Manufacturing

Supply chain planning.
Past the optimiser.

Supply chain teams are usually not short of an optimisation engine. They are short of trustworthy inputs — lead times that reflect reality, supplier commitments that mean something, and early warning that a constraint has moved.

GARBAGE IN, OPTIMISED GARBAGE OUT

A planning system runs on master data: lead times, minimum order quantities, yields, capacity, supplier calendars. In most plants that data was entered years ago and has drifted from reality. A supplier whose quoted lead time is four weeks and whose actual delivered lead time has been seven for eighteen months is producing plans that are wrong before any solver touches them.

The highest-value first project in this area is almost never optimisation. It is comparing master data against actual observed performance and surfacing the differences — a task that is straightforward, unfashionable, and reliably produces findings that change plans immediately.

Where language models genuinely add capability is in the unstructured half of the supply chain: supplier emails, order confirmations, shipping documents and advance shipping notices, which carry commitments and changes that never reach the planning system because they arrived as prose. Extracting those into structured signals is a real gap and one nobody had a good way to close before.

WHERE IT HELPS

Master-data reality check

Observed lead times, yields and reliability compared against what the planning system believes. Usually the first deliverable, usually uncomfortable reading, and usually the change with the largest immediate effect on plan quality.

Reading the correspondence

Order confirmations, delay notices and shipping documents parsed into structured events that update the plan. Today a planner reads these and updates a spreadsheet, if they have time.

Supplier risk signals

Monitoring public sources for events affecting suppliers — financial distress, disruption at a site, regional events — and connecting them to the parts and lines they would actually affect. The connection to your bill of materials is the part that makes it useful rather than a news feed.

Scenario explanation

Translating what a plan change means in operational terms: which lines are affected, which orders are at risk, what the options are. Planners spend a great deal of time producing this explanation by hand for other people.

WHAT WE WOULD NOT DO

Replace a working optimisation engine with a language model. Constrained optimisation is a mature class of problem with excellent classical tooling, and a model that produces a plausible-looking plan without a feasibility guarantee is a downgrade dressed as an upgrade.

Automate purchase order commitments. A system that places orders without human confirmation converts a data error into a financial one. The agent prepares; a person commits.

Promise disruption prediction. Detecting that something has happened and propagating its consequences quickly is achievable and valuable. Forecasting the next unprecedented shock is not, and any vendor claiming it should be asked what they predicted last time.

CHECK THE DATA BEFORE THE SOLVER

PLANNING PATH

Master data

lead times, yields, capacity
Reality check

observed vs assumed performance
Read correspondence

confirmations, delays, ASNs
Structured events

written back to the plan
Existing optimiser

kept, and fed better inputs
Purchase commitment

a person, every time

Solid: the input path we build Dashed: the commitment that stays human

The language model reads what the planning system cannot. The optimisation itself stays with the classical solver that already does it well.

QUESTIONS

QUESTIONS — 3

No, and we would advise against it as a first move. This work sits alongside them — reading what they cannot read, checking what they assume, and writing structured signals back. Replacing a planning system is a multi-year programme with a poor success record and should be a separate decision.

It is the project. Master-data quality is where the value is, and a system that surfaces the gaps between assumption and reality usually shows value before any model is deployed.

It can prepare the position — historical performance, current exposure, comparable terms — which is genuinely useful and saves real time. The conversation stays with your buyers, who are managing a relationship rather than executing a transaction.

Keep reading

Cross-Platform Workflows

→

Reference

Document Pipeline Architecture

→

30-minute technical call · no deck