Marketing. AI. Strategy.
A sourced database of 140 brands using AI in marketing: the tool, the result, and where I found it. Updated for 2026.
The database
Search, filter by category, save the ones you want to come back to, or share a single entry. Every card links to its primary source.
A few brands run more than one documented initiative, so the example count runs ahead of the brand count.
The most public-facing AI work, and the bucket most people picture when they hear "AI in marketing." Most of these are about scaling creative production or letting platform algorithms do the bid allocation. Heinz, Coca-Cola, JPMorgan, and Cadbury are all here.
Beverages
Recreated its 1990s "Holidays Are Coming" commercial with generative AI. Head of generative AI Pratik Thakar says the film scored an average success rate on acceptability, understanding, and entertainment value, but passed Coca-Cola's own and competitors' Christmas ad benchmarks. The Create Real Magic platform also turned its 1931 Santa into a character people could converse with.
Social Media & Advertising
Advantage+ shopping campaigns automate targeting, creative, placements, and budget in real time from the advertiser's inputs. Meta reports the setup helps increase return on ad spend by 32%.
Digital Advertising
Performance Max allocates budget across Search, YouTube, Display, Gmail, and Maps from a single campaign, using Google AI for bidding, budget optimization, audiences, creative, and attribution against the conversion goals the advertiser sets.
Food & Beverage
Agency Rethink gave DALL-E 2 the instruction to depict a ketchup bottle, and the outputs all looked like Heinz. Rethink calls it the first AI-generated ad campaign, built from initial pitch to release in under a month, and it swept The Drum Awards Festival 2023.
Food & Beverage
Machine learning turned a single Shah Rukh Khan film into ads that doubled as spots for thousands of local businesses hit by the pandemic. Built by Ogilvy Mumbai and Wavemaker Mumbai, the campaign won a Cannes Lions Titanium and an LIA Grand.
Financial Services
Signed a five-year deal with Persado after a pilot in which Chase saw click-through lift on Persado-written ad copy as high as 450%, against others in the 50 to 200% range.
Automotive
Albert AI ran digital media buying for the New York dealership, starting with a "48 bikes in 48 hours" push to clear seasonal overstock that nearly doubled the previous sales record. After six months the dealership credited 40% of its motorcycle sales to Albert.
Advertising & Marketing
Production Studio gives clients direct access to WPP's generative content engine through the WPP Open operating system, built on NVIDIA's Omniverse platform and the OpenUSD standard for 3D workflows. WPP pitches it as exponentially more content per campaign.
Advertising & Marketing
CoreAI is a groupwide system meant to put AI capability into every part of the business, backed by a 300 million euro commitment over three years with 100 million allocated to 2024, split evenly between people and technology. Publicis announced it off the back of 6.3% organic growth in 2023.
Consumer Goods
Beauty AI Studio runs the full content lifecycle, from ideation through personalisation across priority markets, for Unilever's Beauty & Wellbeing power brands including Dove, Vaseline, POND'S SKIN INSTITUTE, and CLEAR.
Food & Beverage
Ada is PepsiCo's internal AI platform, described by one executive as "like our Google." On creative development the team uses it to test work from early-stage ideas through to finished executions, building on the best ideas until they land on a strong one.
Professional Network
Accelerate reads an advertiser's website, LinkedIn page, and prior ad accounts, then builds the creative and targets the audience, with Microsoft Designer available for refinement. LinkedIn reports it drives 52% lower cost per action than classic campaigns and cuts campaign creation from 15 hours to 5 minutes.
Beverages
Named WPP its global agency partner for shopper marketing and commerce across the beer portfolio, with WPP Open, the group's AI-powered operating system, supporting the work. It covers in-bar, retail, and e-commerce activity plus shopping integrations around sponsorships including Formula 1 and the UEFA Champions League.
Consumer Goods
Leaning on what CEO Shailesh Jejurikar calls an "enormous wealth of consumer data" plus AI to handle a fragmented media reality where "retailers are becoming media platforms and media platforms are becoming retailers." P&G mines product and shopper research, social posts, and brand fan sites to inform product innovation and campaigns.
Automotive
'Driven by Intuition' was the world's first advert scripted entirely by AI and shot by an Oscar-winning director. IBM Watson was developed against past award-winning luxury adverts plus data on human emotional responses, and Kevin Macdonald directed the result. It launched the new ES executive sedan in Europe in November 2018.
Apparel Retail
Uses AI in media planning and buying rather than creative production, with CMO Craig Brommers saying the brand avoids AI-generated creative over concerns about generic output. About 40% of the retailer's media business is already supported by AI. The brand produces roughly 500 pieces of content per week.
Digital Publishing
Sells advertising through D/Cipher, an intent targeting tool that matches ads to content signals rather than cookies. The publisher reported D/Cipher deals accounted for half of its direct-sold digital revenue and ran 50% larger on average than deals without it. It has since extended the tool to third-party inventory as D/Cipher+.
Social Platform and Advertising
Upgraded the adtech foundation model behind its Weixin and Tencent Video ad products and launched an automated campaign tool called AIM+. Marketing Services revenue reached RMB 145.0 billion in 2025, up 19% year on year. The company attributes the pricing gains to AI-powered ad targeting and to advertisers using AI to produce more ad variants, noting ad loads stayed below peer levels.
Athletic Apparel
Ran non-branded Performance Max campaigns in Canada and paired them with a measurement setup combining marketing mix modeling, incrementality experiments, and attribution. The work boosted return on ad spend by 8% and cut customer acquisition costs on those campaigns. The strategy won top honors at the Google Search Honours Awards in Canada and was later exported to other markets.
