Custom sales and marketing AI agents, at a fixed price.

One workflow, built into your CRM or ad stack, live in three weeks. Price and date agreed in writing first.

Book a scope call

For companies of roughly 20 to 300 people with a CRM or ad account already in use, and one repeated sales or marketing task that eats hours.

What we build.

  • Lead response and qualification

    Replies to inbound enquiries, asks the qualifying questions and books the meeting, then hands over to a person with the full context.

  • CRM follow-up

    Updates the CRM after each conversation and chases the follow-ups that would otherwise slip.

  • Outreach research

    Researches each account before the first message, so outreach starts informed.

  • Quote and proposal drafting

    Drafts quotes and proposals from your own pricing and past documents, for a person to approve.

  • Ad operations

    Tracks competitor ads and drafts the next campaign brief. For Shopify and D2C stores, Sella already does this as a ready-made tool.

Four things happen on every build.

One-page written scope
Agreed before any build work, with the handover cases written into it.
Weekly working reviews
You test a working agent in your own environment, starting in week one.
Built into your tools
Delivered inside the CRM or ad stack you already run, not a separate dashboard.
Thirty days of tuning
Included after go-live, against real usage, with the engineer who built it.

From scope call to live in three weeks.

  1. Within 48 hours

    Scope call.

    We walk through the workflow as it runs today.

  2. 1-2 business days later

    Written quote.

    A fixed price, a fixed date and a one-page scope.

  3. Days 1-7

    Week one.

    A working agent in your own environment, not a slide.

  4. Days 8-14

    Week two.

    Working review. You test it; we tune it.

  5. Days 15-21

    Week three.

    Live, inside the tools you already use.

  6. 30 days of tuning

    Thirty days after.

    Tuning against real usage, included.

How we work

How much does a custom AI agent cost?

A custom AI agent build starts at $12,000 and typically lands between $15,000 and $60,000, fixed in writing before any build work begins. The price depends mostly on how many systems the agent touches and how much of the workflow is edge cases. If the workflow cannot be scoped yet, a discovery engagement costs $2,500.

See what moves the price
Minimum build$12,000
Typical range$15,000-$60,000
Scoping discovery, if needed$2,500

Other workflows

Support, operations, internal knowledge, compliance and recruiting agents are taken on case by case. Describe the task on a scope call and we will say plainly whether it fits.

What we turn down

  • Workflows that cannot be specified. "Make our customers happier" is a goal, not a scope.
  • Fully autonomous agents making irreversible financial or medical decisions without a human checkpoint.
  • Full multi-tenant platforms.
  • Staff augmentation.

Questions about a custom build

Who can build a custom AI agent for my sales team, and how long does it take?
Aratavan builds one custom AI agent for one sales or marketing workflow, inside your existing CRM or ad tools. The scope call happens within 48 hours, the written quote follows in 1-2 business days, and the build goes live three weeks after the scope is agreed, followed by thirty days of tuning.
What happens if the scope changes halfway through?
We re-quote the difference rather than absorbing it silently or refusing to touch it. Most scope changes are small enough to fold into the existing timeline; the ones that are not get a short, written add-on quote before we proceed.
What if the fixed date slips?
We flag a slip the moment we see it coming, not the week it is due. If the delay comes from our estimate being wrong, we absorb the extra time at no charge. The price you agreed is the price.
Do you require a long-term contract?
No. Each build is scoped, quoted and delivered as its own project. Ongoing tuning after the thirty days is billed separately, only if you want it.
Do you use our data to train a model?
No. Agents run on hosted foundation models via API under standard data-processing terms; nothing we build touches or retrains a base model on your data. Data handling specifics are put in writing during the scope call.

Have one workflow in mind?

Describe it on a scope call. You get a fixed price and a delivery date in writing, with no obligation to proceed.

Book a scope call