AI business consultancy
Each agent gets a named job, like answering enquiries, chasing paperwork or pulling the weekly numbers. It works in the systems your staff already use, checks its answer against your records and stops when a check fails, and your staff approve what goes out.
Every engagement starts with a consultation, and its fee is credited toward the solution we build for you. Or email [email protected].
What you're getting
Think of each agent as a new hire you give one job to. It uses your systems, hands back a draft or a finished step, and stops when something doesn't check out. It doesn't set your strategy, and it isn't a switch that runs the company for you.
"Answer booking enquiries from the reservation record" is a job you can check. "Run the front desk" isn't, so every agent we build is scoped to a job.
It opens the ledger, the inbox or the booking system and works from what is there, instead of answering from memory.
Totals are added up in code and matched to the source. If a check fails, it stops and hands the case to a person.
Someone on your team reviews its work the way they would review a new hire, and the decisions stay with them.
Why not just use ChatGPT or Grok
The agents from the big AI companies are good general tools, and you can give them instructions, but you can't change what they connect to, how they check their work or where your data lives. You get the same product as everyone else, and it changes when they decide to change it.
| Off-the-shelf agent ChatGPT, Grok and similar | Your AgentSmith | |
|---|---|---|
| Who it's built for | Everyone who signs up, in the same shape | Your business, built on your process, your documents and your rules |
| What it connects to | The apps the vendor chose to support | Your CRM, inbox, chat apps, accounting and files, whatever your team logs in to |
| How it answers | From the model's best guess, so ask twice and you may get two answers | From your real records, checked before anyone sees the answer |
| When you need it changed | You can adjust the instructions, but how it works inside is up to the vendor | Tell us what changed in your business and we change the agent |
| Where your data lives | On a shared platform | In its own isolated environment, never pooled with other clients |
| Who keeps it running | You do | We do, and we review it with you every month |
How we work
Every engagement starts with understanding how your business runs today. Only then do we decide what to build, so you pay for the problem that costs you the most and not for features you won't use.
We sit down with you and the staff who do the work, and go through how your business runs today, which systems you log in to, and where the hours and the customers get lost.
You get a written plan that names what we would build first, how long it takes and what it costs, with your consultation fee already taken off. If it doesn't make sense for your business, we stop there.
We build in stages on your real data, and your staff try each part on real cases before any customer sees it. Nothing goes live without your approval.
We host it, maintain it and review it with you every month, and add the next piece when the first one has proved itself. You can scale it back at any time and keep everything that was built.
Agents prepare, check and follow up. Your staff make the decisions that need a person, and any work that needs a licensed professional stays with your licensed staff.
What we build
These are examples, and most clients start with one or two of them. The consultation shows which ones would save your team the most time.
Messages from your website, WhatsApp, LINE, WeChat and email land in one place and get an answer within minutes, in the customer's own language, day or night.
Customers upload their documents once. The agent checks what is missing and follows up until the file is complete, so your staff stop chasing it piece by piece.
Procedures, prices and answers in one place that stays current, so new staff stop asking the same questions and everyone gives customers the same answer.
A weekly summary of enquiries, sales and open jobs, pulled from the systems you already use and sent to the owner without anyone building a spreadsheet.
Viewings, appointments and reminders get booked and confirmed, and the follow-up that nobody remembers to send goes out the same day.
Client records, billing checks, monitoring and dashboards that pull from the systems you already pay for, so the numbers in them match your books.
Why it doesn't make things up
A language model on its own is guessing. Ask it how much a client owes you and it gives you a confident number, and if you ask again you might get a different one. Our agents run inside a harness, which means the model decides what needs doing and real tools do the work.
01 Decides
Reads the request and picks what needs to be done and which tools to use.
02 Does the work
Pulls records from your systems, runs real calculations in code and reads the actual files.
03 Checks
Compares the result with your system of record before anything is sent.
04 Delivers or flags
A checked result goes out. If a check fails, the agent stops and flags it for a person.
A general chatbot, asked what a client owes
"About $48,200"
Plausible and confident, and possibly wrong, because it predicts what an answer should look like.