Advertising Agency Group
Built Converged, a cookieless planning and activation system that uses an AI propensity model to score audience cohorts on media behavior, demographics, and psychographics, feeding past campaign results back in as benchmarks. The UK operation ran over 850 campaigns for 41 clients since the February 2024 launch, with average cost-per-click down 47%, cost-per-lead down 44%, and CPM down 9%.
Prediction Markets
Hired creator PJ Accetturo to produce a national commercial entirely with Google's Veo 3, which aired in the YouTube TV stream of Game 3 of the 2025 NBA Finals. The spot cost about 2,000 dollars and took two days, against a comparable traditional production Kalshi put at seven figures and several months. Accetturo ran 300 to 400 generations to get 15 usable clips.
Visual search, dynamic pricing, conversion lift on product pages. Sephora, Walmart, and Amazon have been at it longest, but the long tail of catalog brands is where the recent gains are showing up.
E-commerce
Lens Live lets shoppers point a camera at a real-world object and find matching products, and integrates with Rufus, Amazon's AI shopping assistant, for product insights. It sits alongside AI shopping guides, AI-enhanced reviews, fit tools, and audio product summaries.
Beauty & Cosmetics
Virtual Artist uses ModiFace technology to map facial features through the phone camera and overlay products on a selfie. Cheek Try On added over 1,000 shades of blush, bronzer, contour, and highlighter, plus an AI colour match tool that reads an image, estimates the shade of any product in it, and links to a similar Sephora product.
Beauty & Cosmetics
Beauty Genius runs on GPT-4o through Azure OpenAI inside a dedicated L'Oreal environment, holding maximum latency below 5 seconds and scaling from 100 to 2,000 simultaneous users in testing. It answers beauty questions and returns product recommendations in conversation, around the clock.
Social Commerce
Upgraded visual search so users can find things they struggle to describe in words, like a vibe or an aesthetic. Tapping a Pin triggers visual language models that generate the words explaining why someone likes it, which then narrow the search. It rolled out first in women's fashion in the US, Canada, and the UK.
Fashion Retail
Uses ViSenze for visual search and visually similar product recommendations across its online store, which ViSenze says serves millions of online shoppers and increased shopper engagement.
Grocery Delivery
More than 70% of Instacart customers have at least one dietary preference. Smart Shop uses machine learning to read patterns in past purchases and searches, then generative AI to link related products. Instacart uses large language models to tell an intentional low-carb shopper from someone who happens to buy chicken breast, without relying on manual tagging.
E-commerce Marketplace
Rolled out an AI writing assistant and bulk listing suggestions for sellers, plus AI-generated review highlights for buyers, all optional rather than mandatory. Etsy also drafts replies in a seller's own voice, pointing to its own data showing sellers who answer buyer questions within two days convert 27% higher than those who wait longer.
E-commerce Marketplace
AI fills in product details and category from a seller's photo, cutting the total number of listing steps by half and speeding up listing time. eBay says over 10 million sellers have used its AI features across more than 100 million listings.
E-commerce Platform
Sidekick is a commerce AI assistant built into every store, handling work like picking a theme, adding product photos, and customizing design through conversation.
Home & Furniture
Muse is a Pinterest-style feed where generative AI suggests similar room images and surfaces products matching their attributes. Shoppers can upload a photo of their own room to see how it might look furnished in a given style. Launched February 11, 2025, building on Wayfair's earlier Decorify tool.
Home Improvement
Magic Apron is a proprietary suite of generative AI tools that answers how-to and product questions, provides project guides, and summarizes customer reviews on HomeDepot.com. At launch in March 2025 it was live on millions of product pages and in the mobile app.
Beauty & Skincare
Skin Advisor is an AI-powered tool on Olay.com that walks shoppers through a diagnostic and returns a recommended routine rather than a generic product grid.
Fashion E-commerce
Zalando Assistant reached six million users in 2025, a fourfold increase year over year. Size & Fit avoided more than 8% of size-related returns and foundation-model matchmaking lifted items added to bags by 13%. AI-generated product content went from near zero to 90% in a year, and campaign creation dropped from six weeks to days.
Social Commerce
TikTok Shop runs an algorithm-first model where content, engagement, and timing weigh as much as the listing itself. It generated 100 million dollars of US sales on Black Friday 2024, three times the year before, and revenue rose 113.8% year over year in Q1 2025 as market share went from 23% to 35%.
Beauty Retail
GLAMlab uses augmented reality to let shoppers try on makeup and hairstyles on web and mobile, with thousands of users trying the tool daily. Ulta says try-on users are more likely to purchase than shoppers who skip it, though it does not disclose conversion figures.
Department Store Retail
Ask Macy's is a shopping assistant built on Google Gemini that takes plain-language requests and returns products, styling suggestions, and virtual try-on. It ran as a beta with about half of site visitors before going fully live across the site and app in March 2026. Shoppers who used the assistant spent about 4.75 times more than those who did not.
Online Resale
Runs AI across product search, discovery, ad buying, recommendations, photography, and flaw detection, including natural language search, image search, and a customer service agent called Dottie. On the Q4 2025 earnings call the CEO credited these tools as active buyers rose 30% year over year to 1.65 million. Revenue grew 18% to $79.7 million and orders rose 27% to 1.56 million.
Where the cost reductions are the most concrete. Dollar Shave Club made a campaign for 400 dollars against 4,000 for its 2012 launch video; Too Faced cut its edit stage from four days to one. The pattern is consistent: AI replaces the assembly line, humans still set the brief.
Design Tools
Launched its own foundational design model, trained on its design elements, that generates designs with editable layers and objects rather than flat images. It works across formats including social posts, presentations, and whiteboards.
