Trust
Reliable AI operations for your work
We define tasks and approval steps for your AI agents. Agents collect, analyze and draft. People decide, check and approve. We separate customer data and review work records and results.
Why this matters now
Checking external messages, data, permissions and results
Gartner has forecast that by 2027, 40% of companies will roll back AI agents they had adopted. It named failed permission design as the cause. The cases below illustrate why operating rules matter.
Air Canada
A chatbot gave refund information that differed from the real refund policy. A customer trusted it and bought a ticket. A Canadian tribunal said the chatbot is part of the company and held the airline liable. It found that the company could not treat the chatbot as a separate legal entity.
2024 · Canadian civil tribunal decision
Samsung Electronics
In 2023, employees pasted chip equipment source code and meeting recordings into ChatGPT, three times within 20 days. The company soon banned outside generative AI on all company devices and turned to building its own AI platform.
2023 · Samsung Electronics internal measure
Replit AI agent
A developer clearly ordered a "code freeze". A coding agent still deleted an entire production database that held records on about 1,200 executives. The permission to decide what to do and the permission to carry it out were not separated.
2025 · Publicly reported development case
Mata v. Avianca
A law firm filed a court brief that cited six court cases made up by AI. The cases did not exist, and nobody checked them. The judge imposed a fine and ordered the firm to send correction letters to all the courts involved.
2023 · U.S. federal court decision
Our five operating practices cover external messages, data, permissions and result checks.
People make the decisions, and agents carry out the work
We define roles for people and agents, supported by five operating practices.
People do this
Decide what to do. Check the result. Approve delivery.
Agents do this
Collect information. Analyze. Draft. Repeat routine tasks.
Five operating practices
Tasks handled by people
People handle the tasks below. Agents help with preparation, drafts and cross-checks.
Money
Payments · money transfers · bank account changes · tax filing
Legal commitments
Signing contracts · signatures · official submissions
People
Hiring · dismissal · HR decisions
Information
Handling personal data · sending or publishing anything to customers or outside parties
Documents, messages and figures sent to customers or outside parties are checked and sent by people.
Permission levels and approval
Every agent is given a level. Any action above that level needs approval from a person.
| Level | What it may do | Approval |
|---|---|---|
| 1 | Viewing · analysis · monitoring | Not needed |
| 2 | Writing internal drafts | Not needed (internal use only) |
| 3 | Internal actions within an approved scope | Within the scope approved in advance |
| 4 | Changes to live systems · sending outside · publishing | Approval by the responsible manager |
| 5 | Contracts · payments · legal matters · HR | Approval by the CEO |
For each agent, we define its team, task, accessible materials, permissions, approver and where results are recorded.
Work records you can trace
We record requests, sources, tools and results. Decisions, supporting information and next steps are documented. File changes can be checked, so you can trace the work later.
Results checked for reports and decisions
We check figures, assessments and forecasts before use in customer reports or decisions. Important results receive repeat checks and a review of opposing evidence.
Software applies the checking rules. Automated tests check them regularly.
Latest check results (re-run on August 11, 2026)
All judged correctly
Every representative test case got the right decision.
No rule broken
Each safety rule was tested against every combination of input conditions it can meet.
We use verified values consistent with earlier measurements as the basis for action.
Separate data and workspaces for each customer
We separate work scope, accessible materials and login details by customer and project. Each training participant has a separate practice environment.
How operating rules work
Separate work decisions from rule checks
People or agents choose an approach suited to the task. Software checks permissions and approval rules.
Decisions suited to the task
People or agents choose the work and resources to suit the task.
Permissions checked before action
Software consistently checks which tasks belong to people and which actions each agent may take.
Tests for each input combination
When we add a rule, we test every input combination it applies to. The results above come from these tests.
Decisions fit the task. Actions follow agreed rules.
Operating practices
Checks tailored to your project
We review assessment methods, permissions, setting changes and project operating practices.
- 01
We compare independent reviews.
We compare findings from independent reviews to check conclusions.
- 02
We separate access permissions and external sending.
Agents have separate permissions. People handle messages and documents sent outside the company.
- 03
We review changes to models and reference materials.
When models or reference materials change, the person responsible decides whether to recheck and carries out the review.
- 04
We agree on operating practices for your project.
We explain which practices are running internally and which will be configured for your project.
During adoption discussions, we explain the practices and supporting materials relevant to your project.
Materials to help you assess adoption
Request materials for your project. We provide them within the agreed scope, including result checks re-run on August 11, 2026.
- Each agent: tasks, accessible materials, permissions and approval responsibilities
- Written confirmation of tasks handled by people and how this applies to your project
- Result-checking criteria and review requirements before delivery
- Data use, storage locations and deletion schedule
- Work record storage and access
- Incident notification, shutdown and recovery procedures
What we agree on before we start
How far agents go and where people take over is decided separately for each type of work. Before we start, we fix the three points below with you.
- 01
Decisions and approvals handled by people
- 02
Materials agents may access
- 03
Who reviews results before delivery
We agree on decision responsibilities, data access and reviewers before work begins.
Find out how much of your work agents can take on
Tell us about your work and data. We will propose which tasks people handle and which tasks agents handle.
Contact us about adoption