👋 Hey,
Welcome to the latest edition of The AI Driven Marketer!
Every week, I cut through the noise to bring you the most practical marketing AI news, tools, and a battle-tested automation workflow you can plug directly into your business.
Let's get into the build.
⚡ The AI Pulse: What Marketers Need to Know
OpenAI Previews GPT-5.6 Sol, Terra, and Luna, Undercutting Claude by Half
On June 26, OpenAI previewed its new model family in three tiers: Sol (the flagship), Terra (balanced for everyday work), and Luna (fast and cheap). Sol is priced at $5 input / $30 output per million tokens, roughly half the cost of Anthropic's Claude Fable 5 at $10 / $50, with Terra at half of Sol and Luna at less than half of Terra. OpenAI plans to make all three generally available in the coming weeks.
So What? The story here is not the model, it is the price. When the most capable models drop to half the cost, the AI work you do every day (drafting, research, ad variations, support replies) gets cheaper to run at volume. Budget for a step down in your per-task AI costs, and decide now which high-volume jobs you will scale the moment the cheaper tiers open up.
OpenAI limits its own GPT-5.6 rollout after a government request, and says it shouldn't become the norm
Image Credit: The Verge
Also on June 26, OpenAI confirmed it narrowed the release of its GPT-5.6 lineup (Sol, Terra, and Luna) to a small group of trusted partners at the Trump administration's request, following the same pattern set by Anthropic's suspension two weeks earlier. OpenAI was direct about its discomfort with this: "We don't believe this kind of government access process should become the long-term default... It keeps the best tools from users, developers, enterprises, cyber defenders, and global partners who need them."
So What? Two of the three major US labs have now had a flagship model's release shaped by a government access request inside the same month. That's a pattern, not a one-off. If your workflow depends on having the newest model the day it ships, it's worth having a fallback model and a fallback vendor mapped out before you need one. Worth reading the full bonus section below if this is the first you're hearing about either restriction.
Anthropic's Claude Tag Turns Slack Into a Place You Delegate Work to AI
Announced June 23, Claude Tag lets you add Claude to chosen Slack channels and connect it to your tools, data, and even your codebase. Anyone can tag @Claude to hand off a task, and it works in the background over hours or days, building context from the channels it sits in. Anthropic says 65% of its own product team's code now comes from an internal version of the same tool. It is in beta for Claude Team and Enterprise plans and replaces the older Claude in Slack app.
So What? This moves AI from a tab you open to a teammate you tag. For a small team or a solo founder working with contractors, that means handing off research, data pulls, and first drafts inside the tool where the conversation already lives. The skill that pays off now is writing a clear brief in a Slack message, the same way you would for a junior hire.
Google Puts Computer Use Into Gemini 3.5 Flash, So AI Can Operate Your Apps
On June 24, Google built computer use directly into Gemini 3.5 Flash, letting agents see a screen and act across browser, mobile, and desktop, clicking buttons and filling in forms. It is available to developers through the Gemini API, with optional safeguards that ask for confirmation before sensitive actions and stop a task if a prompt injection is detected.
So What? Computer use is how AI starts doing the click-heavy work no API ever reached: pulling numbers from an ad dashboard, updating a CRM, reconciling two tools that do not talk to each other. It is developer-facing today, but it is the engine behind the "AI does my busywork" products you will buy next quarter. Worth watching which of your manual workflows turn into a one-line request.
Canva Grow 2.0 Turns One Platform Into Your Whole Ad Workflow
Unveiled June 26 at Cannes Lions, Canva Grow 2.0 brings ad creation, publishing, and performance tracking into one place. It generates static and video ads from your brand context, publishes them to Meta, TikTok, and LinkedIn in a single workflow, tags your creative so you can see what is driving results, and feeds real performance data back into fresh ad ideas.
So What? For a solo founder or a lean team, this collapses the create, publish, and measure loop that used to need three or four separate tools. The advantage goes to whoever ships and tests the most ad variations, and a single workflow makes that far less painful. If you already work in Canva, your ad operation just got a lot shorter.