Design Tools
Figma Make turns a text prompt into an interactive prototype, aimed at ideation and testing how viable an idea is through to final implementation. It sits alongside Figma Sites for marketing pages, sharing much of the same underlying technology.
Creative Software
Firefly generates and edits images, video, audio, and designs using Adobe's own models alongside third-party models from partners including Google, and is integrated across Adobe's creative tools.
Stock Media
Built generative AI directly into the stock library: an AI image generator, Magic Brush for painting in changes by description, Expand Image, Variations, Background Remover, and a Design Assistant. All of it works on generated and stock images alike.
Fashion Retail
Created AI digital twins of 30 real models and ran them in social posts and marketing, using each model's likeness only with their permission. The images went up on H&M's Instagram in July 2025, four months after the brand announced the plan.
Digital Media
Uses OpenAI technology to move past curation into creating personality quizzes that ask users questions and generate written results from their answers. BuzzFeed said AI-created content would move from an R&D stage into part of the core business that year, covering quizzes, staff brainstorming, and content personalization.
News / Media
Wordsmith turned quarterly earnings filings into publishable stories at about 3,000 per quarter, ten times what AP reporters and editors produced before, and with fewer errors than the manual versions. The point was to move reporters off data processing and onto higher-level reporting.
News / Media
Heliograf published around 850 articles in its first year. It started with roughly 300 short reports and alerts on the Rio Olympics, then covered congressional and gubernatorial races on Election Day and D.C.-area high school football games.
Financial Media
Cyborg automates first drafts of company earnings stories, turning filings into publishable briefs within seconds and helping reporters produce thousands of earnings articles each quarter. Roughly a third of Bloomberg News content uses some form of automation.
Website Builder
The AI Website Builder generates a complete site from a conversation about the business, producing copy, images, layout, and product catalog rather than starting from a template.
Productivity
AI Meeting Notes transcribes and writes up calls automatically, and agents take on repetitive work. Enterprise Search pulls answers across a workspace, sitting alongside the knowledge base and docs.
Beauty and Cosmetics
Produced its first fully AI-generated ad for Ribbon Wrapped Lash Mascara using Adobe's Firefly Video Model, trained on existing product imagery to match brand aesthetics. The spot took roughly two weeks across a five-stage workflow. The edit stage dropped from four days to one, and the savings went to influencer partnerships and additional model shoots.
Sports Media
The All England Lawn Tennis Club works with IBM on AI features across the Wimbledon app and website, including Match Chat, an assistant that answers live match questions in Wimbledon's editorial style. Match Chat across Wimbledon and the US Open served about 1 million users with an average response time of 6.25 seconds. The club reported engagement up 16% year over year in 2025 and myWIMBLEDON registrations up 39%.
Short Video Platform
Sells its Kling AI video generation model to commercial creators, with marketing and e-commerce named among its main professional sectors. Kling passed 20 million dollars in monthly revenue in December 2025, an annualized run rate of 240 million dollars, 19 months after launch. As of that month it served over 60 million creators, had generated more than 600 million videos, and counted over 30,000 enterprise users.
Sports and Events
The USTA and IBM deployed Match Chat, Ask The Open, Live Likelihood to Win, article summaries, and an AI-commentated 3D replay system for the 2025 US Open. Matches featuring Jannik Sinner and Daniil Medvedev generated more than 5 million fan interactions during early use of the tools. Match Chat answers open-typed questions on player stats, head-to-head records, and name pronunciations.
Advertising Agency
Built an end-to-end agentic pipeline covering insights, strategy, creative, execution, media, and asset production, and used it to make a 30-second spot for PUMA. Running the workflow on NVIDIA NIM microservices accelerated delivery of agentic workflows by over 2.8 times. Monks also stood up a roughly 50-person agentic AI advisory group.
Advertising Agency
Used Adobe Firefly to generate the goldfish characters for its Fishy.AI campaign for IBM, building brand-safe assets rather than sourcing stock or shooting. The campaign delivered over 600 million impressions. Ogilvy also runs synthetic focus groups made of virtual personas modeled on real consumers to pre-test creative.
Direct-to-Consumer Grooming
The in-house team used Higgsfield and Claude to produce its 250 Years, No BS campaign, keeping scripting and comedy human while using the tools to iterate. The campaign cost 400 dollars to make, against 4,000 dollars for the brand's 2012 launch video, and went from brief to finished in about a week with three days of core creative work.
Less a feature, more how the product works. Spotify's DJ has shaped listening for 94 million Premium subscribers, and the Financial Times rebuilt its paywall around a model that decides who gets asked to subscribe. The brand examples here are about getting closer to that bar without rebuilding the whole stack.
Streaming
Recommendation and search sit at the heart of the product, extending past the homepage into personalized messaging and notifications. Netflix publishes its work on recommender systems, contextual bandits, reinforcement learning, foundation models, and causal inference, with the stated goal of minimising time spent browsing and maximising enjoyment.
Music Streaming
AI Playlists generate a playlist from a written prompt, using large language models to read intent and Spotify's own personalization data to fit the listener. Users refine the result with commands like "less upbeat." It launched in beta on Android and iOS in the UK and Australia in April 2024.
Music Streaming
DJ picks the music and talks between tracks the way a radio host would, reading each listener's taste. Spotify says it has shaped listening for 94 million Premium users since the 2023 launch, and in May 2026 expanded it to over 75 markets, adding French, German, Italian, and Brazilian Portuguese.
Travel & Hospitality
Listing embeddings power Similar Listings and real-time personalization in search. Together, Search Ranking and Similar Listings drive 99% of Airbnb's booking conversions.
