AiTrust · The platform for AI digital workers
AI digital workers in your processes. AI Governance with built-in policies. One platform. Your control.
AiTrust is an application developed by Polski Instytut AI Sp. z o.o., in which AI digital workers operate in your organization's processes. They analyze information, prepare documents and carry out successive stages of work, drawing on your organization's knowledge.
Polski Instytut AI stands behind AiTrust
The Institute holds certified management systems for information security, privacy, quality and artificial intelligence.
- ISO/IEC 27001
- ISO/IEC 27701
- ISO 9001
- ISO/IEC 42001
Processes for your sector
Choose a sector. See how AI digital workers carry out a process.
- Each department's own processes
- An AI digital worker at every stage
- Control and approvals throughout the run
We tailor the process to your organizationYou start from a template and end with your own process.
Result and record
From AI output to a documented decision.
- Scope
- Plan
- Data
- Analysis
- Draft
- Report
- Recommendation
- Sources
- Data control
- Response credibility
- Approves
- Corrects
- Stops
- Audit log
- Evidence package
- PDF export
The scope of the record depends on the features and configuration of the deployment. The evidence package gathers the work record from a chosen period for review and is not a certification or a declaration of compliance with the AI Act. The offer sets out the scope of features in your deployment. Response assessment supports people but does not determine whether the result is correct.
See the run in the applicationAI Governance and policies
An AI digital worker under control.
Policies, permissions and the record apply at every stage of AI work. Where a person decides, the work waits for their approval.
Built-in policies cover every stage
- Policies
- Permissions
- Data control
- Response assessment
- Approvals
- Record
- Goal
- Plan
- Data
- Analysis
- Draft
- Verification
- Decision
- Result
Resources shared by the AI digital workers
- Documents
- Knowledge
- Models
- Tools
What will AiTrust do with the data in a task from your sector?
The simulator is an interface concept that runs in the browser on fictitious data and sends nothing. The example uses fictitious Polish identifiers, such as the NIP tax number and the PESEL personal number. Press Send, and the control window shows the content before and after masking and the H.A.D.A. zone. Masking replaces detected identifiers with tokens and limits data exposure, but on its own it does not mean irreversible anonymization. The built-in patterns do not mask people's names, addresses or medical terms, so they stay in the content unchanged. After sending, you will see the AI send report and the response assessment.
Sector
Content to send
Sample content is waiting in the field below. Press Send to see the control window before sending.
Prepare an offer for annual maintenance of an equipment fleet for Delta Serwis Sp. z o.o., NIP 525-000-22-33. The client budget is PLN 480,000 net. Contact person Joanna Lis, tel. 600 000 101, j.lis@delta.example. Service starts in January 2027.
H.A.D.A. is a methodology of human oversight of AI in four steps. Zones A to D classify content before it is sent to the model, in order of increasing data sensitivity and impact. Zone A covers content with no sensitive patterns detected, and the user sends it unchanged. Zone B covers financial and business data, and the user approves sending the version with masked data. Zone C covers personal and contact data, such as an e-mail address, and sending waits for approval by an authorized person other than the requester. Zone D covers critical data, such as PESEL, NIP, account and phone numbers and medical data, and sending is held. This handling applies to the standard profile. When several rules match, the most restrictive zone applies. The organization sets the rules that assign zones and the handling in each zone. In zones C and D, this demonstration uses an organization policy that provides for restricting the content, that is, sending the version with masked data. The recorded example on the H.A.D.A. page, where zone D has the Redact decision, is configured the same way.
Interface concept · simulation in the browser, fictitious data, nothing is sent.
What happens to your data before AI performs the task?
See how data access, content control, policies, models, result assessment and human decisions work together in AiTrust.
An explanatory diagram, not a live feed from a production system.
- 01 · IntentUser and taskStage 1 of 7 · Intent
User and task
The instruction is created in the context of the project, the documents and the assigned process.
- Project
- Documents
- AI digital worker
- 02 · Input controlData and contentStage 2 of 7 · Input control
Data and content
The system checks data and content before passing them on, according to the rules of the process.
- Access and document control
- Protection against prompt injection
- Data sanitization
Before controlPrepare a reply to a customer complaint. Contact details from the report: e-mail jan.kowalski@example.invalid, phone +48 600 000 000.Version sent to AIPrepare a reply to a customer complaint. Contact details from the report: e-mail [EMAIL], phone [PHONE].An example on synthetic data. Detected identifiers are replaced with tags; the scope of detection depends on the configuration.
