Your client’s details
We work on material you choose and redact yourself. You see what each tool sends out, where it is stored, and the setting that stops your data being used for training, from the suppliers’ own written policies.
Consulting
Find out what AI can and cannot do in a firm like yours.
AI agents are software that does a staff member’s routine work by itself: chasing overdue invoices, matching supplier quotes, answering rate enquiries. In two working sessions, you and the people who run your firm put agents to work on your own material, and decide what they are worth to you.
Twenty minutes, no deck, no charge. We go through what you have tried and what you have seen others claim, then tell you where you actually stand and whether this fits.
Prefer to write first? sales@8hats.ai.
What you actually need
The first skill is telling a real result from a good story
It applies twice. To what you read about other companies, and to what your own team reports back to you.
You cannot judge either from the outside. It takes having worked with agents yourself, on your own tasks, long enough to know what good looks like and what it actually costs to get. Until then every claim sounds equally plausible, and budgets get approved on tone of voice.
A real announcement — Klarna
“Our AI assistant handles two-thirds of customer service chats… doing the equivalent work of 700 full-time agents.”
A sentence like that tells you nothing by itself. It has at least three readings.
Reading 1
True and empty
The chats it took were the scripted ones. Cost to serve did not move.
Safe to ignore
Reading 2
Real work, wrong constraint
Genuinely working. Support was never what limited the business.
Their win. Not your move
Reading 3
Their economics moved
Cost to serve fell far enough to change how they price or staff.
The one to act on
Only the third should change your plan. The announcement reads the same in all three cases, and so does your own team’s progress report.
This is not a character flaw. Your people have automated real work already, and that is exactly the point. Automating a task is one problem. Running a company where people and agents share the work is a different one, and it is new enough that there is no playbook you failed to read.
What we can show you
See it running before you decide anything
Technology company, about 80 people, six months in. Subscriptions bought, pilots running, an in-house ML team hired. No effect the business could see. The problem was not the technology. The people at the top asked for adoption and had never worked with an agent themselves. We started with them. Afterwards one of them was able to judge his own AI work, and acted on it. Full case under NDA.
About a hundred agents run our own company. We show them live on the screen, not slides. Four areas, with numbers a calendar can check:
Marketing
1–2 hours
First draft of a page or a deck. It used to be days and five rounds. This page was rebuilt inside one day.
Paperwork
a morning check
Receipts, contracts, reports: agents read and draft. It used to be someone’s whole job.
Testing ideas
overnight
Market research, and ideas tested on AI stand-ins for your buyers, before you build.
Calls and deals
before you hang up
Minutes and tasks from every call, ready as it ends.
It is faster for the person who runs the business to learn what AI does than for a technical person to learn the business. That is why we start with the people who sign.
All 38 technologies we use are on a public map, each labelled by how strongly it is backed: deployed and measured, working prototype, specified, or still frontier. Most suppliers show you the first category and let you assume the rest.
These are our own numbers, not a promise of yours.
What does not move
The part that stays with you
Whatever you decide about AI, the responsibility stays inside your firm.
A supplier’s contract usually caps what he owes you at what you have paid him over the last year. Advice you give a client stays with the person who gave it. No contract moves that to a vendor, and no consultant can sign it for you. That is why the people who sign are the ones who should learn this, not only the people who install it.
It is also why waiting is not neutral. These tools are already on your staff’s phones. If nothing is written down about what may go into them, that is your firm’s current rule on AI.
Your people
Nothing here requires you to change a single job. The first work software takes is the retyping: the same figures keyed into three systems, five years of history copied out of a PDF. We do not advise on headcount. It is not what we sell.
The offer
Two sessions with the people who decide
One price for the firm, however many of you are in the room.
Session one · 90 minutes
We show what runs in our own firm, on the screen. Then we pick one job from your own week.
Two weeks of our support
Your team runs that job itself, on your own material. We stay reachable.
Session two · 90 minutes
What actually happened, reviewed. Your page of supplier questions, written and answered.
- The full program
The AI-Native Organization
USD 8,000per firm
Two working sessions of 90 minutes for your top team, online or in person, with real agent use in between. Step by step detail in the program PDF.
After the first session, if you decide it is not worth continuing, we stop and return half.
You leave with:
- a written page of questions any supplier has to answer before you sign, with our own answers to them
- one job from your own week that your team has run itself
- one number you already keep, written down before we start and read again on day thirty
- If you want a smaller first step
Strategy consultation
USD 1,000
Ninety minutes, one to one, on your own plan. You send us what you are working with beforehand. We go through what AI can and cannot take over in your firm, and what that is worth in hours and money. You leave with it written up, not only discussed.
Private. Nothing goes to your team unless you take it there yourself.
Not worth the money? Say so within seven days. Full refund, and you keep the write-up.
If the plan says build, the workflow your team picked goes into a pilot afterwards, set up for everyone who took part. We quote it then, on what the plan says.
Who would do it
The people you get
Some of us are in the United States, some in Singapore. For each engagement we put together the people whose experience fits your business, and two of us are in every session: no associates, no handoff.
Built CourseFactory and sold it to Smartcat. Twenty-six years building software. The engineer behind our setups.
Research · United States
Alexander Volkov, PhD
LinkedIn for Alexander Volkov, PhD (opens in a new tab)Designs the measurements behind every number we publish, and the way we tell a real result from a plausible one.
AI in production · Singapore
Mariam Mamedli, PhD
LinkedIn for Mariam Mamedli, PhD (opens in a new tab)Fourteen years in data intelligence for banks and fintechs. Head of Product, Asia at CEIC (ISI Markets), quants product line globally. PhD in Economics.
Built Examus and sold it to Constructor Tech. Runs 8Hats day to day, with agents doing most of the routine work.
We measure the business, not the activity. More drafts produced is not a result. It counts when it moves something you already count — and we name that number before the work starts.
Everything we run ourselves is listed on our public technology map, each tool labelled by how strongly it is backed.
Start here
Book a 20-minute call
Twenty minutes, no deck, no charge. We go through what you have tried and what you have seen others claim, then tell you where you stand and whether the program fits.
Prefer to write first? sales@8hats.ai.
The program PDF has the team, the case, how we tell a real result from a plausible one, what happens week by week, what you have at the end, and the price. Three appendices at the back carry the frame we use and a finished workflow, if you want to see one.