Travel & Hospitality
Smart Pricing uses hundreds of factors about a listing and its area to move the nightly rate with demand, staying inside the minimum and maximum the host sets. It is built into every listing at no cost.
Food Delivery
Customized offer carousels on the homepage are built from historical click-through and conversion rates, among other factors, and the checkout screen suggests additions to an order. The update rolled out August 1, 2024.
EdTech
Video Call lets Max subscribers call Lily, an AI character who holds a free-flowing conversation in the language being learned and adjusts to the learner's comfort level. Calls run about a minute for beginners and stretch to three minutes as learners advance, across English, Spanish, French, German, Italian, and Portuguese.
Fitness
Peloton IQ builds weekly workout roadmaps from a member's preferences and goals, estimates performance from workout history, and sharpens its recommendations with every session. It also identifies patterns and writes weekly summaries.
Fashion Retail
Algorithm scores rank the options before a shipment request routes to a human stylist, who makes the final call. The same approach runs new style development, where a model flags attribute combinations with a high probability of being loved and human designers vet the shortlist.
Video Streaming
VP of engineering Cristos Goodrow says recommendations drive a significant amount of overall viewership on YouTube, more than channel subscriptions or search. The system's job is to find the audience that exists for almost any video.
Music Streaming
The Music Genome Project has spent over two decades cataloguing music, identifying hundreds of musical details for each song across genres, decades, emerging artists, and new releases. Pandora calls it the most comprehensive analysis of music ever undertaken.
Grocery Retail
Invested 70 million pounds in the machine learning behind Your Nectar Prices, its personalised pricing and recommendations for more than 18 million digitally registered Nectar customers. The engine generates over 260 million personalised offers every week. Automated health scoring on recommendations produced a 200% improvement in product health scores and a 130 million portion increase in fruit and vegetable sales.
Airlines
Alaska Inspires is a natural language destination search that takes open-ended prompts and returns bookable trip options. The airline reported the tool cuts destination planning time by 75% and converts at 7% against a 5% standard rate. It also reported a 90% guest satisfaction rating, with 87% of users saying they would use it again.
Online Travel
TripGenie turns natural language requests into itineraries and bookings. The company reported that users who plan with TripGenie are 30% to 40% more likely to return to the platform and convert at twice the rate of other users. Those users also spend about 20 minutes longer in the app than average.
News Publishing
Replaced a fixed metered paywall with a model that uses machine learning to decide when to prompt each reader to subscribe. Since launching in January 2025 the FT reported conversion rates up 290% and lifetime value up 7% to 10% among the segments exposed to it. The AI paywall applies to roughly 30% to 40% of the audience.
The category where Klarna's reversal got the most press. Plenty of these are wins, but the lesson is that AI service works best as a layer over a real team, not a replacement for one. Read the sources.
Fintech / Payments
AI handles 2.3 million customer conversations per month, equivalent to 700 full-time agents. Repeat inquiries dropped 25% and the company attributed $40M in profit improvement to the program.
Food & Beverage
Green Dot Assist answers barista questions on in-store iPads in conversation, replacing manual lookups for things like the ingredients in a seasonal beverage. Starbucks announced it in June 2025 and piloted it in 35 coffeehouses.
Financial Services
Erica passed 3 billion client interactions in August 2025, seven years after launch, with nearly 50 million users and more than 58 million interactions a month. Clients have received over 1.7 billion proactive, personalized insights from it.
Aviation
BlueBot, or BB, books tickets through Messenger in conversation and hands off to one of 250 human service colleagues when it cannot finish the job. KLM's social service handles more than 16,000 cases a week across its channels.
Fast Food
FreshAI takes drive-thru orders in natural language, built with Google Cloud. During the pilot it handled 86% of orders without a restaurant team member stepping in, measured as orders completed without intervention. It went live across four company restaurants in the Columbus, Ohio market before wider expansion.
Travel
Smart Filter takes a plain-language description of an ideal stay and applies the matching filters across the whole inventory. Property Q&A answers specific questions by pulling from the listing, traveller reviews, and photos, and Review Summaries condense guest feedback. The AI Trip Planner launched in the US in June 2023 and has since expanded to the UK, Australia, and New Zealand among others.
Travel
Romie, Expedia's AI travel assistant, helps travelers plan, shop, and book through conversation, including joining group chats to summarize trip plans and monitoring disruptions with alternative suggestions. It launched in alpha through EG Labs in May 2024 alongside an AI itinerary builder and AI review summaries.
Aviation
Delta Concierge is an in-app assistant for SkyMiles members covering flight details, seats and gates, baggage tracking, and benefit questions by text or voice, and it hands off to a human agent when needed. The beta opened October 29, 2025 to a randomly selected group of app users, with a phased expansion running through 2026.
Hospitality
RENAI By Renaissance answers guest requests for dining, entertainment, tours, and activities over text or WhatsApp. Recommendations combine ChatGPT and open-source tools with input from hotel staff called Renaissance Navigators, and a human vets every one before it goes out. Announced December 2023 at three US hotels, with expansion to more than 20 properties globally planned.
Financial Services
Eno is a virtual assistant that monitors transactions for fraud and unusual charges, sends spending alerts, and answers account questions through chat or text. Available across the Capital One mobile app and SMS, Eno was one of the first natural-language banking assistants at scale.
Hospitality
Connie, the first Watson-enabled robot concierge in hospitality, was piloted at the Hilton McLean in Virginia from March 2016. It greeted guests, answered questions on amenities, services, and hours, and drew on WayBlazer's travel data for local recommendations. It learned from each interaction.