- 03 · AuthorizationWhat AI may doStage 3 of 7 · Authorization
What AI may do
The process rules define the permitted model and tools. Before sending, the user decides whether to send a restricted version. Operations of designated classes wait for the decision of an authorized person before execution. The authorized person approves or rejects the operation in the approvals inbox before it is executed.
- Policy and H.A.D.A. zone
- Access to models
- Tool permissions
- Send restricted version
- Edit and check again
- Submit for approval
Execute (scope permitted for the task)Approve operation (consent of an authorized person before execution)Stop (the operation is not executed)The H.A.D.A. zone and the organization's policy determine the next step. The policy can also limit the content by sending a masked version.
- 04 · AIThe model performs the permitted taskStage 4 of 7 · AI
The model performs the permitted task
Only the scope permitted for the given task and configuration goes to execution.
- Local model
- Approved external model
- Project tools
Local modelApproved external modelThe model receives the context after the specified data categories have been sanitized. The choice of model follows from the configuration.
- 05 · Result validationThe response is checked tooStage 5 of 7 · Result validation
The response is checked too
The result can be checked before it is used further in the process.
- Response check
- Verification against sources, if enabled
- Check of content returned by tools
- Response check before display
- Verification against sources, if enabled
Limit of the assessment: The assessment does not determine whether the result is correct in substance; a person assesses it.
- 06 · HumanThe decision belongs to a person where the process requires itStage 6 of 7 · Human
The decision belongs to a person where the process requires it
After result validation, a person accepts the result or sends it back for correction. The acceptance covers the prepared result, while consent to the operation is given earlier, at authorization.
- Accept result
- Send back for correction
- Accept result
- Send back for correction
Decisions in the run diagram. The acceptance covers the prepared result after its validation, before it is used further in the process. Consent to the operation itself is given earlier, at authorization.
- 07 · EvidenceThe run leaves a recordStage 7 of 7 · Evidence
The run leaves a record
The organization can keep information about the actor, the model used, the policy, the decision and the result, to the extent offered by the given option.
- Send report with the response
- Event log
- Record export
The scope and order of the mechanisms depend on the specific deployment and process; some controls apply conditionally, according to the configuration.
Data → AI → Human
Data and rules define the context. AI performs the permitted scope of work. People keep authority over decisions that require oversight.
Changing the model and vendor
You change the AI model. The process and control stay.
Each process and each stage can have its own AI model.
Cloud models
Open source and local
The deployment configuration decides where the application, the model and the data run, and with a local model the data does not leave the organization.
- A model assigned to the AI digital worker, department and stage
- Knowledge, permissions and rules stay in the platform
- A new model is verified before use
Assigning models to the roles of AI digital workers
Regardless of model and vendor
Rules don't end with a document. They apply while AI works.
Deployment, control and vendor
Ready to work in your organization
Your processes. Your models. Your infrastructure.
Processes tailored to your documents and roles. On-premise too.
Deployment optionsControl and policies are part of the platform, not an add-on.
Permissions, data protection, result assessment, approvals and evidence.
AI GovernanceWho is behind it
Technology developed by Polski Instytut AI.
AI control methodologies, deployments, publications and company registration details.
About the Institute (in Polish, opens in a new tab)AI Act and GDPR training for companies.
How to use AI and AI digital workers safely and in line with the law. Delivered by EnclavaAI, the platform of Polski Instytut AI.
See the trainingSuwerenne AI (Sovereign AI), knowledge about control over AI.
A platform of Polski Instytut AI about sovereignty and control over artificial intelligence.
Suwerenne AI (in Polish, opens in a new tab)How to buy
From choosing a process to work in your organization
One-time launch. Fixed subscription. No fees based on the number of users.
- BEFORE YOU START, RECOMMENDEDYou train your teamTraining on the AI Act and GDPR, then an audit and AI policies, funding available
- STEP 1You choose a processSector, department and tasks for the AI digital workers
- STEP 2You see it in actionStages, the work of the AI digital workers, results and control points
- STEP 3We tailor it to your companyData and documents, models, permissions, policies and tests
- STEP 4We launch it, you sign offVerification in your environment, then controlled operation
Demo
Let's see how to launch the chosen process in your company.
- AI digital workers
- Data
- Decisions
biuro@aiinstytut.pl+48 22 295 12 14