AI-native Promo Co-Pilot
for CPG Sales Teams

An AI Promo Co-Pilot decides the promotion for each outlet-SKU pair — rather than for a channel or a region. For each pair, it:

Works out which outlets will actually respond to a discount, and which will buy anyway

Selects the SKUs with the highest sales, distribution, or cross-sell opportunity in each store

Recommends the offer type, discount depth, slab, duration, frequency, and eligibility

Simulates the expected sales uplift, distribution growth, ROI, and budget consumption before anything is activated

"Co-Pilot" rather than "autopilot" is the accurate description: it produces the recommendation and the predicted impact, and your team decides what goes live.
What makes it different from a planning tool is that the recommendation doesn't stop at a spreadsheet. It publishes to eligible retailers, applies the offer logic during ordering, pushes the resulting order to the DMS, and measures redemption and business impact automatically — all on one platform.

Traditional planning applies the same scheme across outlets and SKUs that behave completely differently, because that's all a planning cycle has time to do. The result is predictable:

Discounts flow to outlets that would have bought anyway

Offers land on SKUs where they can't move demand

Nobody can tell afterwards which spend created sales and which just cut margin

The Co-Pilot inverts two things. It changes the unit of planning from the scheme to the outlet-SKU pair, and it changes the sequence by predicting impact before activation rather than reporting it after.
Structurally this is possible because the promotions engine and pricing engine sit in the platform's business logic layer, with the recommendation and prediction models in the Intelligence Core directly above them. So the model that recommends the offer is reading the same live transaction data that the engine applying the offer will write to.

Yes — and this should be said early and plainly, because no trade marketing team will hand discount authority to a system they cannot control.
The Co-Pilot recommends and predicts; your team approves what activates. You set the business outcome you want — higher order value, higher purchase frequency, or wider distribution — and the AI builds the promotion to serve it within the guardrails you define.
Those guardrails include:

Budget caps

Maximum discount depth

Eligible outlets

Excluded SKUs

Which promotions require human approval before publishing

A co-pilot that cannot be constrained is not a co-pilot.

A hyper-personalized promotion is one built for a single outlet-SKU combination — rather than applied across a channel, region, or retailer class.
Two shops on the same street can receive different offers, on different products, at different discount depths — because their buying behavior, current assortment, order frequency, and discount sensitivity differ.
The reason this is now possible is that discount sensitivity can be estimated per outlet rather than assumed per segment. And that's where the money is:

Roughly 20%+ of outlets show little or no discount sensitivity yet keep receiving discounts that don't change what they buy — which is pure margin given away.

Personalization at outlet-SKU level is how that spend gets redirected to the outlets where a discount actually changes behavior.

The full commercial range, not a fixed set of templates:

Total order value offers

Bundles and cross-sell

Time-based and slab-based incentives

New SKU and premium distribution pushes

Assortment depth offers

Each carries its own discount depth, slab structure, duration, frequency, and eligibility conditions. The underlying promotions engine supports a large library of discount types and variations, so the recommendation is constrained by commercial logic rather than by what the system can express.
Selection is driven by the outcome you choose, not by the offer catalogue:

Pick higher order value → it builds bundles and basket-value offers

Pick purchase frequency → it builds time and slab incentives that bring retailers back sooner

Pick distribution → it targets outlet-SKU gaps