Services · Automate
The work that repeats, done. Within the limits you set
Agents and integrations. The work that repeats every week, done by agents that only have the tools you allow.
Agents and integrations
What you get
Agents for the work that repeats
Reports, checks, follow-ups and admin, on a schedule you choose.
Connectors to your systems
The ones that exist today, or a new one built for the system you use.
Approval where it matters
Everything that leaves the company waits for a person.
A trail of every action
What an agent did, when, and with which tool.
Limits
Every action has a level
Each tool an agent can use is marked read, write or irreversible. You decide which tools an agent has, and which of them wait for a person.
The agent looks, and changes nothing. Reading a calendar, a document or a number.
The agent adds or changes something that can be undone. A note in the knowledge base, a message in a channel.
The agent sends, pays or deletes. It waits for a person every time, unless you decide otherwise.
Examples
Two agents, and what they may touch
Meeting preparation
before every standing meeting
3 allowed · 1 does not exist for this agent
Questions and handovers
every hour · new tickets
3 allowed · 1 waits for a person
Connectors
Wired to the systems you use
These are connected today. If your system is not on the list, we build the connector, with the same levels, the same approval and the same trail as the others.
- Slack
- Gmail
- Notion
- Google Drive
- SharePoint and OneDrive
- Moneybird
- Klaviyo
- The one you need
Step by step
How it runs
-
Pick the stream
We look for the work that repeats, takes time and follows rules. That is where an agent pays off first.
-
Set the toolbox
Per tool you decide: allowed, waits for a person, or does not exist for this agent.
-
Run it next to your people
The agent starts with approval on everything. As it earns trust, you loosen what is safe.
An agent works best on a knowledge base that is already filled. Implement does that first.
What is an agent?
An agent is a piece of recurring work done by AI, such as a weekly report or the preparation of a meeting. It has a toolbox with only the tools you allow, and a schedule. Every action it takes is recorded, and what could do harm waits for a person.
What can an agent not do?
Anything you did not give it a tool for. Each tool has a level: read, write or irreversible. You decide which tools an agent has and which of them wait for a person. A tool that is off does not exist for that agent, so it cannot be tried.
Which systems can agents work with?
Slack, Gmail, Notion, Google Drive, SharePoint and OneDrive, Moneybird and Klaviyo today. If the system you use is not on the list, we build the connector. It follows the same rules as the others, so it has levels, approval and a trail.
How do we keep control?
An agent starts with approval on everything that matters. You see what it did on a run page, and you can switch it off at any time. Every action leaves a trail, so you can answer later what happened, who approved it and why.
Services
The other services
Advise
AI consultancy
Where AI pays off in your company, which tools to use, and a plan to start.
Read more Every teamImplement
Setup and onboarding
We interview your people, fill the knowledge base and connect your AI tools.
Read more Product ownersBuild
Custom software
Rapid prototyping and safe, production-grade development with AI.
Read moreStart with a call
Thirty minutes about your company. After the call you receive a proposal with a fixed price.