Built on cybersecurity and data security
AI / SI agents
Plan business agents around specific tasks: finding information, organising content and connecting workflows. Assess data sources, tool permissions and human review, then validate the deployment approach through a limited pilot.
Business knowledge agents
Find answers in authorised documents and product information, separate access by staff role and provide sources for review.
Knowledge search and document organisationService information agents
Answer service questions using approved public information, organise enquiries and hand follow-up work to the team.
Customer enquiries and team collaborationWorkflow agents
Prepare summaries, drafts and tasks through defined steps, with permissions and human approval agreed before connecting systems or taking actions.
Workflow automation and system integrationSI means Super Intelligence, following the related US government terminology for AI technologies. Agent capabilities, connected tools and permitted actions are tested and agreed for each project.
SI terminology: White House, September 2026 ↗Development through prototypes, testing and acceptance
Turn requirements into a working prototype, then develop in stages against features and acceptance criteria, with business users involved throughout.
Requirements and working prototypes
Confirm users, workflows and data sources, try key screens, and define first-phase features and exclusions.
Stage deliverablesWorkflows, prototypes and acceptance criteriaDevelopment and scenario testing
Implement features and test roles, validation and error handling. Check interfaces and licensing for external services.
Stage deliverablesTestable releases, test records and issue logsLaunch and maintenance handover
Confirm deployment, backups, administrator accounts and operational checks, then provide usage instructions, version handover and support responsibilities.
Stage deliverablesApproved release, deployment and administration documents
Checks before handover
- Main workflows and permissions for different roles have been tested.
- Error notifications, backups and rollback are arranged.
- Source code, data, third-party costs and maintenance responsibilities are specified.
Features, timelines, source code and intellectual-property arrangements follow the project agreement. Third-party interfaces, platform approvals and AI output quality require separate verification.
Start with the purpose and authorised knowledge sources
Possible uses include staff knowledge search, assistants for public customer information, document preparation, draft replies and work summaries. Discuss whether these belong in a website, app or internal system. Configure functions around the project, using data the company has the right to use and identifying its owner, update date and intended audience. Use public, de-identified or approved material for testing.
- Staff searches across internal guidance, product information or work documents.
- Customer access to approved public service information, or draft replies reviewed by staff.
- Summaries of documents, meetings, tasks or reports.
- Data and third-party content permissions, with clear access limits for each role.
- The tool's retention, processing locations, supplier terms and handling of outdated content.
Limit access first; require approval for important actions
Begin with read-only knowledge search, then assess whether other tools need to be connected. Accounts and tools should receive only the resources and read/write permissions needed for their work. Sending email, deleting or changing data, making payments or changing permissions should require a person to check and approve the specific content, recipient, amount, file or access scope.
- Separate user identity, data permissions and tool actions.
- External documents, webpages and emails may contain malicious instructions; handle source content separately from operating instructions.
- Check outputs and action parameters independently rather than relying on prompts alone.
- Keep access revocable so the responsible person can stop a connection or function.
Use real questions to test answers and safety
Plan source references and human review in the prototype, testing common business questions. Alongside normal answers, test missing answers, outdated material, unauthorised content and malicious external instructions. AI can make mistakes; confident wording is not verification. Designated staff should still check sources and confirm important content.
- Define which sources answers should use and how to respond when information is missing.
- Test search permissions and data separation for different roles.
- Check references, summaries, tool parameters and approval flows.
- Pilot with limited data and use cases before choosing further functions.
Plan deployment, logs, updates and access removal
Specify deployment, suppliers, accounts, data retention and charges for the project. Query and action logs need a defined purpose, access controls and retention arrangements without unnecessary sensitive content. Retest after document or tool updates. Plan deactivation, permission removal, backups and any necessary recovery before use.
- Bring your assistant idea, intended users, data sources and examples of common questions.
- Explain whether it will be used in a website, app or internal system and which tools it needs.
- Clarify whether it will only retrieve or draft information, or perform approved actions.
- Assign responsibility for reviewing answers, approving important actions and updating information.
- Specify prototype, testing, deployment, later changes and support responsibilities in the proposal.
Functions, integrations, acceptance and ongoing support are subject to the agreed proposal.
Information and references
Suitability and enquiry preparation
When is this service suitable?
Common needs
- Finding company documents and FAQs takes time, prompting permission-aware knowledge search.
- You want an AI tool to help prepare drafts, documents or service enquiries.
- An AI assistant needs company data or tools, requiring clear read/write and approval scope.
Which service do I need?
- Workflow automation and system integration
Connect assistant tasks to approved company workflows.
- AI usage and business data security planning
Confirm permitted data, tool permissions and review responsibilities.
- Cybersecurity solutions
Coordinate authentication, data protection and system security.
Prepare for an enquiry
- Tasks, users and success criteria for the assistant.
- Describe knowledge sources and data types, or provide samples stripped of personal and confidential information.
- Planned platforms, tool licences, access responsibilities and actions requiring human approval.
For an initial enquiry, describe the issue and equipment only. Do not submit passwords or verification codes. The team will arrange a secure channel for sensitive documents and incident records.
EnquireAI work assistants and knowledge search toolsCybersecurity and AI useExplore security planning, responsibilities and acceptance considerations
Cybersecurity · Safe AI use
Software development and integration: security planning priorities
「AI work assistants and knowledge search tools」 includes the following security priorities; measures, tests and deliverables are specified in the proposal.
- Data and sign-inArrange authentication, data access and approvals around users and workflows, with clear purposes for company, development and test data.
- Integration and recordsCheck external interfaces, accounts and licensing, protect credentials, and plan logs for important actions, errors and changes.
- Testing and recoveryTest normal operation, invalid input and permission boundaries. Confirm release, backup, shutdown and rollback arrangements before deployment and maintenance handover.
AI considerationsAI queries must follow authorised data and user permissions. Restrict tool read/write access and retain human approval for important operations. Test malicious instructions in external content, data leakage and incorrect results.
Planning and acceptance recordsRequirements and acceptance checklist, software version, function and permission tests, deployment documents, and administration and maintenance responsibilities.
You may also want to know
Can we begin with a few non-sensitive documents?
Yes. Select approved, public or de-identified material, define users and common questions, and discuss a limited prototype and tests. Assess further data, permissions and deployment needs afterwards.
How are AI answers checked, and are they guaranteed to be correct?
Source references, test questions and human review can check whether answers are supported and how missing or outdated information is handled. AI can still make mistakes. Accuracy cannot be guaranteed for every answer, and important content requires confirmation by a responsible person.
Can the tool be restricted to reading without changing or sending information?
Read-only use can be specified as a requirement, with permissions configured for the selected platform and connections. Any later action features need designated approvers, checks of specific parameters, and tests for access removal, deactivation and necessary recovery.
