Procurement and spending
Is the agent still authorised to make this commitment within the applicable limit?
plusChain™ for Agentic AI
Patent pending
An autonomous agent may request an action in another system or organisation. Authority granted earlier is only the starting point: the receiving side needs to determine whether the agent is still authorised, and whether its current authority is sufficient for that particular request.
From granted authority to current authority
A principal grants an agent authority within defined scope, limits and duration. Before the agent acts, some authority may have been used or committed. It may also have expired, been revoked or been reduced. Changes to applicable policy can affect what the agent is permitted to do. The relevant question is what authority remains at the point of action.
Behavioural guardrails are not proof of authority
Guardrails help govern an agent’s behaviour, but they do not by themselves establish its current authority for a particular action. Following local policy does not, on its own, answer the receiving system’s question about valid and sufficient authority.
An independent check at the receiving side
The receiving system initiates a check using plusChain verification infrastructure and relevant evidence. It determines whether the requesting agent has valid and sufficient current authority for the proposed action, independently of assertions from the agent or the organisation operating it. A completed check can find authority invalid or insufficient.
plusChain supplies the verification infrastructure. The receiving organisation retains the decision about whether to accept the action.
Delegation remains bounded
A parent agent may delegate narrower authority to a sub-agent. That authority remains dependent on the parent’s current authority. If the parent’s authority changes, what the sub-agent may do can change too. The receiving side checks the requesting sub-agent’s current authority for its particular action.
Potential application areas
Examples of authority questions across different settings:
API and tool access
Is the requested operation within the agent’s current authority?
Data access
Does the agent still have authority to access the requested data?
Resource allocation
Does the allocation fall within the agent’s remaining authority?
Illustrative procurement example
An organisation gives a parent agent authority to commit up to $10,000. The parent delegates authority to commit up to $4,000 to a sub-agent. After the sub-agent has committed $2,500, it requests a further $2,000 purchase through another organisation’s system.
Only $1,500 remains within the sub-agent’s delegated limit. The receiving system uses plusChain verification infrastructure to check the relevant current authority independently. The $2,000 request exceeds that limit, so the action is withheld. plusChain supplies the verification infrastructure; the receiving organisation retains the substantive decision.
Discuss your requirements
Tell us where autonomous agents act across your systems or organisations, and what the receiving side needs to verify.
Contact us