Illustrative scenario with a fictional organisation. Not a customer reference, result or forecast.
A free zone told to adopt AI, and to answer for it.
A fictional robotics free zone wanted agents in procurement and tenant service. Its audit authority wanted continuous evidence of control. Both can be true.
One proposed incentive, five governed steps
Khor Sidr Robotics and Advanced Manufacturing Zone is fictional. A restricted agent prepares a synthetic incentive proposal. Teradius does not pay, transfer or hold money and gives no financial or tax advice.
A Khor Sidr agent proposes a fictional AED 48,000 tenant incentive. Rules hold one scope exception for review; it is reviewed and resolved. Accountable officer Samir Noor approves the exact amount, recipient and version, AED 48,000 to Tenant T-0412 (fictional), proposal version 1, under reference R-31406. Authority is re-checked at execution. The receipt carries the same reference, R-31406, with the decision record and board pack.
What would change, as designed An incentive payment would reach a tenant only after a named approver saw the exact amount and recipient.
Read the full scenario
At a glance
- Organisation
- Khor Sidr Robotics and Advanced Manufacturing Zone (fictional), a free zone authority in the United Arab Emirates for robotics and automation firms
- Size
- Mid-sized public body: a few hundred staff across the group, hosting several hundred licensed tenants
- Entities
- The authority (licensing and regulation), a developer company (land, buildings, utilities) and a services company (business set-up, visas, shared facilities)
- Customers and stakeholders
- Licensed tenants that build and integrate robots; a board; the government audit authority the zone answers to
- Governance
- A board; an officer newly accountable for the use of AI across the zone's services
- Teradius used in this scenario
- Core (approvals, receipts, audit trail): true now. Agents: true now, tested against recorded model responses and not yet run against a live model. CRM and Intelligence: true now. Strategy (the Board Room): true now. Approval screen: in development. Connectors to existing accounting systems: in development. Third-party agents and evidence of the AI run: in design. United Arab Emirates pack and special economic zones pack: in design.
The situation
Khor Sidr Robotics and Advanced Manufacturing Zone (fictional) licenses and hosts companies that build and integrate robots for logistics, manufacturing and healthcare. The zone is run as a small group: the authority itself, a developer company that owns the land and buildings, and a services company that handles business set-up, visas and shared facilities.
Its money moves in both directions. It bills licence fees, lease charges and service fees to several hundred tenants, with the refunds and credits that follow. It administers incentive payments to tenants under the industrial strategy it was created to deliver. It procures buildings, utilities and services. And it reports upward: to its board on the zone's directives, and to the audit authority it answers to, which in this scenario expects evidence of control rather than assurances of it.
Tax, payroll and e-invoicing obligations, for the authority and for its tenants, are met through the systems and providers the zone already uses.
The problem
In this scenario, the zone's government has asked every public entity to adopt AI across its services and to name one officer accountable for it. The finance team wanted agents to match supplier invoices; the services company wanted agents to triage tenant requests; procurement wanted approvals that did not wait for a chain of emails.
The question the newly accountable officer kept being asked: when an agent proposes an incentive payment to a tenant, or a change to a supplier's bank account, who approves it, and what exactly did they approve? The audit authority's version of the same question was broader: show us, continuously and without our having to take your word for it.
Three sets of books, approvals by email and a quarterly board pack assembled by hand made both questions hard to answer.
What they did with Teradius
In this scenario, the zone keeps its ledger, payroll, e-invoicing and payment providers and places Teradius in front of them as the approval and evidence layer. Connectors to existing accounting systems are in development; the scenario assumes them. Teradius does not pay, transfer or hold money; the connected system pays.
Money actions become catalogue actions. Fee invoices, incentive disbursements, procurement commitments and supplier-master changes are proposed, reviewed in plain language, and approved by a named person with a stronger sign-in step. Authority is re-checked at execution, so an approval given by someone whose authority has since been revoked is refused before anything is written. The approval is bound to the exact effect, and a change after review is refused before the connected system pays. (Core: true now.)
Agents propose matches and triage, and nothing more. Each agent has a restricted identity with only the roles it was given, proposes only catalogue actions, carries a budget, and can be stopped at any step. (Agents: true now; tested against recorded model responses and not yet run against a live model.) The approval screen is in development. Agents built elsewhere calling Teradius-governed actions through an open agent protocol, and the evidence of the AI run itself, are in design.
Tenant relationships in the same records. Tenant accounts, contacts and support cases sit under the same permissions and audit trail as the money, so a fee credit raised from a tenant's case and routed through Teradius takes the same approval path as any other money action. (CRM: true now.)
Three companies, one model. Each company keeps its own records under one set of permissions and one audit trail, with every change recorded with who made it and when, and every approval and receipt in one chain across the three. (Core: true now.)
Directives, key results and guardrails for the board. The zone's directives, occupancy, tenant growth and delivery of the industrial strategy, are set up with key results and guardrails, and their health is computed from published metric definitions. The quarterly pack is frozen from those definitions at issue, and a target met while a guardrail is breached is reported as unhealthy. Board-decision reasons are visible only to board-permission holders. (Intelligence and Strategy: true now.)
Jurisdiction rules as a pack, when designed. The United Arab Emirates pack and the special economic zones pack are in design. They are designed to carry the jurisdiction rules, figures and entity reporting for entities in the Emirates and in special economic zones, with scope to be confirmed with local advisers. Nothing in this scenario relies on them; the zone's tax, payroll and e-invoicing stay with its existing providers.
What would change
Qualitatively, and as the product is designed to work:
- An incentive payment would reach a tenant only after a named approver saw the exact amount and recipient, and a change after review would be refused before the connected system pays.
- The audit authority could be given decision records and evidence packs that it can check without relying on the zone or on Teradius.
- The board would read directives, decisions and the evidence behind them in one chain, in a pack nobody prepared.
- The agents would do the matching and the triage; the decision, and the proof, would stay with named people.
- Jurisdiction-specific rules would be layered on as a pack once designed with local advisers.
We were asked to adopt AI and to be accountable for it in the same sentence. Being able to show every proposal, every approval and every receipt is what makes the second half possible.
— Chief Financial Officer, Khor Sidr Robotics and Advanced Manufacturing Zone (fictional). An invented line for this fictional scenario; not a testimonial and not a statement by any real person.
True today and bounded
- Approval binding, execution re-checks, receipts and signed decision records are true now on synthetic data.
- The approval screen and accounting connectors are in development.
- The agent runtime has been tested against recorded model responses and has not yet run against a live model. Third-party agent calls, AI-run evidence and jurisdiction packs are in design; scope is to be confirmed with local advisers.
True now means built and working in our pre-release environment on synthetic data. In development means being built now. In design means designed, not yet built. Nothing on this site is generally available.
Teradius is pre-release, and early access is by invitation.
If your organisation is deciding how AI agents may be allowed near money, request a briefing.