A new generation of AI-assisted menu intelligence is giving food and beverage manufacturers a monthly read on where their categories are growing — and where they are not.
For most of the past three decades, an Australian manufacturer wanting to understand the foodservice channel had two options: commission a research project, or wait for a syndicated report. Both delivered a snapshot of a market that had already moved on.
That gap is closing. A cluster of data platforms has emerged over the past five years that gathers menu, price and location data directly from the operator universe, classifies it automatically, and refreshes it every month. Among them is Brizo by Datassential, formerly Brizo FoodMetrics, which became available in Australia and New Zealand in January 2025 through Sydney-based research firm Food Industry Foresight.
Two million establishments, refreshed monthly
Brizo’s database covers more than 2.1 million foodservice establishments across six countries, made up of more than 1.4 million independents and around 660,000 chain locations belonging to more than 10,000 chains. Against those establishments it has observed more than 2.5 billion menu items and more than 171 million menu prices, alongside 125 or more cuisine classifications and detail on more than 1,100 restaurant technologies in use.
In Australia and New Zealand, the platform covers more than 170,000 establishments. Records are drawn from more than 40 marketplaces, ordering platforms and review sites and rebuilt monthly, with pre-opening venues tracked weekly .
The scale is the point. A cafe adding a plant-based patty, a pub lifting its schnitzel price by two dollars, a new dark kitchen opening in Marrickville — individually these are noise. Aggregated across a national menu set every month, they become a leading indicator.
Where the machine learning sits
The heavy lifting is classification. Raw menu text is inconsistent, abbreviated and frequently misspelled, and the same dish appears under dozens of names. Turning that into structured, searchable fields — dish, protein, ingredient, dietary claim, cuisine, price band — is a language-processing problem at a scale no analyst team could match manually.
Brizo’s parent, Datassential, has been layering generative AI capability across its wider platform, including a natural-language search function that returns cited sources, menu-tagging and concept-ideation tools, and an interface that lets its data be queried directly from assistants such as ChatGPT and Claude.
It is worth being precise about what the AI does and does not do here. It is not forecasting demand or writing strategy. It is reading, standardising and indexing an enormous volume of unstructured commercial text so that a category manager can ask a specific question and get an answer in minutes rather than months.
What manufacturers are using it for
Four use cases dominate for manufacturers.
Market sizing. Rather than estimating a total addressable market from top-down channel splits, a manufacturer can count the establishments that actually serve the relevant occasion. Upper Crust Enterprises, a United States panko breadcrumb producer, sized its opportunity by filtering the establishment universe for venues serving fried food, then searching menus at ingredient level for panko and tempura references.
Operator segmentation. Protein processor Tyson uses establishment-level cuisine and ingredient data together with median meal price to group operators, then matches products to the price points each group can carry.
Menu trend tracking. Because the data is a monthly time series rather than a single cut, the rate of change is measurable. A manufacturer can see how quickly an ingredient, claim or format is spreading, in which channels, and at what price — and distinguish a genuine trend from a metropolitan fashion.
Distribution and white space. Comparing where a brand’s products appear on menus with where the addressable venues sit exposes the gaps, by suburb, by channel and by cuisine.
Brizo reports customers cutting prospecting time by up to 75 per cent. The more defensible benefit is simply that questions which previously required a research brief can now be answered inside a working session.
Access and integration
The platform is used through a browser for immediate answers, with results exported to spreadsheets or pushed into a CRM. For manufacturers wanting to track metrics over time or blend external menu data with their own sales history, the same data can be delivered into a cloud data warehouse such as Snowflake or BigQuery. Access is sold as an annual subscription across several tiers.
That warehouse option is where the sharper analysis tends to happen. Menu data on its own describes the market; joined to internal sales data, it describes a manufacturer’s position in that market — penetration by delivery area, share of the venues that could stock a product, and the price gap between a brand and the category average.
The limits worth knowing
These tools read the visible market, which means the public-facing menu. Menus that are not published online, back-of-house specifications, contract catering tenders and institutional foodservice are all under-represented. However, the huge amount of data that is visible gives strong, clear and ongoing evidence of what foodservice operators are putting on menus. This means we can see what they in turn must be buying to be able to serve those menu items.
The direction of travel
The underlying shift is not really about AI itself. It is about frequency and the ability to analyse huge amounts of data. A market that used to be described annually can now be described monthly, at establishment level, for a subscription cost well below a single custom research project.