Example AI agents

Meet the agents that do the job, not just help with it.

An AI agent owns a role. It works through the week on its own, makes the routine calls, and brings you only the decisions that are yours. Here are three we build, each shaped around the business it works for, then hosted and maintained by us.

Sofia

Social media manager

Sofia keeps your Instagram busy while you're busy with clients. She writes each batch of posts in your voice, turns them into on-brand graphics and carousels, and publishes only what you've approved, in the time slots you chose.

What she does all week

  • Writes a fresh batch of posts in your voice whenever your approved queue runs low
  • Renders graphics with your logo, colours and exact wording, so the text on every image is right
  • Uses your own photos, or carefully vetted ones, never generic stock
  • Learns from every approval, edit and rejection

What you do

  • Review each batch on your phone: approve, edit, ask for changes, or reject
  • Send her photos whenever you have something you'd like posted
  • Nothing posts without your approval
  • Every draft checked against your industry's advertising rules
  • A weekly cap on how much she produces

From Sofia · to you

10 new posts are ready for your review

Carousel · 5 slides

Scheduled for Friday, 11:00

"Spring is the easiest time to start. Here's how to tell you're ready…"

Illustrative example of Sofia's review email.

Marcus

Customer finder

Anyone can buy a list. Marcus finds businesses that actually need what you sell, and shows his work. He researches each one from its own website and public listing, works out which of your services fits, and hands you a brief with the evidence quoted and linked.

What he does all week

  • Searches for businesses that match your ideal customer, by industry and area
  • Reads each company's own website and public listing
  • Matches what they need to the services you offer, with a quote and a link behind every claim
  • Finds the right person to contact and, if you want outreach, drafts a personal intro

What you do

  • Tell him once who you sell to and what you offer
  • Approve each prospect before anything is sent
  • No evidence, no claim: anything he can't quote from the source is cut
  • Pitches only services you actually offer, and never guesses an email address
  • Unsubscribe, suppression lists and sending limits built in

Prospect brief · ready for review

Harbor Heating & Air

Residential HVAC · 2 locations

Best fit Service reminders & dispatch
“Call the office to book your seasonal tune-up — we'll get back to you within two business days.” harborheating.example/services
Contact Operations manager · listed on the About page
Illustrative example of a prospect brief. The business is fictional.

Nora

Win-back specialist

Your cheapest new revenue is the customer who already loved you and quietly stopped coming. Every week Nora finds who's due back, writes each one a short personal note in your voice, and sends it from your own email once you've approved it. Not a newsletter, not a blast.

What she does all week

  • Works from the customer export your booking or sales system already produces, so there's no integration project
  • Picks who's due using clear rules you set, with the reason written down for each person
  • Writes one personal note per customer, grounded in their real history with you
  • Checks your next export to show you who came back

What you do

  • Upload your customer list from a bookmarked link, about once a month
  • Approve each note on your phone before it goes out
  • One note per customer, at most one follow-up
  • Grounded or skipped: a note that mentions anything not in the record is thrown out
  • Opt-outs honoured permanently, with daily sending caps

Awaiting your approval · 1 of 12

To: Dana R.

Why now Usually rebooks every 8 weeks · last visit 14 weeks ago

Hi Dana — it's been a little while since your last appointment with us, and we wanted to say we'd love to see you again.

If you'd like your usual slot, just reply and we'll find a time that works.

Illustrative example of a win-back note. The customer is fictional.

Beyond these three

Other roles an agent can own

Sofia, Marcus and Nora are examples. Every agent we build is shaped around one business: its voice, its rules and its systems. If a role in your business is mostly chasing, sorting and following up, it is probably a good candidate.

  • Inbox managerAcknowledges every inquiry in seconds, sorts the rest, and drafts replies for your review.
  • ReviewsAsks happy customers for a review at the right moment and drafts replies to the ones that come in.
  • CollectionsFollows up on unpaid invoices politely and persistently, and escalates the stuck ones to you.
  • QuotesTurns incoming requests into draft quotes from your own price list, ready to check and send.

How you get one

From first call to an agent at work

The same path as everything we build. You tell us about the role, and we do the rest.

  1. 01

    Find it

    We trace how the work actually moves today and agree what the agent owns, what it escalates, and to whom. You get a written scope with a fixed build price.

  2. 02

    Build it

    We build it around your voice, your rules and the tools you already use. No new platform for your team to learn.

  3. 03

    Switch it on

    It runs alongside your current process, with you approving its work, until it has proven itself. Then it takes over.

  4. 04

    Keep it running

    We host it, monitor it, and tune it as your business changes. You hear about failures from us first.

Hosted, maintained, and billed simply

By default we host your agent in a dedicated cloud environment of its own, never shared with another client. If you'd rather own it, we deploy it into your own cloud account instead, and you can revoke our access at any time.

You pay a fixed price for the build, a monthly fee for hosting and maintenance, plus the AI model usage it runs on, billed monthly. No seats, no licences, nothing to install.

Questions

What people ask about agents

Are Sofia, Marcus and Nora off-the-shelf products?
No. They are examples of agents we build. Each agent is built for one business: your voice, your rules, your systems and your approval steps. The names are only for this page; yours can be called whatever you like, or nothing at all.
Will the agent act without my approval?
Only where you decide it should. Anything that reaches a customer or the public waits for your approval, at least until it has proven itself, and you approve from your phone. You can tighten or loosen that at any time, and every action it takes is logged.
What do you need from us to get started?
A 30-minute call, then a working session with whoever does the work today. After that it is mostly approvals: the agent brings you its work to check, and your edits and rejections are how it learns what you want.
Can it run in our own cloud account?
Yes. By default we host it in a cloud environment dedicated to your business, never shared with another client. If you'd rather own it, we deploy it into your own cloud account instead, and you can revoke our access at any time. We still monitor and maintain it either way.
What does it cost?
It depends on what it has to do and which systems it touches. After the first conversation and a working session with your team, you get a written scope with a fixed price for the build. After that you pay a monthly fee for hosting and maintenance, plus the AI model usage it runs on, billed monthly.

Which role would you hand off first?

Book 30 minutes and bring one job that eats your team's week. We'll tell you whether an agent can own it, and what it would take.