What have we learnt about pricing analytics in the Gen-AI era?

 Over the past years, our work on pricing analytics projects has given us a clearer view of what Gen-AI changes, and what it does not.

AI can support the full pricing chain, from extracting information in unstructured documents to estimating price elasticity, testing scenarios and explaining recommendations. Three lessons stand out:

1. Pricing starts with reliable data;
2. Analytical AI remains central to estimating demand and optimising prices;
3. Recommendations only create value when decision-makers can understand and trust them.

Pricing starts with better data

Pricing information is often spread across internal systems, PDFs, catalogues, contracts and competitor websites. Before building a model, this information needs to be collected, cleaned and structured.

Gen-AI can extract prices, discounts and rebate terms from unstructured documents, support competitor price monitoring, match free-text product descriptions with master data, and identify duplicates, inconsistencies and outliers.

This stage may receive less attention than modelling, but it is essential. Poor-quality inputs lead to unreliable elasticity estimates and weak pricing recommendations.

Analytical AI remains the engine

Machine learning can identify patterns between prices and sales volumes. Causal models help separate the effect of price from promotions, seasonality, competitor actions and other factors.

These models estimate price elasticity, showing how demand changes when prices move across different segments, products or market conditions. Optimisation models can then test different price combinations against strategic rules, margin floors, price caps and target metrics.

The result is a price recommendation aligned with the company’s strategy, rather than a purely statistical output.

From model output to a business decision

A pricing model only creates value when people can use its recommendations. Decision-makers need to compare scenarios, understand trade-offs and see why a particular price was proposed.

A useful pricing interface should allow teams to ask questions in plain language:

• What would happen if we increased prices for this segment?
• Which price list offers the best balance between revenue, margin and volume?
• Why did the model recommend this price?
• How sensitive is the recommendation to competitor prices or the margin floor?

The system should connect each recommendation to the underlying elasticity, competitor and business logic, while showing the expected impact on demand, margin and portfolio value.

The human role remains central

AI can accelerate data preparation, modelling and scenario analysis, but pricing decisions still require business judgement. Commercial context, regulatory constraints and customer knowledge remain essential.

The most effective approach keeps people involved in validating assumptions, adjusting guardrails and approving price changes. AI expands the range of scenarios teams can analyse and makes the reasoning easier to access. Accountability for the decision remains with the business.

Our main conclusion is that Gen-AI adds value across the pricing chain, but only when it works alongside strong data, analytical models and human judgement. That combination makes pricing decisions faster to test, easier to explain and better connected to business strategy.