Home Improvement Retail
Deployed Mylow Companion, an AI assistant built with OpenAI, to associates across more than 1,700 US stores, reaching roughly 300,000 employees. Associates query it from sales floor handhelds for product details, project advice, and inventory. Lowe's calls it the first retail deployment of its kind at this scale.
Banking
Fargo is a virtual assistant inside the bank's mobile app that handles routine tasks such as Zelle transfers, bill payments, and balance questions. In March 2026 the bank reported Fargo had passed 1 billion customer interactions in under three years since its 2023 launch, alongside 33 million mobile active users. More than 3 million Spanish-speaking customers have used it over 160 million times.
Insurance
Uses OpenAI GPT models to draft claims correspondence, with adjusters reviewing and approving instead of writing from scratch. About 23,000 representatives send roughly 50,000 claim communications a day, and nearly all of those emails are now AI-written. Allstate said the AI drafts use less internal jargon than the versions reps wrote themselves.
Insurance
Routes first notice of loss through AI Jim, a claims bot that collects the claim, triages it, and settles simple cases without a human. In its 2025 annual report the company said 96% of first notices of loss were taken without human intervention and roughly 55% of claims were fully automated. Lemonade ended 2025 with 1,282 employees.
Airlines
ConnectionSaver decides whether to hold a departing aircraft for connecting passengers without delaying those already on board. United reported the system has saved more than 3.3 million customer connections since its 2019 launch. A 2025 beta of related app features covered more than 350,000 customers with a 98% connection success rate.
Telecommunications
Built two internal generative AI tools for frontline staff: One Sentence Summary, which condenses a customer's interaction history, and Ask Telstra, which searches internal knowledge bases in natural language. In trials, 90% of employees using One Sentence Summary saved time and worked more effectively, and follow-up contact dropped 20%. Over 80% of Ask Telstra pilot participants said it improved customer interactions.
Telecommunications
Deployed a Gemini-based assistant to its 28,000 customer service representatives, trained on roughly 15,000 internal documents so reps could find answers during calls. Piloted in July 2024 and fully deployed by January 2025, the company said sales through the service team were up nearly 40% since deployment. Verizon reskilled care agents toward selling rather than cutting headcount.
Telecommunications
Replaced its rules-based chatbot with a generative AI virtual assistant across its contact centres. Session abandonment fell from over 40% with the previous chatbot to 9% across nearly four million interactions since 2024. Bell reports the assistant correctly recognizes 97% of customer queries.
Fast Casual Restaurants
Deployed an AI assistant for customers on its digital ordering platform. The company reported a 46% uplift in platform engagement after the rollout. It was the only one of three major chains in the report to disclose a hard performance figure.
The examples that don't fit cleanly into ads, content, or personalization. Usually this is research, planning, or reporting work. It never shows up to the customer, but it cuts how long the team spends getting to a decision.
Sportswear
Uses AI in product design to compress sneaker development from months to days, and cuts prototype waste by testing and refining virtually before anything is physically produced. The reported outcome attached to the program is a 37% sales increase.
Beauty & Personal Care
Marked 20 years of Real Beauty by becoming, in its own words, the first beauty brand to commit to never using AI to represent real people in its advertising. It also published the Real Beauty Prompt Guidelines, guidance on generating more representative images in the most popular generative AI programs.
Financial Services
Mastercard Commerce Media targets offers and content using permissioned insights from the roughly 160 billion transactions Mastercard processed in 2024. The network reaches 500 million enrolled consumers through 25,000 advertisers and their bank and publisher channels, and Mastercard reports up to 22 times return on ad spend across retail, travel, dining, and everyday spend.
Food & Beverage
NesGPT, built on the same technology as ChatGPT, is used across sales, product innovation, marketing, and legal for drafting, proofreading, idea generation, and data analysis. Over 7,000 US employees ran nearly 230,000 prompts in the first three months, and users report saving an average of 45 minutes a week.
Toy & Entertainment
Employs machine learning engineers who build the pattern-recognition and prediction models behind LEGO products, and has shipped products combining machine learning with augmented reality to bring LEGO creations to life. The company tells the story through its engineers rather than published performance figures.
Enterprise Software
Brought predictive AI into CRM workflows with Predictive Lead Scoring and Predictive Forecasting, then layered generative AI for call summaries and outreach drafting on top, all sitting behind a Trust Layer.
Quick Service Restaurants
AnyWare lets customers order through 15+ channels including Alexa, Google Assistant, smart TVs, smartwatches, cars, and SMS. The unified ordering platform turns the ordering interface itself into a marketing surface.
Food & Beverage (CPG)
Partnered with Google Cloud and MightyHive on personalized marketing at global scale: 20M personalized assets across 150 countries, supporting a 10%+ lift in global marketing ROI and 20% in the US.
Beverages
Fuel Light 360 models marketing and commercial investment scenarios in real time, so teams argue about the decision instead of whose data is right. When rivals pushed ad pressure in sugar-free sodas, teams weighed television, digital, and in-store trade-offs at once. A consultant on the project put the decision cycle at two weeks down to one hour.
The lowest-effort, highest-ROI rollout for most teams. Klaviyo, Braze, HubSpot, Yum! Brands. Subject line testing, send-time optimization, segment generation. Almost none of this requires a new vendor.
Marketing Software
Agent Hub, formerly Breeze Agents, is where HubSpot customers build and run AI agents grounded in the customer data already in their CRM, with no separate setup to connect it. Agents are reachable from the Breeze Assistant wherever a team works, priced at 10 cents per answer.
