Retail and consumer
Prices that differ between the shelf, the app and the weekly ad. Fresh stock counted once a day. Offers built on last week's basket. We rebuild the data layer so retail AI can act on live signals, inside controls you can show.
Updated October 19, 2026
Retail AI is only as good as the minute its data was last true.
If you run a value or grocery business, your margin is a few cents a unit and your shoppers check the price every visit. The assortment, the stores and the site are working. Somewhere a team is trialing personalization, availability promises or a service assistant, and each one is waiting on data that arrives late. For consumer brands selling direct, the same delay sits between the storefront, the marketplaces and the warehouse.
Price and shrink
A price that differs between the shelf, the app and the ad, or fresh stock counted once a day, turns into markdowns, waste and refunds. In grocery there is no room for any of them.
What waiting costs: margin lost one item at a time, where nobody sees it.
Inventory blind spots
Batch-updated stock data means stockouts lose the sale, and overselling creates fulfillment failures and returns that erode margin.
What waiting costs: margin you lose on both sides of the same number.
Supply chain fragility
When the path from supplier to shelf to customer is not event-driven, a disruption reaches the customer before your dashboard.
What waiting costs: failures the customer sees first.
The personalization gap
Direct-to-consumer competitors personalize at scale while enterprise retailers batch their recommendations overnight.
What waiting costs: a customer experience gap that widens with each season.
| Market | What supervisors are looking at | What it means for an agent |
|---|---|---|
| United States | PCI DSS version 4.0.1 for payment pages and card data, set by the PCI Security Standards Council and checked by your assessor. State consumer privacy laws, such as California's, enforced by state regulators. The FTC on unfair or deceptive practices, including claims about AI. For grocers, the FDA Food Traceability Rule for foods on its list, with compliance required from July 2028. | Card data never enters an agent's context. Customer data is used only for the purposes recorded. What an assistant tells a customer about price or stock is logged and true at the moment it was said. Lot and supplier events are kept so traceability records are ready when the FDA asks for them. |
| India | The Digital Personal Data Protection Act 2023 and its rules, and CERT-In incident reporting directions. | Consent is recorded per purpose before customer data reaches a model, and incidents can be reported from the log. |
| The Philippines | The Data Privacy Act of 2012 and the National Privacy Commission. | The questions are the same and the filings differ. We confirm the local rulebook at the start of every engagement. |
The data exists. It arrives late, in batches, through point-to-point links nobody can trace, so every AI project starts by waiting for the nightly job. Each new agent that touches customer data or payment pages also lands inside payment card and privacy rules.
If a customer buys the last unit at 2:14 p.m., when do your site, your store and your supplier find out?
2:14 p.m., the last unit
Batch world
The item keeps selling until the nightly job.
Event world
Site, store and supplier update in the same second.
Retailers that move their data from batch to events first can let AI act while the customer is still on the page.
One price event published to shelf labels, the app, the site and the ad, as event streams on Google Cloud with Pub/Sub, with a check that flags any channel that disagrees.
You hold one price, true in every channel at the same minute.
Vertex AI, BigQuery and dataInventory, order and supplier events on one governed stream, built as event streams on Google Cloud with Pub/Sub.
You hold: real-time stock and order events reaching every channel together.
Vertex AI, BigQuery and dataApigee between the store, the site, suppliers and marketplaces, replacing point-to-point links.
You hold: one API layer with an owner and a policy for each connection.
Apigee and APIsModels on Vertex AI and BigQuery and agents on Gemini Enterprise for replenishment, service and personalization, inside spend and action limits.
You hold: one agent in production on live data with a decision log.
Vertex AI, BigQuery and dataEasy to start
Choose one workflow, name one owner, and give us time with the people who sign it off.
We bring engineers and the gateway design, we work inside your environment under your security rules, and we write the review pack alongside the build.
A good first workflow is price and availability for one fresh category across store, app and site. See the fixed-price ways to start.
What the work feels like
One workflow is chosen and one person owns it. The questions your reviewers will ask are written down before any build starts. Your merchandising team sees live availability in the first sprint demo.
Governance goes in before the first agent does: the gateway, the audit trail and the access rules. Your risk team sees them working.
A working demo you can show your own review board, not a status slide.
A decision log that can rebuild any action, and a go-live date that was agreed at the start.
Seen it before
Production readiness check
A short readiness check on a live or planned agent workflow. You get a score and the principles worth another look.
Take the readiness checkNo. We work in the integration and data layer around it.
Spend and action limits at the gateway, and a buyer approval step for anything above them.
With event streams on Google Cloud, built on Pub/Sub, so inventory and order events reach the site, the stores and BigQuery together. If you already run another streaming platform, we connect to it instead of replacing it.
Yes. The pressure there is price accuracy, fresh stock and shrink. The work is the same: live events, governed APIs and agents inside limits.
For foods on the FDA's list, you keep key records for each shipment and lot, and you must be able to hand them to the FDA within 24 hours of a request. Supplier, receiving and lot events on one governed stream make those records easy to produce.
Tell us what you are working on. Our Customer Partnership Team will reply to arrange a conversation.
Get answers to your questions
AI assistant. It can make mistakes, so check anything important with our team.
