What we have built

Five AI employees, and what came of them.

Each of these was built for one business, around the tools it already ran on and its own rules about what a person sees first. None of them is a product you can buy off a shelf.

One of them has a number we can publish. Four are running for clients and have nothing we are allowed to show you yet. Each is labelled, so you never have to guess which you are reading.

with a published result
1with a published result
running, no numbers we can show
4running, no numbers we can show
built from a template
0built from a template
  1. Revenue manager01

    Ten short-term rentals, priced every day

    Gina Holiday · Sydney

    A floor, a low-season, a shoulder and a peak rate for each of ten properties, plus weekend and event multipliers — New Year, a marathon on the route, a festival in the district. Then a ranked list of what to change first, with a dollar figure against every line.

    What it hands over

    Unit 7C · floor $210 · peak $340 · NYE ×3.0 · marathon ×1.3

    Result

    12% more revenue than the year before, across 10 properties.

    Where the human stays Every price change is theirs to approve, one by one.

  2. Revenue manager02

    A second portfolio, same diagnosis

    Running for a client

    Occupancy, nightly rates and booking channels read together to show which listings are priced wrong and by how much. The same method as above on a different portfolio — which is the part worth knowing, because it is the first time we ran it twice.

    What it hands over

    14 listings · 6 underpriced on weekends · 2 never discounted

    No published numbers yet

    Running for a client. Nothing we can publish yet.

    Where the human stays Same rule: nothing moves until someone approves it.

  3. Bookkeeper03

    A ledger that reconciles itself overnight

    Running for a client

    Cleans up the ledger, matches payments against bookings, categorises expenses, prepares owner reports and gets the paperwork ready for tax season. What it cannot match, it sets aside and asks about rather than guessing.

    What it hands over

    $1,240 in · matched to booking #4471 · Cleaning, March

    No published numbers yet

    Running for a client. Nothing we can publish yet.

    Where the human stays Their accountant still signs off on the tax return.

  4. Social media manager04

    Posts, replies, and the ones worth answering

    Running for clients

    Writes and schedules posts, adapts each one to the platform it is going to, watches what people say back, and hands over the replies worth the owner’s time. Running today on LinkedIn, Instagram, Facebook, X, TikTok and Xiaohongshu.

    What it hands over

    LinkedIn · Tue 09:00 · draft ready for review

    No published numbers yet

    Running for clients. Nothing we can publish yet.

    Where the human stays They decide per platform what goes out on its own and what waits for them.

  5. Daily briefing05

    Three things that need the owner today

    Running for a client

    Reads the day’s messages, meetings and loose ends across Feishu and DingTalk, and turns them into the short list of what actually needs the owner — so the thing that matters is not buried under four hundred messages that do not.

    What it hands over

    3 things need you today · 11 handled without you

    No published numbers yet

    Running for a client. Nothing we can publish yet.

    Where the human stays It suggests the list. They decide what stays on it.

The one you can actually see

What the social manager put out.

Four of these five produce work that lives inside someone’s accounts, which is why this page can describe them but not show them. Social is the exception — the posts are public, so here they are. Real output, published by the system, on real client accounts.

  • Account feed

    Account feed

  • Product post

    Product post

  • Wellness post

    Wellness post

  • Fashion feed

    Fashion feed

  • Jewellery post

    Jewellery post

  • Creator pair

    Creator pair

On LinkedIn it also answers people.

Publishing on a schedule is the easy half. The half that takes judgement is reading what comes back and deciding which of it deserves a reply from a human — these are two it handled without one.

  • An article it published on its own

    An article it published on its own

  • A reply it left on someone else’s post

    A reply it left on someone else’s post

The next one is built the same way.

We come and watch how the work is actually done, build the role with you rather than for you, and you see it running on your own data at about two weeks.