Email & SMS Marketing
Personalized Send Time predicts the best delivery moment for each subscriber instead of sending to the whole list at once. Klaviyo reports a 35% increase in click rates, measured on the top 15% of campaigns using the feature between October and November 2025 against a control group sent at a fixed time.
Enterprise Marketing
Extended Einstein Copilot to marketers and merchandisers for daily marketing and merchandising tasks, alongside tools for unifying business and commerce data and an AI-powered personalization decision engine that tailors interactions using data from any source.
Customer Engagement Software
BrazeAI decisioning picks the message, channel, and timing per user instead of running fixed rules. Kayo Sports reported a 14% increase in subscriptions, a 105% increase in cross-selling, and a 20% rise in average subscription price. Fashion retailer 24S saw a 35% higher purchase conversion rate on abandoned cart campaigns.
Quick Service Restaurants
Built an internal AI factory that generates customer marketing messages across Taco Bell, KFC, and Pizza Hut, drawing on a first-party data program holding over 140 million permissioned customer records. The company sent more than 200 million AI-powered communications, which it reports were up to five times more effective than traditional approaches. Taco Bell hit a record 41% digital sales mix in Q2 2025.
The internal-rollout category. JPMorgan's LLM Suite, Microsoft and EPAM's Copilot programs, Amgen's phased rollout, Walmart with STRIVR. These are the case studies enterprise leaders ask me about most.
Financial Services
Rolled out its internal LLM Suite assistant to 140,000 employees, announced by President and COO Daniel Pinto in September 2024, alongside assessments of how to optimize "every single process" with large language models. Pinto raised the bank's projected AI value to closer to 2 billion dollars, much of it tied to fraud prevention.
Technology
Microsoft was the first enterprise to fully deploy Microsoft 365 Copilot, and published its own playbook. Microsoft Digital drives usage and adoption to more than 300,000 employees and vendors worldwide, running it as Customer Zero with governance settled first and employee champions inside each cohort.
IT Services
Measured Copilot impact across 2,000 employees over a 6-month period. The data-driven adoption journey reported a 20% reduction in external collaboration hours.
Biotechnology
Phased generative AI rollout to ~20,000 employees from 2023 to 2025. Built governance and testing rigor between waves; tracked qualitative and quantitative adoption metrics through every expansion.
Fintech / Payments
Deployed AI agents on inbound queries through 2024 and progressively cut reliance on human agents for first-contact resolution, eliminating roughly 700 customer service roles. Customer satisfaction forced the reversal, and Klarna moved to a hybrid model where AI handles tier-one volume and humans keep escalations and judgment calls.
Retail
Used STRIVR's AI-driven VR platform to train associates on store operations and customer scenarios. 15% improvement in employee performance, 95% reduction in training time, higher participation than legacy training.
Consumer Packaged Goods
Runs AI pilots as proofs of value, with human-only control groups running alongside AI-augmented workflows, using OpenAI and Gemini models through Databricks rather than training its own. Creative asset adaptation runs 30% faster across its 700-person marketing team. Brand performance reports combining Nielsen and other datasets now take about 15 minutes, saving marketers 20% to 40% of their time.
Grocery Retail
Completed the rollout of AI.Carrefour, an internal assistant built on Google Gemini, across eight countries including Spain, Poland, and Romania. The tool reached 125,000 employees. Carrefour reported employee feedback as fast, useful, and intuitive, and disclosed no productivity figures.
Grocery Retail
Runs Sage, an internal generative and agentic AI platform where associates check schedules, request time off, view pay stubs, and ask policy questions, while store leaders get real-time labor data. Sage has up to 150,000 active associate users and has grown from about 30 use cases at its mid-2024 launch to more than 95.
Beauty and Cosmetics
Built ConsumerIQ, a Copilot Studio agent that centralises consumer data across 25 brands and 150 countries so employees can query surveys, clinical trials, and product usage data in natural language. A second agent, Trend Studio, uses Azure OpenAI to detect market trends and generate tailored marketing copy. Microsoft reports answers return in seconds rather than hours, with no performance figures disclosed.
Wealth Management
Built AI @ Morgan Stanley Assistant on OpenAI models to answer advisor questions against its internal research library, then added Debrief to summarize client meetings and write follow-up notes into Salesforce. The firm's chief analytics and data officer reported that 98% of financial advisors actively use the assistant.
Commercial Insurance
Gave US Middle Market underwriters Sixfold, a generative AI tool that assembles and summarizes risk information before the underwriting decision. In the first year, 80% of submissions handled by 16 underwriters were processed with it, saving an estimated 60 minutes per submission. Zurich launched it in four offices in January 2025 and is expanding countrywide.
Hospitality
Applied AI to group sales, where the company fields more than 1.5 million corporate RFPs a year, so salespeople start from a drafted response instead of building each one manually. CEO Mark Hoplamazian said the tools freed up a full day a week per salesperson and that Hyatt grew market share in group business.
Telecommunications
Rolled out Microsoft 365 Copilot to 68,000 of its roughly 100,000 employees after an initial trial run with Microsoft and KPMG. The trial measured around three hours saved per person per week on drafting, meeting summaries, and internal search, and the legal and compliance team found drafting a new contract took an hour less than usual. Ninety percent of trial users said they benefited and wanted to keep using it, and 60% said it improved the quality of their work.
Enterprise Networking
Built an internal AI assistant for its own workforce and reports it is now the most used AI tool at the company. As of November 2025 it had served over 100,000 users and handled more than 45 million interactions, averaging 156,000 per day. Users reported average time savings of five hours per week, and 73% reported increased productivity.
