The Agentic AI Observatory · cross-domain comparison
How the same research questions change across seven domains
The domains are not expected to produce the same answer. This comparison shows what each question means in systems with different objects, authorities, institutions, physical constraints and market structures.
Procurement carries evidence. The other six columns describe what the research is designed to test, not what it has established.
Select a domain
Agentic Procurement Requisition, purchase order, contract and supplier relationship
Reversibility
Does the ability to undo an action affect how much authority agents receive?
What can still be undone after an approval, an order, a goods receipt or a payment state, and which of those states an agent may set without a second party.
Physicality
Do physical systems preserve incumbent control?
Goods receipt and acceptance, where an official record has to be matched against a physical delivery before the transaction can complete.
Institutional authority
Does regulation restrict autonomy or make delegation more explicit?
Delegation-of-authority schedules, contract terms and audit requirements, and which of them name a human approver explicitly.
Multi-party orchestration
Does coordinating many parties create a new intermediary position?
Requester, approver, buyer, supplier and finance, coordinating inside one firm and across the boundary to another.
Formal versus practical control
Can practical control shift before formal authority does?
Whether formal approval stays with a manager while guided buying determines which suppliers and catalogues are ever seen.
Decentralization versus recentralization
Does distributed agentic activity disperse coordination or concentrate it?
Whether buying disperses toward requesters or concentrates in whichever suite holds the authoritative record.
Agentic Commerce Offer, order and customer relationship
Reversibility
Does the ability to undo an action affect how much authority agents receive?
Returns, refunds, cancellation and payment reversal, and whether the buyer, the merchant or the platform holds the right to trigger each one.
Physicality
Do physical systems preserve incumbent control?
Fulfilment and returns logistics sitting behind an otherwise digital transaction, and what a digital instruction cannot recall.
Institutional authority
Does regulation restrict autonomy or make delegation more explicit?
Consumer-protection rules, payment-network rules and platform policy, and which of them bind an agent acting for a buyer.
Multi-party orchestration
Does coordinating many parties create a new intermediary position?
Buyer, merchant, platform, payment provider and logistics provider, each holding a different part of one purchase.
Formal versus practical control
Can practical control shift before formal authority does?
Whether merchant authorization stays intact while buyer agents control discovery, comparison and intent.
Decentralization versus recentralization
Does distributed agentic activity disperse coordination or concentrate it?
Whether many agents transact directly with many merchants, or a small number of intent layers intermediate them.
Agentic Travel End-to-end journey assembled from interdependent bookings
Reversibility
Does the ability to undo an action affect how much authority agents receive?
Itinerary modification and recovery across providers, and who absorbs the cost and the rebooking when one leg fails.
Physicality
Do physical systems preserve incumbent control?
Seats, rooms and physical presence, which cannot be duplicated, held indefinitely or instantly re-created.
Institutional authority
Does regulation restrict autonomy or make delegation more explicit?
Carrier conditions of carriage and the regulators governing denied boarding, cancellation and refund.
Multi-party orchestration
Does coordinating many parties create a new intermediary position?
Airlines, hotels, aggregators, insurers and payment providers, each with its own rules over one part of a journey.
Formal versus practical control
Can practical control shift before formal authority does?
Whether provider rules stay intact while an orchestrator decides which itineraries are ever assembled and offered.
Decentralization versus recentralization
Does distributed agentic activity disperse coordination or concentrate it?
Whether agents disintermediate aggregators, or a new orchestration layer forms above them.
Agentic Logistics Shipment, inventory and custody state
Reversibility
Does the ability to undo an action affect how much authority agents receive?
Physical release or diversion after custody has changed, and whether a digital instruction can recall an object already moving.
Physicality
Do physical systems preserve incumbent control?
Custody of the object itself, which is the constraint that defines the domain rather than one factor within it.
Institutional authority
Does regulation restrict autonomy or make delegation more explicit?
Customs, carrier authority and the documents that release a shipment, and which of them a coordinating agent can produce.
Multi-party orchestration
Does coordinating many parties create a new intermediary position?
