Agencies and white label5 min read

Your clients already ask you to update their pages, fix a form or add a section. More and more often, they also ask for “a chat with AI”. Many agencies answer by installing a consumer tool, then realise they control neither the bill, nor the brand shown to visitors, nor the quality of the answers.

There is another way: turn the AI assistant into a service of your agency, one you configure, install and look after for your clients, at your own price. This article explains why the offer belongs with you, how to build it in practice, and what to settle before you sell it.

Why this is agency work

A useful AI assistant answers from the site’s own content: opening hours, prices, delivery terms, services, frequently asked questions. You already know that content. You often wrote or built the pages, and you know which information is current and which is stale.

Three things make the assistant a natural extension of your work:

  • You have technical access. Installing a widget means adding a script to the site’s pages. Your client may not know how; you can do it in minutes.
  • You have the relationship. The client trusts you with their site and usually prefers a single point of contact.
  • You already have a maintenance rhythm. An assistant needs follow-up: adding a page, removing an expired offer, adjusting an instruction. That fits into an existing maintenance contract.

What your client is really buying

Your client is not buying an AI model. They are buying better-informed visitors, fewer repetitive questions by phone or email, and a presence outside business hours. Pitch the offer in those terms.

They are also buying your reliability. An assistant that invents a discount or a delivery time damages trust. That is why the value of your offer lies as much in the setup and follow-up as in the tool: picking the right sources, writing clear instructions, testing before going live.

Setting up an assistant for a client with blabla.bot

blabla.bot is built for exactly this case: an agency or freelancer managing several client sites. Here are the steps, in the order you will take them in the console.

  1. Create the client and their site. The console organises your portfolio into clients, then sites, then agents. Each site has its own agent, knowledge and consumption.
  2. Add knowledge. Add a page address (the agent then explores the site from there), a sitemap, or PDF files. Before each indexing run, the console shows an estimate of its cost and asks you to confirm.
  3. Define the behaviour. Start from a template by sector and use case (customer support, pre-sales, appointments, FAQ) that fills in the role, tone, greeting and forbidden topics. Review, adjust, save.
  4. Pick the AI option. Four options are offered: two models, each in Simple or In-depth mode. The console shows an estimated cost per conversation for each one.
  5. Style the widget. Preset, the client’s main colour, logo, launcher position, suggested questions. Text contrast is checked: a colour that is too hard to read blocks publishing, and a correction is suggested.
  6. Test. The “Test” simulator sends your questions through the same answer path as the real widget. Test conversations are excluded from statistics.
  7. Install. Copy the embed code into the site’s pages. The widget only opens on the addresses you have allowed for that site.

You can also enter a contact email: when the agent cannot help, it offers that address to the visitor.

Setting your price

The principle is simple: you pay blabla.bot for actual consumption, and you bill your client however you see fit.

On the blabla.bot side, there is no subscription. You buy prepaid credits by card, and each answer is debited at the AI provider’s cost multiplied by two, input and output tokens included. With the catalogue’s current prices, a conversation costs a fraction of a cent.

On the client side, you are free: a monthly package, a package bundled with maintenance, a one-off setup fee followed by monthly follow-up. blabla.bot stays out of that relationship and sends nothing to your client.

With a fixed package, one precaution matters: a client whose site suddenly gets a lot of traffic should not eat your margin. You can set a monthly budget per site. Once it is reached, that site’s assistant alone pauses until the next month or until you raise the budget. An alert warns you beforehand, at 80% of the budget by default.

Questions to settle before you sell

Before offering this to a first client, clarify these points with them, in writing.

QuestionWhy it matters
Which sources may the agent use?The agent answers from what you give it. A forgotten page means an unanswered question.
Who approves the instructions and forbidden topics?The client knows their business and its limits better than you do.
What happens when the agent does not know?Provide a contact address to offer to the visitor.
Who updates the knowledge, and how often?An offer that changes on the site must change in the agent too.
What volume is included in the price?That is what lets you set the site’s monthly budget.

Be careful with promises. An assistant reduces repetitive questions; it does not replace a team. It can make mistakes, hence the importance of testing and follow-up. Channels other than the website widget (messaging apps, phone) are not offered today, so do not sell them.

Start small

The easiest start is one or two clients whose sites already contain clear information: a shop with opening hours and terms, a practice with its services, an online store with its returns policy. Measure real consumption for a month, adjust your price, then extend the offer to the rest of your portfolio.

To estimate cost and margin before you start, use the calculator on the Pricing page: it takes the number of sites, conversations per month and the AI option, and works out your margin if you enter a resale price.

Tags

  • agencies
  • resale
  • white label
  • AI assistant