Industrial Manufacturing
Extended its Industrial Copilot to maintenance work, giving technicians a generative AI assistant for diagnosing and resolving equipment faults on the shop floor. First pilot deployments saved an average of 25% of reactive maintenance time. The offering builds on the Industrial Copilot already used across engineering and operations.
Pharmaceuticals
Rolled out plai, an internal AI decision-support app built with Aily Labs, across R&D, manufacturing, and supply chain. The company reports 20,000 employees use it every day. In supply chain it predicts 80% of stock disruptions using probabilistic planning and correlates 65% of those risks with a root cause.
Enterprise Technology
Applied AI agents to more than 70 of its own business areas across a workforce of 270,000 and reported 3.5 billion dollars in productivity impact over two years. Its AskHR agent automates 94% of simple requests such as vacation bookings and pay statements. The AskIT agent cut calls and chats to the IT support team by 70%.
Healthcare
The Permanente Medical Group deployed an ambient AI scribe that listens to patient visits and drafts clinical notes, reaching 7,260 physicians. Across more than 2.5 million patient encounters between October 2023 and December 2024, users saved an estimated 15,791 hours of documentation time compared with non-users. In surveys, 84% of physicians said it improved patient communication and 82% said work satisfaction improved.
Meta Advantage+, Google Performance Max, TikTok Symphony. The platforms doing the AI work for advertisers, often without naming it. These are already in your stack if you spend a dollar on paid.
Social Media
Symphony Creative Studio added Image to Video, which turns product photos and brand assets into branded video, Text to Video for generating and testing concepts without visuals, and Showcase Products, which puts products in front of Symphony Digital Avatars holding them, wearing them, or displaying an app.
Social Media
Meta AI restyle edits Instagram Stories from a text prompt: remove objects, add elements, change an outfit or a video background, or apply preset effects. Users can push restyled edits out through Add Yours stickers to start a trend.
Social Media
Machine learning built on Reddit's 23 billion plus posts and comments powers ad targeting keyed to community conversations, including AI-enhanced contextual keyword targeting. Reddit's automated Max Campaigns cut cost per acquisition 17% and lifted conversions 27% on average in early split tests.
Streaming & Advertising
Magic Words analyzes scenes across Disney's library to tag content, brands, and emotional tone, then lets advertisers buy against the mood of a moment rather than a demographic. It runs on Disney+ and Hulu, and went into beta with six agency groups: Dentsu, Omnicom, Horizon Media, GroupM, IPG Mediabrands, and Publicis Media.
Social Media
AI Sponsored Snaps put brand AI agents inside Chat, the surface where Snapchatters sent over 950 billion messages in Q1 2026. More than half a billion Snapchatters have messaged My AI since launch. Snap reports the format drives 22% more conversions at nearly 20% lower cost per action, and twice the conversions per full-screen ad view versus other inventory.
Less marketing, more operations, but it shows up in margin and product availability. PepsiCo, Unilever, and Walmart on demand forecasting and logistics.
Retail
Route optimization technology packs trailers efficiently and minimizes miles travelled. Walmart eliminated 30 million unnecessary miles driven, avoided 94 million pounds of CO2, and bypassed 110,000 inefficient paths, work that won the Franz Edelman Award in 2023.
Retail
The Inventory Ledger handles up to 360,000 inventory transactions per second at peak and serves as many as 16,000 requests for inventory positions per second, built to scale further.
Fashion Retail
Applies AI across the chain from design and production through to pricing, using predictive intelligence for real-time demand forecasting and inventory management and RFID for garment-level visibility. The Inditex-owned brand reports 7.1% year-over-year sales growth off the digital transformation work.
Food Delivery
Every machine learning use case at DoorDash runs through Sibyl, its prediction service cluster: search and recommendation, fraud detection, dasher dispatch optimization, and delivery time prediction. It is deployed as a Kubernetes service on AWS with a Kotlin gRPC server in each pod.
Automotive
The AIconic Agent is a generative AI multi-agent system for procurement and supplier networks. It builds on Knowledge Navigator, Offer Analyst, and Tender Assistant, the GenAI tools BMW Group Purchasing introduced in 2024 as central knowledge databases for analysis.
The smallest category here, but the one with the longest track record and the most measurable savings. Visa, Stripe, HSBC, Commonwealth Bank. AI has been load-bearing in this space for over a decade.
Payments
A payments foundation model turns transactions into embeddings used to make real-time predictions. Authorization Boost's retry logic recovered more than 6 billion dollars in legitimate declined transactions for Stripe users in 2024, a record. Card testing attacks are up across the industry but down 80% on Stripe.
Fintech
A scam detection feature assesses whether a customer is being coached into making a card payment as part of a scam and declines it, then walks them through an in-app intervention with educational content and access to a fraud specialist. Revolut reported a 30% drop in fraud losses from scams where money was sent for investment opportunities.
Payments
Risk Manager lets sellers write rules that decline a payment, trigger a risk alert, or invoke 3D Secure when conditions are met. Rules can key off Square's own risk evaluation, an AVS mismatch, or an invalid CVV.
Payments
Reported that its AI-based security systems identified and blocked 144% more suspected fraudulent activity over the 2025 Black Friday and Cyber Monday weekend globally than over the same weekend a year earlier. Visa attributes the capability to 13 billion dollars in technology investment over the prior five years.
Banking
Deployed an agentic AI system that detects emerging fraud patterns in transaction data and writes the detection rules to intercept them. It monitors more than 80 million signals and 20 million payments a day and sends about 40,000 proactive customer warnings daily. The bank reported fraud losses fell over 20% in the first half of the 2026 financial year, and the agent contributed to three quarters of its card fraud rules.
