Services · Intelligence
Not a chatbot bolted onto a website. An agent wired into the system your business runs on: it reads your records, answers in plain words, prepares the change you asked for, and waits for a person to confirm it.
- A company whose people spend hours finding, copying and updating records across one system.
- A team that wants AI on its own data without sending that data to a tool nobody controls.
- An owner who wants to ask the business a question in plain words and get an answer with the records behind it.
What we do
The work, in plain words.
- 01
Agents with your permissions
The agent sees what the person asking may see, and nothing more. Roles decide what it can read and change; who may see a price or a cost is settled on the server, before anything reaches a model.
- 02
Changes a person confirms
The agent finds the records and shows exactly what would change, from what to what. Nothing moves until someone confirms it, and every change is tagged in the audit log next to the person who approved it.
- 03
Wired into the real system
Tools that call your own records (orders, products, customers, shipments, tickets) instead of a copy pasted into a chat window.
- 04
Models chosen per job
We build on the major model providers and pick per task. What a model may receive is decided in code, not in a prompt.
- 05
Run after launch
Agents need looking after: new tools, new models, costs kept in check, answers checked. We run them the way we run the rest of the system.
How it starts
Three steps, each in writing.
- Step 1
A conversation
Which work eats the most hours, and which system holds the data.
- Step 2
One job first
We pick one task the agent can do end to end, scope it in writing, and build it into your system.
- Step 3
Then the next
Once people use it, we add tools and run it on the monthly retainer.
Proof
Running today, not a promise.
Cardi, the agent inside Samplyn
Built into the platform we run for a national apparel wholesaler: staff ask in plain words, Cardi reads and edits styles, orders and shipments, shows money only to admins, and changes nothing until a person confirms. Every change is tagged in the audit log.
samplyn.com/cardi ↗The apparel platform case
How the platform and its AI layer are built and run: the buyer portal, retailer intelligence, photos, and the agent.
Read the case →Our own company memory
Every AI session we run shares one memory of how our business works. The same approach is on offer for yours.
AI memory for companies →
What we do not do
Where we draw the line.
- We do not give an agent more access than the person using it.
- We do not let an agent change records without a person confirming.
- We do not claim an agent replaces your team. It takes the repetitive work off them.
Questions
What people ask first.
Is our data used to train a model?
We use model providers through their business APIs and send only what a task needs. Which provider and what it may receive is decided per system and written down for you.
Can the agent make mistakes?
Yes, which is why it prepares changes and a person confirms them. It shows exactly what would change before anything moves, and every change is on the record.
Does it work with the system we already have?
That is the point. We build tools that call your system's own records. If the system has no way in, we build one, or we take the system over first.
Can our customers talk to it?
An agent can serve staff, customers or both, each with their own permissions. We usually start with staff, where the hours are.
What does it cost?
The founder quotes after the first conversation, in writing. Model costs are part of what we watch and keep in check every month.
No prices on this site: the founder quotes in writing after the first conversation. How engagements work.
Tell me what you need.
Philip Radvan, founder. Every message lands with me, and it's me on the call.