By Akeneo
AI is changing how products are compared, evaluated and selected. Justin Thomas, VP of Sales, EMEA North at Akeneo, argues that pricing can no longer be managed separately from the product information that defines its value and context.
AI is removing the traditional separation between pricing and product information. Merchandising and commercial teams may set the price, while product and ecommerce teams manage the information surrounding it, but AI increasingly requires both to work together.
Pricing is becoming a product data problem because the quality of every pricing decision depends on the context delivered through product attributes, images, and descriptions. If that context is incomplete, inconsistent or inaccessible, even a commercially sensible price can look wrong to the systems increasingly shaping discovery and demand.
When an AI agent compares two washing machines, it needs more than two prices. It needs comparable capacity, energy performance, dimensions, noise levels, programmes, warranty terms and delivery conditions. If one model is described fully and another has key attributes buried in prose or omitted altogether, the comparison is distorted before price is considered.
The cheaper product may be overlooked because the agent cannot verify that it meets the customer’s requirements. The more expensive product may appear poor value because its distinguishing features are not structured clearly enough to justify the premium.
The same applies across categories. A food product’s price makes sense in relation to pack size, ingredients, provenance and dietary status. A fashion item depends on fabric, fit, construction, care and availability by size and colour. Price is meaningful only when those attributes are complete, normalised and comparable.
Confidence versus reality
Akeneo’s survey of 200 UK IT decision-makers reveals an apparent contradiction. 95% rate the product and operational data used by their AI systems as good or excellent, including 43% who describe it as excellent.
Yet respondents estimate that an average of 43% of total AI project effort is spent preparing data through cleaning, structuring and labelling. Almost three in ten say data preparation consumes more than half of project effort. Organisations may trust their data in general while still paying a substantial tax to make it usable for individual AI applications.
Good data quality depends on context. Information designed for a website or ERP process may be inadequate for an AI agent to support product comparison and discovery. AI requires greater consistency, granularity and semantic clarity across the product catalogue.
The survey bears this out; 8% cite data silos as a challenge to scaling AI, 15% lack of data governance and 14% inconsistent or incomplete product attributes.
For AI to make commercially useful pricing comparisons, it needs reliable data from structured enterprise systems such as PIM, ERP, DAM and CRM. 71% of respondents say their organisations already use these systems as primary data sources for AI.
As AI depends on these systems to understand products, the quality of their data can influence which alternatives are surfaced, how value is assessed and whether a recommendation is made.
The effect extends beyond a company’s website. 60% expect autonomous shopping and comparison agents to be among the channels most influenced by AI during the next three years, ahead of direct websites at 51%. Already, 42% of respondents use AI for search, discovery or personalisation.
Product evaluation will increasingly happen in marketplaces, conversational interfaces, search summaries and independent agents. In those environments, brands cannot rely on page design or persuasive copy to repair missing product facts. The structured product record has to do the heavy lifting because otherwise, bad data distorts price.
First, it weakens comparability and competitive intelligence. If measurements, bundles or pack quantities are inconsistent, AI may compare products that are not genuinely equivalent or calculate unit value incorrectly.
If product matching relies on weak taxonomy or incomplete attributes, the competitor set itself may be wrong. A precise algorithm applied to false equivalence produces a confidently mispriced result.
Second, bad data quality conceals differentiation. Premium pricing depends on evidence of additional value. If superior materials, longer warranties, better performance or sustainability credentials are absent or expressed inconsistently, AI may treat a differentiated product as a commodity and favour the lowest visible price.
Third, it encourages indiscriminate discounting. When businesses cannot identify which attributes drive preference, price becomes the easiest lever to pull. This may generate a short-term conversion response while eroding margin and brand position. Better product information helps distinguish a genuine price objection from a failure to communicate value.
Finally, inconsistent data fragments pricing across channels and markets. An offer may be correctly priced in one region but misunderstood elsewhere because units, language, regulatory attributes or variant relationships have not been localised and governed consistently. The number travels; the meaning does not.
Pricing governance must include product data governance
Installing a PIM does not resolve every commercial issue. Pricing, product, ecommerce, data and AI teams now share a dependency that must be governed deliberately.
Organisations should identify both the attributes that make products comparable and those that justify price differences within each category. They should establish common definitions, units, taxonomies and validation rules; connect price and promotional data to the correct products, variants, packs and markets; and monitor completeness and consistency from the perspective of the AI use case rather than merely the source system.
AI will make pricing more dynamic, but dynamism is not the same as intelligence. Changing a price quickly has limited value if the system does not understand what is being priced, which products are genuinely comparable or why customers should pay more for one option than another. Product data is the evidence through which that price is interpreted.
Published 01/10/2026
