Key Highlights
- AI agents are changing how consumers research products and make purchases, requiring marketers to think beyond websites and buying experiences designed exclusively for humans.
- 3M's Ask 3M assistant uses verified technical documentation to help industrial customers compare products, troubleshoot applications and connect with authorized distributors.
- Home Depot has rolled out its Magic Apron AI assistant to more than 2,000 U.S. stores, providing product recommendations and in-store guidance tailored to customers' locations.
- Meta's Muse agent can handle shopping tasks and complete purchases with user approval, introducing new considerations for product pages and checkout experiences.
- Marketers should test whether AI agents can navigate their websites and complete purchases while ensuring product information is structured, searchable and accurate.
Welcome to Unprompted: The AI Marketing Brief, where I cut through the noise in AI news and research to show marketers what’s happening — and why it matters for your work, your team and your career.
If I asked you to describe your audience, could you do it? Of course you could. You'd start listing personas, rattling off pain points and bragging about all the behaviors you track. You eat, sleep and breathe demographics.
But your answer might be missing one important detail. Your most important audience member might not be a person at all.
AI is reading your content as closely as your most loyal subscriber, and sometimes more often. If you aren't building your campaigns with AI in mind, you could be leaving sales on the table.
I can hear your rebuttal now: "We already optimize for search engines. Isn't that enough?"
Not anymore. Optimizing for generative engines follows most of the same principles as SEO, but it doesn't account for the rise of AI agents. It's one thing to make your content easy for AI to find, read and recommend. It's another to build a site that humans and agents can both actually use.
So if your list of personas doesn't include Arthur the AI, your audience strategy might need a rethink. In this edition of Unprompted, we'll look at how AI is changing the way people research and buy, whether they're shopping online or standing in a store aisle.
AI shopping assistants are here to help
Some big, recognizable brands are already using AI to rethink how customers research and buy their products. 3M recently launched Ask 3M, an AI-powered assistant at ask.3m.com that gives industrial customers self-service access to technical expertise. For now, it focuses only on industrial adhesives and tapes, but that narrow scope doesn't make it any less impressive. Users can compare products and troubleshoot applications in plain language, so you don't need to be an engineer to get a useful answer. Ask 3M also links its suggestions to authorized distributors and lets users download source documents right in the conversation.
So what makes Ask 3M different? Most chatbots pull from information scattered across the internet. Ask 3M's answers come only from verified documentation across 3M's 49 technology platforms. That helps keep its guidance accurate and relevant. It also shows how a company can turn its technical documentation into a self-service tool that sends buyers straight to the people who sell its products.
Home Depot is making a similar bet on a much bigger scale. The company just rolled out Magic Apron, its AI shopping assistant, to all 2,000-plus U.S. stores. It might sound like just another chatbot, but it's actually a rare blend of digital and in-person shopping. Magic Apron draws on local store knowledge, so customers get product advice, directions to the right aisle, and project help tied to the store they're in. They can type, talk, upload a photo or ask in multiple languages, either through Store Mode in the Home Depot app or by scanning a QR code in the store. It's already fielding millions of questions a month.
While many companies are still running small pilots, Home Depot has put real numbers behind a national rollout. That makes Magic Apron one of the most useful reference points out there if you're building the business case for AI-assisted customer service or in-store guidance.
Meta gives users access to an easy-to-use agent
So far, we've looked at brands using AI to answer their customers' questions. Meta is putting AI on the other side of the counter. It just launched Muse, a personal AI agent that handles tasks like sending emails, booking travel and shopping, and it can keep working on longer-term goals even after the app is closed. It checks in before sensitive actions like purchases and supports checkout through Stripe's Link, including purchase protections. For a lot of people, Muse could be their first real experience handing a task to an AI agent.
That could change the entire buying cycle. When an AI agent can browse, fill out forms, negotiate and complete purchases on someone's behalf, it changes who (or what) you're actually selling to. Until now, product pages, checkout flows and offers have been designed for humans. Brands that want to stay competitive may increasingly need to design for an agent, too.
Claude gets more accurate, but isn't perfect
All of this depends on AI finding the right information and interpreting it correctly. Anthropic recently released Claude Opus 5.5, which the company says writes more clearly than its previous version. It's built to put the most important information first, follow the writing rules users give it, and invent fewer figures and quotes than earlier models.
Anthropic tested that last claim by having the model write a quarterly earnings report using only sources it could find on a copy of the web. Any invented figure or quote counted as a failure. Opus 5.5 passed 16 of 18 times. Neither Opus 5 nor Anthropic's top-tier Fable 5.1 passed once.
If those results hold up, marketers could get cleaner drafts with fewer fabricated stats and less time spent fixing style problems. But it still failed twice, and those are Anthropic's own tests. Before taking any claims at face value, run your own tests and check the results carefully. It's also a good reminder of why 3M kept its assistant away from the open web.
Things to do this week:
- Put Home Depot in your next deck. Add Magic Apron to your next AI-in-retail comparison deck or pitch as a named, sourced example. Cite the "2,000+ stores" and "millions of questions per month" figures straight from Home Depot's announcement.
- Check whether your docs could feed an Ask 3M. Review your product documentation and technical spec sheets to see if they're structured well enough to power a similar assistant. Flag any gaps in searchable, verified content now, because you'll have to fix them before a tool like this is even on the table.
- Send an agent to your checkout. Have an AI browsing agent, like Claude or ChatGPT with browsing, try to complete a purchase on your site. Note every product page or checkout step where it fails or gets stuck.
- Test Opus 5.5 against your style guide. Give it your brand style guide and have it draft one piece that cites data. Check every figure and quote against its source, and count how many edits you had to make. That number is your baseline for comparing tools.
About the Author
Alexis GajewskiAlexis Gajewski
Contributor/AI Expert
Alexis Gajewski is a media strategist with nearly 20 years of experience in editorial leadership, content strategy, digital content creation, and search optimization. She is known for building the systems, training, and standards that help writers, editors, digital creators, and marketers produce better work more efficiently, without sacrificing accuracy or voice.
Alexis specializes in turning complex, emerging concepts into clear, actionable frameworks. Whether the subject is AI, search engine optimization (SEO), or editorial standards, she develops approaches teams can adopt quickly and apply with confidence. Her training serves every level, from foundational AI literacy to advanced workflows.
She is recognized for her work at the intersection of journalism, AI, and search. As audiences increasingly find content through AI assistants and answer engines, Alexis helps organizations adapt through answer engine optimization (AEO) and generative engine optimization (GEO), ensuring their content is well structured, authoritative, and readily surfaced and cited. She believes AI should strengthen sound writing and editorial judgment, not replace them.
An experienced educator and collaborator, Alexis has designed and led professional development programs for more than 150 editors across dozens of B2B brands and has aligned editorial, marketing, and audience development teams around shared goals. Her years covering the manufacturing industry instilled a continuous-improvement mindset that shapes her approach: identify inefficiencies, implement solutions, measure results, and refine.
Alexis also works as a freelance writer, editor, and consultant for publishers, agencies, and brands, specializing in content optimization and discoverability, AI-assisted workflows, content strategy, audience development and analytics, and editorial quality.
Resources
Quiz
Related
dreamstime_xl_415807152
Customer journey concept editable stroke outline icon isolated on white background flat vector illustration.Elevate your strategy with weekly insights from marketing leaders who are redefining engagement and growth. From campaign best practices to creative innovation and data-driven trends, MarketingEDGE delivers the ideas and inspiration you need to outperform your competition.


