Need to Know News - July 24th, 2026
In this week's Need to Know News edition:
🤖 Google's flagship model still isn't ready, so it put out three leaner, lower-cost Flash models instead... aimed at anyone running AI at scale.
🤖 Synthesia AI now puts your sales reps in a mock pitch with an AI that challenges every point... then scores each run for their manager.
🤖 If you're using AI to churn out high volume content... a new TikTok study of 400 marketers found the leaders are playing a different game.
And a whole lot more!
90% of Workers Want Creative AI. Only 9% Actually Use It
A new Adobe survey turned up a gap almost comic in its size: 90% of workers say they want creative AI tools at work, while only 9% report using any. Most who do use AI for technical chores like summarizing documents, and leave the creative work they claim to want untouched.
📋 The Details: The survey leaned on full-time US employees who already use AI weekly, so if anyone's reaching for creative AI, it should be them. Even they park AI on technical work and figure they lose 7.5 hours a week to tasks creative AI could take off their plate. One in three would hand image and video jobs to it... they just haven't.
🎯 Why You Need to Know: If your team keeps AI penned into research and cleanup while the creative work stays hand-done, you're sitting in the 91%, and the survey says the appetite to change that is already in the room.
📡 Watch For: Whether that 9% climbs as creative AI lands inside the apps people already open. Interest isn't the holdup, and next year's adoption line tells the story.
Disney Opens an AI Ad-Maker for Connected TV, in Closed Beta
Disney is handing smaller advertisers a door into connected-TV ads, a room once gated by production budgets. It kicked off a closed beta of Disney Ad Creative Studio, a tool inside Disney Campaign Manager that turns a brand's logos, product shots, and guidelines into streaming video ads.
📋 The Details: A select group is in first, letting Disney pressure-test it on real workflows before a wider release. Under the hood it strings several AI models into one pipeline, spinning brand materials into ad variations for different audiences and goals. It's aimed mainly at small and mid-sized businesses, though people still steer the approvals and the guardrails around likeness and IP.
🎯 Why You Need to Know: Connected-TV has been hard for a small brand to make creative for, and Disney is putting the ad-building inside its own campaign tool. If it opens up, a local shop could run streaming spots without booking a production crew.
📡 Watch For: The jump from closed beta to open access, and what Disney charges. That's the moment CTV creative stops being an agency-scale project for businesses shut out until now.
🚀 WATCH: How These AI Copy Bots Are Producing World-Class Sales Copy 50X Faster Than Even The "BEST" Copywriters On The Market…
(Plus… They Don't Get Sick, Miss Deadlines, Or Ask For Raises Either!)
Watch the full AI Copywriting Tell-All Video Here
Claude's Voice Mode Can Now Think Through Your Hard Problems
Some problems you can't type your way through. Claude's voice mode ran only on the quick Haiku model, fine for speed but shallow on anything knotty. Now Opus and Sonnet run in voice too, so you can talk out a half-formed idea and have Claude question it back instead of handing you an answer.
📋 The Details: It's in beta today for every chat user on mobile, desktop, and web, though it works best from your phone. And it reaches your connected tools now... tell Claude to push a Calendar meeting, turn a pitch into a Canva one-pager, or triage the day's email.
🎯 Why You Need to Know: An assistant that questions you out loud changes what a commute is good for. You can work a pricing call, get gaps poked in your logic, and land on what you really think before opening a doc.
⚡ Your Move: Next time you're stuck on something you'd normally chew over alone, open voice mode and talk it through, then ask Claude to name the weak spots in your reasoning. Free users get Haiku and one tool to start.
TikTok's New Report: Winning Brands Learn Fastest, Not Loudest
Pumping out more content is fast becoming the losing move. A new report from Warc and TikTok, drawn from 400 marketers across four countries, found the brands winning now are the ones learning fastest from the people they serve... reading what a community values while it's still forming.
📋 The Details: Most marketers in the report agreed AI has settled into the creative process, good for making assets faster. But they rated it weakest on copywriting, scripting, and real originality, and worried about sameness. Warc calls the missing piece "cultural intelligence," the read on what's catching on that a model can't fake.
🎯 Why You Need to Know: The report says a volume-first AI plan is pushing on the thing that no longer sets you apart.
⚡ Your Move: Before scaling another batch of AI assets, spend an hour on where your best content landed and why, then feed that read into the briefs. A learning loop beats an output count.
Michaels Built a Shopping AI in Six Weeks. It's Had 75,000 Chats
Michaels went from idea to a live shopping assistant in six weeks. Called Ask Mike and built on Google Cloud's Gemini Enterprise, it lets shoppers describe a project in plain words, "help me plan my kid's birthday party," and hands back curated picks instead of a keyword search. Since May, it's fielded 75,000 conversations.
📋 The Details: More than 60% of those chats are people hunting for products, treating Ask Mike less like a search box and more like a creative sidekick. It runs on Michaels.com and the mobile apps, with product-page overviews planned next. And Google Cloud's retail lead called six-weeks-to-production a signal of where the industry's headed.
🎯 Why You Need to Know: The six-week build is the part to sit with. A conversational storefront reached live commerce faster than most teams schedule a rebrand, which resets what "we don't have time to build that" means for a mid-sized retailer.
⚡ Your Move: Run a messy, plain-language request through Ask Mike on Michaels.com and watch how it answers, the cheapest way to feel conversational commerce before you pitch it internally.