A Your AgentSmith agent, same question
$47,380.00, verified
It pulled the invoices from your accounting system, added them up in code and matched the total to the ledger before anyone saw it. Same question, same answer, every time.
Illustrative figures.
Proof, not promises
These are businesses I own and run. The agents work in them the way employees would, each with assigned jobs, and I decide what we make and what ships. The numbers come straight from each platform's dashboard, and they measure the business, not the agents.
Music channel
An orchestral music channel on YouTube. I write and judge the music. The agents prepare each release, write the titles and descriptions, schedule uploads, post across platforms and pull the analytics, and the views come from the music and its audience.
Source: YouTube Studio, 28 Sept 2025 to 27 Sept 2026.
Streaming service
Our own music streaming service, so the channel's audience can hear the whole catalogue on the web and on Android. I decide what it does, and the agents wrote the code and add each new release.
Video game
A mech action RPG that I design and direct, and it is still being built. The agents take implementation passes on the code, the art, the store copy and the patch notes. It is in Early Access on bravemech.com with a free demo on Steam, with 125,738 Steam store impressions and 12,800 store page visits so far.
Source: Steamworks store traffic (page visits, not sales).
Client, hospitality
A restaurant and event venue in Caledon East. Its agent sorts the inbox every morning, sends the owner a daily email rundown, prepares the Monday schedule and answers questions about the menu, hours and reservations from the inn's own notes. The owner decides what gets sent.
Who we work with
For a property developer we built an agent that reads architectural drawing sets and measures them into a bill of quantities, and we checked it line by line against a quantity surveyor's figures before trusting it. Most of our clients aren't technical, and we keep their names private unless they tell us otherwise.
Who's behind this
A YouTube channel with 5.38 million views in a year, a music streaming service and a game in Early Access, run by one person with agents doing the jobs employees would do. My first business was music, and when AI upended that industry I put agents to work instead of fighting them. That's how AgentSmith started.
How we charge
Every business runs differently, so a price list would only be a guess. We start with a consultation, and we put a price on the solution once we understand your business.
Step one
We go through how your business runs, which systems your team uses and where the hours go, and you leave knowing what is worth building first. You pay online and pick a time on the calendar straight after.
If you go ahead with us, the $300 is credited in full toward the solution we build for you, so it isn't added on top of the fee later.
Then a written proposal
After the consultation you get a proposal for your business, with a one-time fee for the build and a monthly fee for running and improving it. You see the full price before you commit, and if it doesn't make sense for you, we stop there.
Then we build and run it
We start with the problem that costs you the most and add the next piece once the first one has proved itself. You can scale it back at any time without losing what has been built.
Questions
No. If you go ahead with us, the full $300 is credited toward the solution we build for you, so it comes off what you pay for the solution instead of being added on top.
After the consultation we write a proposal for your business, with a one-time fee for the build and a monthly fee for running and improving it. The price depends on what we solve first, how many systems and channels it connects, and how many of your staff use it.
No. We handle the models, the servers, updates and the training cycles. Your part is telling us how your business works and approving what goes live.
Most of our clients aren't technical. You talk to us, we build it, and your team uses it the way they would message a colleague.
ChatGPT is a good general tool, but you set it up, you check its answers and you live with what it can't connect to. We build a team of agents that learns your business, connects to your systems and checks its own work, with people behind it who keep it running.
Where they already work. Telegram, WeChat, WhatsApp, Discord and the web are all supported, and a desktop app is available as an add-on.
We pick the right model for each kind of work and change it when a better one comes out, so you don't have to follow the AI news. Which providers are allowed to handle your work is agreed with you.
Each client's agents run in their own isolated environment with their own data, on a private network, and your documents are never pooled with another client's.
Yes. You can scale it back at any time, and nothing that has been built is deleted, so stepping back up later restores everything your agents have learned.
That's what the consultation is for. If what we'd build doesn't make sense for you, we tell you and we stop there.
The offer
We'll build the system that solves it, and we'll run it with you. It starts with a consultation, and its fee comes off the solution if you go ahead.