Your AI employees act for your business.
You set the rules.
AI employees create real value only when they rest on governance: a clear mandate, a person in the loop for what matters, and a record of everything they do. Fless builds all three into every employee.
The business stays responsible
An AI employee is not a legal person. It cannot own rights or carry responsibility for you. Laws on electronic transactions in many countries treat automated systems as tools of the business that uses them. Courts have held companies responsible for what their own chatbots told customers.
So the question is not whether you are accountable. It is whether you can show that the employee acted within a clear mandate, that a person approved the risky steps, and that you kept a record.
This page is general information, not legal advice.
Good AI governance, built in
The practices lawyers and risk teams recommend for AI that acts on its own, and how Fless puts each one into practice.
Start with the mandate
Refunds employee · mandate
Active“Settle refunds under ₦50,000 on your own. Send anything bigger to Ada in Finance.”
Give every employee a written mandate
Decide what each employee may do, for how much, with whom and when, and write it down before it starts work.
On Fless
- An employee can only use the actions in its job, and only through the accounts you connect to it.
- Set each action to Off, Ask first or Automatic.
- Add limits: daily and monthly spend caps, allowed recipients and domains, daily volume and working hours.
- Cap a whole app for every employee in your workspace, so no one employee can go past your ceiling.
- Print each employee's mandate as a one-page document to file or share with a counterparty.
Require a person above your risk threshold
Routine work can run on its own. Anything above your threshold, or with unusual terms, waits for a named person.
On Fless
- Ask first actions wait in an approval queue with a preview of exactly what will happen.
- Route approvals by type and amount, for example money above ₦250,000 goes to your finance lead.
- Only owners, admins and the approvers you name can decide. Approvers can edit before approving, and requests expire after 24 hours by default.
- Approve from the dashboard. Owners can also approve by replying on WhatsApp.
Agree how mistakes are spotted and corrected
Decide in advance how people will know they are dealing with AI and how errors are put right.
On Fless
- Outbound calls say they come from an AI assistant. Emails, and new WhatsApp and website chats, carry an AI disclosure line by default.
- Undo supported actions from the activity log.
- Edits and rejections become lessons the employee remembers, so the same mistake is less likely next time.
Keep records and evidence
Keep a record of what was done, who approved it, and how the employee was tested before it went live.
On Fless
- Every action attempt is logged, including blocked, approved, rejected and failed ones, with who approved it and what it cost.
- Export the activity log as CSV for your auditors.
- Admin changes go to a tamper-evident audit log: each entry is chained to the one before, and you can check its integrity at any time.
- An employee cannot be published until it passes its test scenarios in the sandbox.
Make one person accountable
Name one executive who owns AI risk and reports on it, to the board if you have one.
On Fless
- The Trust dashboard shows runs, success rate, approvals and mistakes caught for each employee.
- Replay any run step by step to see what the employee did and why.
- Your Chief of Staff includes what needs attention in a daily brief, and owners get a monthly governance report.
- Owners can pause any employee at once.
The highest-risk actions are locked down
Money and contracts carry the most legal risk, so Fless treats them differently.
Money waits for a person
Transfers, payouts and refunds start as Ask first. For connected payment apps they can never be switched to Automatic. Spend caps and amount-based approvers sit on top.
Employees never sign for you
There is no action that signs a contract on your behalf. An employee can prepare a document and send it for signature, and sending always waits for approval.
Tested before it can be published
Each employee runs its test scenarios in a sandbox with no real accounts and no real side effects. It cannot be published until it passes.
Blocked actions are recorded too
The log records what an employee tried to do as well as what it did, so you can see where your limits stopped it.
Who does what
Fless provides the controls. You decide how they are set, and you stay accountable for the outcome.
Fless:Enforces the actions, limits and approval rules you set, on every action.
You:Decide each employee's mandate and who approves what.
Fless:Logs every action attempt and keeps admin changes in an audit log that cannot be edited or deleted.
You:Review the log, and export it for audits.
Fless:Tests employees in a sandbox before they can be published.
You:Check the test scenarios match your real work.
Fless:Keeps credentials encrypted and each workspace's data separate.
You:Connect only the accounts an employee needs.
Fless:Lets you pause an employee at once.
You:Name an accountable person, and pause when something looks wrong.
Bring your team with you
The fastest way to lose the value of AI is a team that fears it or never uses it.
Be open about why
Tell your team what you are adopting AI employees for, the outcome you expect, and what it means for their own work.
Start with the people who do the job
They know where the real pain is. On Fless you describe the job in plain words, so the person who does it today can write the brief.
Upskill as you go
Make your people the approvers and reviewers. Every edit they make teaches the employee, and the hours saved go to higher-value work.
Governance checklist
Go through this before any AI employee acts for your business, on Fless or anywhere else.
- 01Each AI employee has a written job, a list of allowed actions and limits on value, recipients and hours.
- 02Actions that move money, publish, sign or delete are set to Ask first.
- 03Every approval type has a named approver, with a higher approver above a set amount.
- 04Customers can tell when they are dealing with AI, and know how to reach a person.
- 05Someone reviews the activity log every week and exports it for audits.
- 06The employee passed its test scenarios before launch, and is re-tested after its job changes.
- 07Your contracts with counterparties say how errors made by automated systems are handled.
- 08One named executive owns AI risk and reports on it.
Frequently asked questions
Put an AI employee to work, on your terms.
Describe the job, set the limits and choose who approves.
Start building