Banking
Replaced rules-based anti-money-laundering screening with an AI system that flags suspicious activity without being told what patterns to look for. The bank reports it now finds two to four times more financial crime than the previous system while producing 60% fewer false positives. Analysis of billions of transactions dropped from several weeks to a few days, against about 980 million transactions screened monthly.
Across the 147 examples in the database, the work clusters into a few patterns that show up over and over:
For a deeper look at the operating model behind these rollouts, see the AI Enablement & Transformation Hub.
A handful of brands keep showing up across the documented examples, either because the work is unusually good or because they've put enough numbers on it for the rest of us to learn from. The short list I'd send to a CMO trying to benchmark:
| Brand | What they're known for | Reported result |
|---|---|---|
| Klarna | AI customer service | Replaced 700-agent equivalent, then publicly reversed course |
| JPMorgan Chase | Internal LLM Suite + Persado for ad copy | Click-through lift as high as 450% in the Persado pilot; LLM Suite rolled out to 140,000 employees |
| Coca-Cola | Generative AI in advertising | AI-remade Christmas film passed its own and competitors' Christmas ad benchmarks |
| Adidas | Personalized email creative | Sneaker design cut from months to days; 37% sales increase |
| Nestlé | NesGPT for drafting, analysis, ideation | 45 minutes saved per employee per week; 7,000 users in three months |
| Sephora | Visual search + recommendations | AI-driven personalization across app, web, and in-store |
| Spotify | AI personalization as product | DJ has shaped listening for 94 million Premium users |
| Heinz | DALL-E brand association campaign | Swept The Drum Awards Festival 2023; built in under a month |
| Walmart | STRIVR for training, AI for supply chain | 15% improvement in employee performance; 95% less training time |
| Meta & Google | Ad platform automation | Advantage+ ~22% ROAS lift; Performance Max ~13% more conversions |
Filter by the Customer Service, Content & Creative, or AI Enablement chips above for the full list per category.
Three rules I held myself to while building this list:
The list is reviewed and re-verified roughly every 90 days. Each card shows the date of the most recent check. If you spot a number that looks off, or a brand that should be on this list and isn't, send it over.
The database above includes a dedicated AI Enablement filter with documented enterprise rollouts: JPMorgan Chase's LLM Suite, Microsoft and EPAM's Copilot programs, Amgen's phased gen AI rollout, Klarna's reversal, Walmart with STRIVR, and more. Each entry links to the primary source.
For the free tools and frameworks I use when running enablement programs, see the AI Enablement & Transformation Hub. Or get in touch directly.
Retail and e-commerce are way out in front. They've had product catalogs, recommendation engines, and customer data infrastructure for years, so AI slotted in fast. Food and beverage is the next tier (Coca-Cola, Heinz, Cadbury, Nestlé, PepsiCo all show up across this list), and financial services follows close behind on both customer service and fraud detection. The pattern across all three: brands with very high customer interaction volume see payback fastest. The results worth citing tend to cluster around three things, regardless of industry: personalization, ad optimization, and content production. The AI for Brands hub has more context on where each sector actually stands.
Depends on the use case, and being specific matters here because the numbers vary by an order of magnitude. Ad platform automation is real but not dramatic: Meta puts Advantage+ at a 32% increase in return on ad spend, and LinkedIn puts Accelerate at 52% lower cost per action. The bigger numbers tend to show up in content production: Dollar Shave Club made a campaign for 400 dollars against 4,000 for its 2012 launch video, and Too Faced cut its edit stage from four days to one. Internal enablement is the harder one to put a number on, but JPMorgan's LLM Suite reached 140,000 employees and Nestlé says NesGPT saves people an average of 45 minutes a week. Personalization is a different situation entirely. Spotify says its DJ has shaped listening for 94 million Premium subscribers. That's less a “result” and more just how their products work now.
Pick the thing your team touches most often and start there. Email personalization and ad creative testing are usually the safest first bets: measurable, reversible, and you don't have to rebuild anything to try them. Once you have a number to point at, expanding into harder use cases (segmentation, predictive lifetime value, personalized journeys, in-app experiences) gets a lot easier because the conversation moves from “should we try AI” to “here's what AI did for us last quarter.” The mistake I see most often is starting with the most impressive use case instead of the most useful one. Generative video for a brand campaign is exciting; subject line testing is what compounds. If you'd rather skip the trial and error, that's what the AI enablement work is for.
Start with what's already in the tools you pay for. Meta Advantage+, Google Performance Max, Klaviyo or Mailchimp's AI features, Canva Magic Design. These are already on for most teams and genuinely useful. No new vendor, no procurement process. The next layer up, also low-friction: a paid ChatGPT, Claude, or Gemini seat for the team, plus Notion AI or whatever's native to your wiki. That covers research, drafting, and reformatting work and pays for itself the first week. The advanced layer (Persado for ad copy at scale, Jacquard for CRM, internal LLM platforms) requires a real procurement cycle and shouldn't be the first move. For a broader list, the free AI tools directory covers more ground, including a marketing text generator built for brand teams.
Yes, and most teams figure this out after something goes wrong. If you're making customer-facing content, or you're in finance, health, or legal, the stakes are obvious. But even in lower-risk categories, a basic policy saves a lot of awkward conversations. The main things to nail down: what can go out without a human reviewing it, what disclosure language you use when AI is involved, what data your AI tools can actually see (this is the one most policies skip), and who owns the output if a contractor used AI to make it. Write a v1 and plan to revise it. The first version is going to be wrong in small ways and you'll learn what to tighten the first time someone asks “wait, can I do this?” The free AI policy generator covers all of that as a starting point.
Marketing Strategist & AI Expert