Shipper, carrier, broker, customs and warehouse, coordinating along a chain of custody rather than around a table.
Formal versus practical control
Can practical control shift before formal authority does?
Whether carrier and customs authority stays intact while a coordination layer decides routing, priority and exception handling.
Decentralization versus recentralization
Does distributed agentic activity disperse coordination or concentrate it?
Whether visibility disperses across parties or concentrates in whoever holds the authoritative custody record.
Agentic Financial Transactions Money, account position and financial mandate
Reversibility
Does the ability to undo an action affect how much authority agents receive?
What can be reversed after authorization and what can be reversed after settlement, and the window in which each remains possible.
Physicality
Do physical systems preserve incumbent control?
No physical object. The binding constraint is the settlement infrastructure and the institutions that operate it.
Institutional authority
Does regulation restrict autonomy or make delegation more explicit?
Financial regulation, identity and consent requirements, and the delegated mandate that defines what an agent may initiate.
Multi-party orchestration
Does coordinating many parties create a new intermediary position?
Payer, payee, their institutions, the networks between them and the settlement infrastructure underneath.
Formal versus practical control
Can practical control shift before formal authority does?
Whether mandates stay explicit while an agent determines timing, counterparty and amount within them.
Decentralization versus recentralization
Does distributed agentic activity disperse coordination or concentrate it?
Whether delegation disperses across providers or concentrates in the institutions holding identity and the settlement infrastructure.
Agentic Hiring Candidate, role and employment relationship
Reversibility
Does the ability to undo an action affect how much authority agents receive?
Contestability before and after a rejection or an offer, and whether a candidate screened out early can re-enter at all.
Physicality
Do physical systems preserve incumbent control?
No physical object. The binding constraint is a person's consent, availability and willingness to proceed.
Institutional authority
Does regulation restrict autonomy or make delegation more explicit?
Employment law, anti-discrimination rules and the duty to keep a decision attributable and contestable.
Multi-party orchestration
Does coordinating many parties create a new intermediary position?
Employer, candidate, agencies, platforms and background providers, each holding part of one hiring decision.
Formal versus practical control
Can practical control shift before formal authority does?
Whether formal human approval stays in place while ranking and screening shape who is ever considered.
Decentralization versus recentralization
Does distributed agentic activity disperse coordination or concentrate it?
Whether sourcing disperses across employers or concentrates in the platforms holding the candidate graph.
Agentic Energy Distributed physical asset, capacity and energy position
Reversibility
Does the ability to undo an action affect how much authority agents receive?
What can be reversed after dispatch, a trade or a physical response, and whether a grid state can be unwound at all.
Physicality
Do physical systems preserve incumbent control?
Generation, storage and load, which have to balance physically and in real time rather than on a ledger.
Institutional authority
Does regulation restrict autonomy or make delegation more explicit?
The system operator, market rules and safety regulation, and which actions they reserve to a licensed party.
Multi-party orchestration
Does coordinating many parties create a new intermediary position?
Asset owners, aggregators, retailers, system operators and market platforms, coordinating continuously.
Formal versus practical control
Can practical control shift before formal authority does?
Whether formal authority stays with system operators while coordination platforms shape practical dispatch choices.
Decentralization versus recentralization
Does distributed agentic activity disperse coordination or concentrate it?
Whether distributed assets trade peer to peer, or aggregate under a small number of coordinators.
This comparison sets out the structural questions and hypotheses applied to each domain. It does not present cross-domain results. Each cell states what the question would require the research to observe in that system; where evidence exists, its standing is stated on the claim itself rather than inferred from this table.
Where to go next
-
The Observatory
The portfolio, the shared three-level method and what would establish a reshuffle.
-
The method applied to evidence
Procurement, where these questions have been put to a corpus a reader can inspect.
-
How to read a comparison like this
What kind of record counts as evidence, and what a claim drawn from it may say, are two separate readings and neither substitutes for the other.
-
Apply the method
Build an Observatory for a client, map the reshuffle around a company, or commission co-authored research.