Raffles Consulting Services

Agentic AI tool permissions: build the operating register

Control agentic AI tool permissions through named identities, allowed actions, data limits, approval points, monitoring and tested revocation.

Agentic AI tool permissions should be narrower than the technical actions an agent could perform. An agent may combine data, software tools and multi-step autonomy, so one over-broad credential can turn a reasonable use case into an uncontrolled operating channel. The company needs a register that ties every permission to a purpose, owner, limit, human checkpoint and reliable stop mechanism.

Inventory every identity, tool and allowed action

List the agent, orchestration service, model, service accounts, application programming interfaces, plugins, databases, files and external systems. For each connection, record the named business owner, technical owner, authentication method, read or write scope, sensitive data, transaction limit and environments. Do not use a shared administrator identity because individual actions cannot then be attributed or removed cleanly.

IMDA’s Model AI Governance Framework for Agentic AI emphasises risk bounding, meaningful human accountability and technical controls. The PDPC guide to data protection practices for ICT systems covers access, monitoring, incident response and recovery. Apply contractual and legal duties to the actual systems and data.

Connect permissions to consequence and approval

Classify actions such as viewing, drafting, updating, sending, purchasing, deleting or changing access. Set amount, volume, time, customer and system limits where useful. Require human review before irreversible, high-value, legal, employment, financial or customer-impacting actions. A person needs enough context and time to make the approval meaningful.

Define prohibited combinations, such as creating a payee and releasing payment, changing a control and approving the same change, or extracting data and sending it externally. Use separate identities and approvals where duties conflict. Record why each exception is needed, who approved it and when it expires.

Test monitoring, expiry and revocation

Generate test actions and verify that logs identify the agent, tool, input, output, approval and downstream result. Confirm that alerts reach a person who can intervene. Rotate credentials, review unused access and expire temporary permissions. Vendor dashboards should not be the only record when the company needs independent evidence.

Run a revocation exercise that pauses the agent, disables credentials, cancels queued work and checks connected systems for incomplete actions. Restore service only from an approved state. Update the register after any model, tool, workflow, data or owner change rather than waiting for the next annual review.

Turn the decision into a controlled file

Use a dated decision record that identifies the applicant, business need, current official source, chosen route, responsible owner, costs, milestones and evidence. Separate facts already verified from forecasts and assumptions. A catalogue listing, supplier statement, consultant recommendation or portal acknowledgement does not establish approval. The company remains responsible for its declarations and commercial decision.

Give each quotation, contract, invoice, payment and deliverable a stable reference. Preserve earlier versions and explain material changes instead of silently replacing them. Where the authority, tax adviser, auditor or another professional must decide a point, record the open question and obtain the required view. Do not convert uncertainty into a confident eligibility statement.

Review before commitment and again before closeout

Run four reviews: suitability, authority, implementation and closeout. At suitability, test the entity, activity and business outcome. Before commitment, freeze the proposed scope, vendor and financial assumptions. During implementation, compare actual work with approved terms. At closeout, reconcile deliverables, accepted results, invoices, full payment, operating evidence and obligations that continue.

Name the person who may approve each stage and the records that person must see. Procurement should not treat a draft application as permission to contract. A project team should not treat supplier completion as company acceptance. Finance should not treat a submitted claim as cash received. Clear status words prevent accidental commitments and weak reporting.

Handle exceptions without rewriting history

Maintain an exception log for changed facts, delayed milestones, supplier variations, cost movements, missing records and failed tests. State the effect on scope, eligibility, timing and expected value. If an authority decision or formal variation is required, use the published route before relying on the change. Silence and helpdesk receipt are not approval.

If a material defect appears, pause the affected acceptance, filing or claim item. Keep original evidence, assign corrective action and set a review date. Do not backdate records, move costs between categories or adjust a baseline after seeing the result. A candid exception with an owned resolution is more credible than a clean-looking file that cannot explain what happened.

Plan handovers before the original project team disperses. The handover should identify the current version, source locations, system access owner, next review date, open exceptions and any condition that continues after payment or filing. Restrict personal, financial and commercially sensitive records to authorised users, while giving each responsible owner enough information to perform the assigned control. Record the external event that should trigger an earlier review, such as a changed programme page, revised authority letter, supplier failure or material business change.

Verify the final outcome independently

Ask someone who did not prepare the final pack to compare the decision, current official source, approved terms, deliverables, payments and operating evidence. Recalculate totals and sample source records. The reviewer should be able to reproduce the conclusion without relying on the preparer’s memory. Record unresolved points with named owners and dates.

Use the agentic AI use-case screen for the wider decision and keep this article as the narrower operating guide. After submission, preserve the receipt and reconcile questions, adjustments and the final registered or paid outcome. Continue any reporting, retention, asset-use or other condition that survives closeout.

Practical review checklist

  • Agent and service identities separated
  • Actions and data scopes recorded
  • High-consequence approvals defined
  • Conflicting permissions blocked
  • Revocation drill completed

Agentic AI tool permissions are controlled when every action can be traced to a justified identity and purpose, high-consequence work reaches a real human decision and access can be revoked without leaving hidden activity behind.