Calculated, not predicted
Five rules for AI in retail, and how an agent earns the right to make the call.
Every retail software company is an AI company now. The word is on every homepage, every trade-show stand and every pitch, and it has almost stopped meaning anything.
So this edition is about what it means here. Laminir is built on AI, and the way it uses AI is deliberately different from most of what you'll hear this season. These are the five rules it's built to, and why each one matters to the people who carry a retail business.
01
Calculated, never generated
Most retail AI is a prediction engine. It forecasts demand, suggests a buy and hands you a number you can't check. When it's wrong, nobody can say why.
In Laminir, every figure is calculated the way a planner would calculate it, from your own sales, stock and orders. The same inputs always give the same number. The AI's job is to read that number, connect it to the others and explain it in plain language. It does the thinking around the arithmetic, never the arithmetic itself.
- Out
- “The model says so.”
- In
- The number, and exactly how it got there.
02
Autonomy is earned, not assumed
Retail AI is heading towards agents that act on their own, inside limits the business sets. That's where Laminir is heading too. It's why they're called agents. The question is how a business gets there without handing over its buy on trust.
Today, every recommendation carries its reason and an owner. The buyer approves it, defers it or overrides it, and every override is recorded with the reason. That record is how autonomy is earned. When an agent's calls on a kind of decision have matched the buyer's, season after season, that decision can be handed over, inside limits the business sets, with every action still on the record.
The routine calls go first. The order written at market stays with the buyer longest.
- Out
- Switched on, and trusted from day one.
- In
- Earned call by call, on your own record.
An agent should earn the decision before it's given one.
03
It knows what you're trying to do
Ask a general AI tool about your stock and it will answer the question you asked. It doesn't know your plan for the year, what you've committed to the board, or that you decided last month to protect full-price margin over volume.
Laminir holds the plan and the strategy alongside the numbers, and every recommendation is read against them. A markdown that clears stock but breaks a margin commitment is shown as exactly that. The AI isn't just answering questions. It's holding the business to what it said it would do.
- Out
- A clever answer to the wrong question.
- In
- A call read against your plan and your priorities.
04
It remembers what you decided
Most AI starts every conversation from nothing. Ask it the same question next season and you'll get a fresh answer, with no idea what you decided last time or why.
Laminir keeps the decision history: what was recommended, what was decided and by whom, the reason for an override, and what stood in the way. When the same brand or the same question comes round again, the recommendation starts from there. That record is the part nobody else can copy, because it's yours.
- Out
- A fresh answer every time.
- In
- The answer, with what you decided last time beside it.
05
Built for the people who carry the business
Most retail AI is built for enterprise: a planning team, a data warehouse, a BI department and a long implementation. The businesses that need the help most have none of those. The owner is often the buyer, and two people carry fifty brands.
Laminir reads the systems you already run, with no warehouse to build and no team to hire. The first full cycle runs in about a week, on your own numbers.
- Out
- Phase one of a twelve-month programme.
- In
- Your own numbers, read in a week.
Why it's built this way
I've spent twenty-five years in retail leadership, and too many of the big calls I've seen were made late on a Sunday from a spreadsheet only one person understood. I didn't want a tool that took the decision away from that person on day one. I wanted one that did the work around it, showed its working, remembered, and earned the right to take on more. That's the difference, and it's deliberate.
Check this week
Ask one question of every number your current tools give you: where did it come from? If nobody can answer in a sentence, you're trusting a black box with your buy.