foryn

Logistics insights / Perspective

Enterprise AI · Systems of record

One shipment. Five systems. Who owns the decision?

A forwarder takes a customer order. The TMS holds the movement and booking. A separate rate repository holds the carrier agreement. The customer's service instructions live in a commercial system. Visibility events arrive through another provider. Finance records which costs have been approved.

The work is one shipment. The information is spread across several applications, each with good reasons to exist.

An AI assistant embedded in the TMS can be excellent at the tasks close to the shipment record. An agent working across systems can assemble the wider picture. Yet more connectivity does not automatically produce a better decision. It can also make it harder to see who authorized what.

Illustrative scenario / Missed vessel connection

Four permitted actions. One unexamined promise.

An agent finds a recovery route, reserves new capacity, cancels an earlier booking, and sends the customer a revised delivery date. Each step might be accepted by the system that owns it. The entire sequence may still be outside the forwarder's delegated authority.

Booking

Reserve alternate space

The reservation is technically valid.

Operations

Cancel the original

A cancellation charge may follow.

Finance

Accept extra cost

The combined exposure may exceed a limit.

Customer

Promise a new date

The service commitment has changed.

A set of allowed software operations is not necessarily an authorized business decision.

Hypothetical illustration; not a reported shipment or operational incident.

Two useful approaches

The right architecture depends on the work being done.

Embedded intelligence and independent coordination solve different parts of the problem. The decision shouldn't be framed as a competition for the most agents.

Inside the TMS

Near the governed transaction.

An embedded assistant can use the shipment record, roles, reference data, and workflow constraints already present. That proximity matters when someone needs to prepare or confirm a booking, summarize an exception, or inspect a cost.

Its limitation is what lies beyond the application: agreements, instructions, communications, and decisions owned elsewhere.

Across systems

Near the whole customer request.

An independent agent can collect missing facts and coordinate work from different sources. That may be especially useful in companies operating several platforms after acquisitions or in global networks with varying local tools.

Its limitation is authority: successful integration does not create permission to combine separate actions into a commercial obligation.

Source of truth is contextual

Different systems own different facts.

A carrier booking confirmation is the best evidence of a reservation. A signed agreement may determine which rate applies. The customer's instruction controls what changes can be made without asking. An accounting record establishes the cost recognized by the business.

Treating all of these as interchangeable data is a mistake. Before an agent uses a value to justify action, it needs to know where that fact came from, who owns it, and whether it is still current.

The same holds for the human interface. A pricing specialist doesn't want an orchestration diagram. They need the proposed offer, its price basis, the missing evidence, and the decision that actually requires their authority. The technology should clarify the job, not advertise its own complexity.

Foryn / Perspectives

Interoperability matters. Accountability still has to cross the boundary.