Microsoft Advertising recently ran a webinar on agentic commerce: AI assistants that don’t just recommend products but buy them on a shopper’s behalf. We listened, so you don’t have to. Below are the five points we think matter most for food and drink brands, and what we’d do about them.

The short version: AI is fast becoming a route to purchase in its own right. The brands that win will be the ones with clean, rich product data, a short path to checkout and a clear view of how AI talks about them.

1. AI is becoming a shopping channel, not just a search tool

Microsoft describes three versions of the buyer journey, running side by side. In the first, people search and buy for themselves. In the second, AI helps them compare, but they still click buy. In the third, the shopper sets a goal and an AI agent completes the purchase. All three will exist together for some time, so brands need to be ready for each.

The numbers Microsoft shared are mostly from the US not UK, but they show how quickly this is moving:

43% of shoppers say they’ve discovered a new brand through AI
82% of heavy AI users rely on it to choose products for them
shoppers referred by AI convert 42% more often than those referred by search

What it means for you: the job hasn’t changed. Brands still need to be found, be chosen and close the sale. What’s changing is who, or what, is doing the shopping.

2. Your product data is your new shop window

AI assistants would much rather read a structured product feed than scrape your website. Adobe research quoted on the call found that two-thirds of product pages can’t be read properly by machines. If your products aren’t in a feed, there’s a real chance AI simply won’t recommend them.

Microsoft calls a well-built feed an “agentic catalogue”. In practice, that means:

Complete: every product you want found, not just the ones you advertise
Accurate: clear titles, good descriptions, GTIN barcodes, categories and decent images
Fresh: price and stock updated at least daily, so AI doesn’t recommend something that’s sold out or last season’s pack
Rich: extra context such as allergen warnings and customer service links, with FAQs and usage guides coming soon

One feed now powers Copilot, Bing and, soon, the Edge browser, and these placements are organic, so there’s no media cost. Microsoft also backs a shared industry standard, the Universal Commerce Protocol (UCP), so data enriched once can serve several AI platforms. Shopify gave an early indicator: conversions were 1.8 times higher when Copilot used its AI-ready catalogue!

For food and drink brands: AI shoppers rarely name a product. They ask for a gluten-free pudding for eight, a midweek tea the kids will actually eat, or a premium canapé range perfect for banqueting. Detailed attributes, ingredients and allergen data will decide who gets picked.

3. Checkout is moving into the AI conversation

Until now, a shopper who found the right product in an AI chat had to leave and pick up the journey on the retailer’s website, and many dropped off along the way. Copilot Checkout lets them buy without leaving the conversation, with their details already filled in. The retailer remains the merchant of record keeping the customer relationship.

Shopify merchants are switched on automatically and can opt out. Others can connect through PayPal or Stripe.

Our take: this mainly affects brands that sell direct to consumers. Microsoft didn’t confirm UK timing or share any performance data, so we’d keep an eye on it rather than act on it today.

4. You can now measure how AI sees your brand

When someone asks an AI assistant a question, it breaks the question into several smaller searches behind the scenes and builds its answer from the sources it trusts. To be recommended, a brand has to earn its place across all of those searches, not just one keyword.

Microsoft Clarity, a free analytics tool, now reports on this in three ways:

AI crawler activity: which AI bots visit your site, how often, and which pages they read
Citations: how often AI uses your site as a source, and your share compared with competitors, split between searches that name your brand and those that don’t
Topic gaps: where competitors are being recommended instead of you, with specific suggestions for content to fill the gap

You can track up to 150 questions a week for free, written the way a real customer would ask them.

Worth knowing: two findings from the demo stood out. AI crawlers favoured guides and articles over product pages. And brands lost out on detailed, specific questions. Detailed, genuinely useful content is what earns a recommendation: recipes, usage ideas, technical specs.

5. AI shoppers behave differently once they reach your site

According to Microsoft’s own Clarity data, visitors who arrive from AI assistants convert at around three times the rate of other channels – that’s pretty impressive. They’ve usually done their research already, so they want to confirm the price, the right variant and the key details, then buy. Visitors from social or organic search tend to browse and compare more first.

Clarity can separate traffic by source and show recordings and heatmaps for each group, so you can see where AI-referred visitors get stuck. It runs alongside Google Analytics rather than replacing it.

What to do: look at your key product pages through the eyes of someone ready to buy. Price, pack size, availability and the buy button should all be visible without scrolling. It’s basic UX design.

What we recommend

None of this needs a big budget to get started. For most food and drink brands, we’d suggest four steps:

Check your product feed. Make sure every product is included, with full attributes, ingredients and allergen information, and that prices and stock update daily. If you already have a Google feed, it can be imported into Microsoft Merchant Center in minutes.
Find out how AI talks about you. Set up Clarity and track the questions your customers actually ask, alongside your main competitors.
Fill the content gaps. Build out the detailed, useful content AI looks for: recipes, serving ideas, comparison guides and clear answers to common questions.
Tidy up your key pages. Make sure the details a ready-to-buy visitor needs sit clearly above the fold.

Some of Microsoft’s newer features are US-only for now, and most of the figures quoted are US-based. We wouldn’t wait for the UK rollout, though. Better feeds and better content pay off in Google and ChatGPT too. And as ever, it’s better to be ahead of the game than chasing your competitors.

How jellybean can help

We’ve spent 39 years marketing food and drink, one of the most tightly regulated categories there is. Allergen law, HFSS and LHF rules and of course listings that live or die on a retailer’s site are our bread and butter. Our search team already works on SEO, AEO and GEO for food and drink brands, so this is a natural next step. Want to see how AI talks about your brand today, or a second pair of eyes on your product data? Get in touch today here