Think
Reads the goal and works out how to reach it. It looks at the ways the job could be done, drops the ones that will not work, and takes the one that will.
Bangkok·AI development agency
We build teams of AI agents. Dozens of them, organised like a real department, working inside the software your team already uses.
They think. They remember. They act.
And every week they work, they get better.
The shift
For twenty years, software was built like a machine. Now we build it like a team.
That is why we stopped calling ourselves a software house. A machine only does what someone wrote down in advance. A team is different: you give it a goal, it works out how to reach it, and it gets better at the job every week. On your screen it still looks like normal software. What sits behind it is not.
An AI agent is not a person, and not a normal tool. It sits in between. You give it a job, the memory to do that job, and permission to use your systems. Then it goes and finds what it needs, decides, and comes back with the work done. What we build is a room full of them. Each one is responsible for a different part of your business, working beside your people while the doors are open, and carrying on after everyone goes home.
Definition
Start with the word everyone is using and nobody explains.
AI agent
Software that can think, plan, and act on its own, to reach a goal you gave it.
Not a chat window waiting for your question. A worker with a brain, a memory, and hands.
The three parts
Three parts. Take one away and it stops being an agent.
Common confusions
Most things sold as AI are one of these two. Neither is an agent.
goal hold tonight's price, or push the set menu?
result hold the price, push the set menu tonight 1 click to approve
The short version
The simplest way to picture it: a new person on your team who never forgets.
Not a tool somebody has to operate. It takes the job, works out how to do it, writes down everything it learns, and comes back when it is finished.
It does not replace your people. It takes the 11pm work away from them: the reading, the reports, the catching up. A person still makes every decision that matters.
Where we come in
Anyone can build one agent. The hard part is making a whole team of them work together.
Building one agent is easy today. Building a team of them is not. Someone has to decide who does what, who checks whose work, what they all remember together, and when a human has to say yes. That is our job: we design the team the way you would design a company, then we build it. It works beside your people during the day and keeps going through the night. It never clocks out.
And this is the part that matters most. Every agent writes what it reads into one shared library: every review, every post, every price change, every order, all kept in the same shape. That library is why the team can answer questions nobody else can answer, and it grows every single day. It is the real asset, and it is yours.
one day on your team● never clocks out
08:12 social read last night's posts and comments, saved to the library
11:40 the manager answered a guest in Japanese, using what the library knows
14:05 your records today's orders and bookings, saved to the library
19:30 your rivals two price changes, three doors down, saved
02:04 reviews read 1,214 new reviews on your street, saved
06:00 the manager → one answer for you, ready for one click
It gets better
Most software is at its best on day one. This one is better every week.
Every day it writes down what it read, what it chose, and what happened next. Then it compares. What worked becomes the new habit. Nobody has to go back and change the instructions. It changes them itself, and you can always say no.
Where it lives
You never see the team working. You only see the answer, on the same screen as always.
Nobody wants one more app to check. So the team works underneath the screens your staff already open every morning: the bookings, the menu, the daily summary. The work happens out of sight, and the answer arrives in front of the person who has to decide. We build every part of it, and connect it to the systems you already have.
What we run
These are not examples. They are ours, and they run every day.
Four products, all built by us, all working right now. Underneath each one is the same team of AI agents. We simply point it at a different business. The one below is live. Ask it anything.
Ask it anything
Where do I get the best carbonara in Rome: the real one, the one people who grew up there keep going back to, not the tourist trap?
reTubeIt reads the creators' videos
the answer
Most lists send you to the tourist spots. The Romans on video keep going back to one small place in Testaccio. They film the guanciale, not the view. Same three words every time: “like nonna’s.” Go before 8pm, it sells out.
Example. Built from the market we keep, not your own records.
the live page
carbonara · every place, every answer Open ↗Ask your own.
100 answers, free.
Try it on your own market, then let the team keep learning it.
Ask a question. Get one answer, with proof.
It has watched 13,723 food videos so you don't have to. Ask where to eat, or how to cook something, and it gives you one answer, plus the exact second of the video it came from.
See how it works →A whole team, working inside a restaurant.
The owner asks what the restaurants around them are doing tonight. The guest asks where to eat. Both get an answer in seconds. Running in three countries.
See how it works →The same team, working inside beauty studios.
It watches what the studios nearby are doing, and writes a week of posts in the owner's own voice. No agency, no photographer, no writer.
Visit ↗The same team, working inside dental clinics.
It brings in new patients using real reviews and Google, and it stays inside the privacy rules every clinic has to follow.
Coming soon · being built nowWho
One Italian director, twenty years building everything but the code. Now AI builds that too.
Italian, and everything that comes with it: food, craft, and a habit of looking at a process and asking what could be simpler, and what could run by itself. Since 2007 that has been the job: simplify first, then automate the parts that can be automated. A consultancy across three continents, more than ten years of it in Australia, deep inside restaurants, marketing and the web. Small restaurants, restaurant groups, franchises. Always the same work underneath: the marketing, the websites, and the systems that keep them running.
Data engineering was the natural next step: get the numbers in order, and a business can finally see itself. AI is what made that scale. I could always design a system. I just could not write the software myself. That part changed. What you see here is more than twenty years of the same work, rebuilt for this new era: one person who understands every part of it, and an AI team that does the work. And when it does not know something, it says so. Saying nothing is better than a confident wrong answer.
AI should give people their days back, not take their jobs.
We do not replace your people. We take the weight off them: the reports, the searching, the catching up at 11pm. And we give back the evenings. Same pay, or more. Less stress. A person still makes every decision, keeps every relationship, and owns the company. Always.
Tell us what it should answer first.
Don't write us a long document. Just one question your business can never answer fast enough. That is where the first agent starts.
No newsletter. No noise. Just a reply, from me.