Category: Uncategorized
Agentic AI and the shift handover problem
An agent that runs continuously meets an organisation that stops every eight hours. What the outgoing shift knew and did not write down is where the decisions break.
Retail AI earns most of its value in assortment
Forecasting and replenishment get the attention. The larger and less contested gain is deciding what to carry, where, and what to stop carrying.
What an agent must know before it can price
Price elasticity is the easy part. The constraints that make a price legal, contractual and commercially survivable are what the model does not have.
Network inventory optimisation is a different problem
Optimising each location separately produces a network that is individually correct and collectively wrong. The interaction between locations is the whole problem.
Multi-agent systems, when decisions genuinely conflict
Several agents optimising separately will reach incompatible conclusions. The design question is not how they cooperate, but who resolves it when they do not.
Demand sensing is not a forecast with a shorter horizon
Running the same model weekly instead of monthly is not demand sensing. Sensing uses signals a forecast never sees, to change a decision a forecast cannot reach.
How do you prepare ERP data for decision-making?
ERP data is designed for transactions, not decisions. Every analytics project that skips that gap stalls in a data quality argument.
The pilot that proved nothing
It ran on clean historical data, in an isolated environment, against a metric agreed afterwards. It succeeded. It told you nothing about whether this works.
How is price elasticity measured?
Measuring elasticity is not comparing two price points. Without separating promotion, season and competitor movement, the number you get is wrong.
Why store-level demand forecasting differs from chain-level
A model that forecasts total demand accurately can be systematically wrong per store. The problem is not the model — it is the level of aggregation.