Synthesia's AI Avatar Now Roleplays Your Sales Pitch, Then Grades It
Synthesia made its name turning scripts into slick training videos. Its newest product chases a different prize: proof the training changed how someone works. Roleplay Sessions lets you rehearse a tough conversation, a sales pitch or a customer complaint, against an AI avatar that pushes back and scores you on a rubric.
📋 The Details: The reasoning under the avatar is OpenAI's, while Synthesia handles the face, the voice, and the rubrics that turn a practice run into performance data a manager can read. It's enterprise-only for now, though the CEO says small businesses and maybe schools come next, within a few months.
🎯 Why You Need to Know: This is where enterprise AI is drifting, from "look what it can generate" toward "show me it moved a number." A tool that scores your team's practice reps hands managers data on who's ready and who isn't, a different sell than a content generator.
📡 Watch For: When Roleplay reaches smaller users, the window the CEO put a few months out.
Coca-Cola Is Turning Brand Consistency Into Software
Coca-Cola is refreshing its look across 200-plus markets, tightening the red-and-white, the Dynamic Ribbon, and the Spencerian script so everything reads as Coke at a glance. The telling piece sits behind the paint: new AI "design intelligence" tools that check work against the brand system before it ships.
📋 The Details: Coca-Cola Zero Sugar gets the clearest facelift: bigger "zero sugar" type and a black Dynamic Ribbon. The refresh is reaching Latin America, Europe, the Middle East, and Asia, but the operational move matters more, pairing the look with a brand center and AI tools to hold governance together as creative gets made market by market.
🎯 Why You Need to Know: When every channel is a production channel, brand consistency becomes a workflow job a brand book can't hold, so Coke is wiring the rules into software that checks assets before they ship. That load grows faster than any team you'd hire.
⚡ Your Move: Pull the three or four assets that carry your brand, the logo, the color, the signature element, and ask whether a stranger could apply them right without calling you. That gap coul be your consistency problem.
OpenAI Introduces Presence
OpenAI put its money where its agents are. Its new enterprise product, Presence, already runs the company's English phone line, and it closes out 75% of inbound issues with no human in the loop. Within 10 days, its self-improvement loop had trimmed human handoffs by 15 percentage points.
📋 The Details: Presence is built for production, not demos. Each agent gets only the access its job needs, so it can answer questions, tap your systems, and take approved actions, then hand off to a person at a wall. BBVA, SoftBank, and IAG are already on it, though getting in means a limited rollout run by OpenAI's own engineers... no self-serve yet.
🎯 Why You Need to Know: The selling point, aimed at anyone weighing a support or sales agent: the hard part now is reliability, and OpenAI is pushing the guardrails and evaluation scaffolding that get an agent there.
📡 Watch For: The open question is whether that 75% holds outside OpenAI's tightly run line, and when Presence loses the white-glove rollout for something you can just buy.
Google Admits It's Behind on AI Coding, Bets on Gemini 4
Google said the quiet part on its earnings call. CEO Sundar Pichai admitted the company still trails OpenAI and Anthropic on AI coding and agentic tasks, even as he waved off the idea Google is falling behind. His answer has a name: Gemini 4, which he called Google's most ambitious yet.
📋 The Details: The admission sits next to a delay: Gemini 3.5 Pro reportedly slipped after missing internal targets on coding and autonomous tasks, where Codex and Claude Code have pulled ahead. So rather than wait on Pro, Google rolled out cheaper Flash models. The money kept flowing regardless... Gemini now counts 950 million monthly users, and Google Cloud revenue jumped 82% to $24.8 billion.
🎯 Why You Need to Know: The Google model you'd want for coding or agent work is the delayed one, and the company itself concedes it's not caught up there. If you're picking tools now, test rivals rather than hold out for Gemini 4.
📡 Watch For: Whether Gemini 4 closes the gap Pichai owned up to, or 3.5 Pro finally ships first. Google's timeline on both is the tell.
Google's New Flash Models Are Cheaper and Faster to Run
While the flagship keeps testing, Google shipped the workhorses: Gemini 3.6 Flash, 3.5 Flash-Lite, and a security-tuned Flash Cyber, all for running AI agents at scale on the cheap.
📋 The Details: 3.6 Flash is the everyday model: it codes better than 3.5 Flash while burning 17% fewer output tokens, at $1.50 per million in and $7.50 out. And Flash-Lite, the cheaper speed pick, runs at 350 tokens a second. Both are live now in the Gemini app, the API, and Gemini Enterprise.
🎯 Why You Need to Know: Anything you run at volume through Gemini, drafting, tagging, summarizing, gets cheaper per task here, which beats any benchmark chart. Fewer tokens for the same job trims the cost of every automation you've built.
⚡ Your Move: Point a live Gemini workflow at 3.6 Flash and watch the token count. Google says it runs leaner, so the same job should cost less untouched.
🧵 The Thread: These Flash models are what Google put out instead of the delayed 3.5 Pro... the coding gap Pichai copped to on the earnings call, met with efficiency instead of a frontier leap.
Thanks for reading.
Until next time!
The AI Marketers
P.S. Help shape the future of this newsletter – take a short 2-minute survey so we can deliver even better AI marketing insights, prompts, and tools.
[Take Survey Here]
