Use AI across your company,
with your rules built in.
Shared controls for company context and agent actions.
Access rights, approvals and records managed together.
AI your people can use,
within the rules you set.
Extend AI across your teams
Let teams use company knowledge in their AI tools, with access checked against their existing rights.
Keep ownership of AI actions
Define which systems agents can use, what they can do and when a person must approve.
Trace AI decisions
Review the sources, rules and approvals behind agent actions, with records available for your risk team and auditor.
How AI stays
within the rules you set.
Connect with the rights you have
Systems connect read-only by default. People keep the permissions they have in your systems, with further restrictions available per team.
One way in
The platform checks access when people and agents request company knowledge through their AI tools.
Limits per agent
Each agent gets only the systems and actions its task needs. Approvals come at the moments you choose.
Actions and approvals logged
Every agent action and every approval is kept, with the rule and the sources behind it.
Evidence on request
Agent risk assessments, decisions and approvals are recorded for your risk team’s AI Act review.
- EExact OnlineInvoices and suppliersRead
- SSharePointContracts and proceduresRead
- OOutlookShared mailboxesRead
- AIChatGPT, Claude, Copilotpeople, with their own rightsAsk
- AG6 agentsin your processesWork
- OutlookRead
- SharePointRead
- Exact OnlineBook approved lines
- Above €2,500Ask first
- ICInvoice check booked a line09:20 · rule v13Booked
- CCCredit control sent a reminder09:32 · template v4Sent
- FOFinance owner approved a change09:41 · rule v14Approved
- Risk classSet per process
- Agent decisions1,240 recorded
- Approvals38
- Human oversightFinance owner
Rights and action limits
checked for each request.
- 1RightsRequests use the person’s rights in your systems or the agent’s assigned rights. Team restrictions can further limit access.
- 2LimitsEach agent works only within its task, in the systems that task needs.
- 3ApprovalAbove a limit, and for every change to how an agent works, the owner approves first.
- 4Audit trailAgent actions and approvals are kept with the rule and the sources behind them.
- 1RightsFinance, book lines
- 2Limit€1,840, below €2,500
- 3ApprovalNot needed below the limit
- 4RecordedRule v13, contract §4.2
- Asked byAgent, for finance operations
Every agent on record,
ready for your risk team and auditor.
- 1OwnerFinance owner
- 1SystemsExact Online, book lines
- 2Tested40 earlier cases, all passed
- 3Versionv3, approved 12 May
- 3Previousv2, can be restored
- 4This month2 exceptions, reviewed with the owner
- AI model providerEU, on record
- 1Who owns itFor each agent, the owner, the task and the systems it may use.
- 2How it was testedThe cases used to test the agent before go-live, with the results recorded.
- 3What changedEach version with who approved it and when, and the version before it.
- 4How it performsQuality and exceptions this month, reviewed with the owner.
The same controls
for every agent and every request.
Rights from your systems
Keep the permissions from your systems. Restrict access further by team.
Permissions per agent
Each agent gets only the systems its task needs. Writing back is switched on per agent.
Agent register
Every agent with its owner, its task and how far it may act on its own.
Continuous evals
Quality measured with evals before go-live and continuously in production.
Versions you can roll back
Every change has a version, who approved it and when, and can be undone.
Agent actions recorded
With the rule and the sources each action used, ready as evidence.
Hosted in the EU
On European infrastructure, with the AI providers per process on record.
Not used for training
Your data and questions do not train AI models.
Support for your reviews
against each framework.
AI Act
For companies whose AI use falls under the AI Act: an agent register, recorded risk assessments, human oversight and a record of agent decisions for your reviews.
GDPR
For every company handling personal data: a data processing agreement, EU hosting with the providers on record, set retention and input for your DPIA.
NIS2
For essential and important entities: supplier details, access control and a record of agent actions for your risk measures.
DORA
For financial entities: an overview of the AI services and providers in use, tested changes and versions you can roll back.
NEN 7510
For healthcare organisations: access by role, a record of access to sensitive information, and EU hosting.
ISO 27001
For companies that hold it or work towards it: access, change and supplier records that fit your controls.
ISO 42001
For companies with an AI management system: an agent register, risk assessments and human oversight to include in it.
Sector supervision
For supervised companies in healthcare, finance and other sectors: decisions with their sources and rules, ready for your supervisor.
What each team
can rely on.
The AI tools your teams use, with your company context and access that follows the rights in your systems.
Owner, tests, versions and every action per agent, ready when your auditor asks.
See where personal data is processed, which providers handle it and how long it is kept.
Every agent has an owner and set limits. Changes come to the owner first, with the evidence behind them.
What would help you use AI with confidence across your business?
Talk to us →