TikTok's Symphony Agent Builds Whole Video Campaigns From a Prompt
Also launched at Cannes Lions (June 22–23), TikTok's Symphony Agent generates entire video campaigns from a text prompt, guided by an analysis of what is trending on the app. It builds example videos with ByteDance's Seedance 2.0 model, creates image and avatar assets, dubs videos into multiple languages, and can even shortlist creators to pitch your brand.
So What? TikTok is handing marketers a trend-aware video studio that runs without a camera. The risk is everyone's ads starting to look the same, so the edge stays with the human input: a sharp hook, a real point of view, and a specific offer. Use the agent for speed and volume, and keep the idea yours.
🛠️ The AI Toolkit
3 Tools specifically curated for digital marketers.
1. ElevenLabs
Best for: Creators and marketers who need professional voiceover, narration, or dubbing without a recording studio.
What it does: Turns text into natural-sounding speech in many voices and languages, clones a voice, and dubs existing videos, so you can add narration to Reels, ads, and explainers in minutes.
2. Profound
Best for: Marketers and founders who want to know how AI engines describe their brand, and improve it.
What it does: Tracks how ChatGPT, Perplexity, Gemini, and Google AI answers mention and cite your brand against competitors, then shows what to fix so the engines recommend you more often.
3. Motion
Best for: Solo founders juggling more tasks than hours who want their week planned for them.
What it does: Takes your to-dos, meetings, and deadlines and builds your daily calendar automatically, then reshuffles it when something moves, so your most important work always gets a time block.
Bonus
Loop Engineering for Claude Code (Without Budget Surprises)
I published a full guide on this a few days back. This is the short, paste-ready version. Check the full article here for more details: Loop Engineering: Your Claude Code /loop and /goal Setup Guide.
For two years, getting work out of an AI agent meant the same routine: write a prompt, read what came back, write the next one. You held the tool the whole time. That part is ending. Boris Cherny, who created Claude Code, put it plainly: "I don't prompt Claude anymore. I have loops that are running. My job is to write loops." Days later, Peter Steinberger, who stewards the OpenClaw project, said the same thing: "You shouldn't be prompting coding agents anymore. You should be designing loops that prompt your agents."
So here is the plain version: what a loop is, the two commands that ship inside Claude Code today, and the one billing setting that stops a runaway loop from eating your budget while you sleep.
💡 Key insight: A prompt is one turn you babysit. A loop is a goal you set once that runs, checks itself, and corrects until it hits the finish line you defined.
The two loop commands (built into Claude Code today)
/goal: run until done. You set a finish line, and Claude keeps working toward it across turns. After each turn, a small fast model reads the conversation and answers one question: is the condition met? A "no" comes back with a reason, and that reason becomes the next instruction. A "yes" ends the loop. (Shipped in v2.1.139.)/loop: run on a schedule. You give it an interval and a prompt, and it fires on that clock until you press Esc. This is the one for watching for a change and reacting to it. (Shipped in v2.1.72.)
Pushing work to a finish line is a /goal. Watching for a change is a /loop. Point one at the other's job and you either burn turns on finished work or spin forever on something the agent cannot move.
The 4 ingredients of every loop
Goal: the finish line, written so the agent's own output can prove it. "Every product description is under 150 words." "Tell me the moment the deploy goes green."
Context: the agent forgets between runs, a file does not. Keep the working state in a file the loop reads on each pass.
Evaluation: the check that decides pass or fail after each turn. This is the part people skip, and it is what makes a loop trustworthy. A passing test or a clean build is a fact. A model saying "looks done" is an opinion. Prefer facts.
Agent: the model doing the work. Get the first three right and almost any capable model will run the loop.
The billing guardrail: set two ceilings, every time
A loop with no ceiling is an expensive way to find out you left it running. Set both of these, every time:
A turn ceiling, inside the goal. Add a stop clause to the condition itself, so a goal that can never be met does not run forever. For example: "Stop after 15 turns if the build still fails."
A dollar ceiling, at the account level. Set a hard spend limit in your Anthropic Console usage settings, so no single loop can run past a set amount no matter what. This is your backstop.
💡 Key insight: Start with one simple loop, make sure it can stop itself, and put a dollar ceiling behind it before you walk away. Set the goal once, let it check and correct its own work, and cap the downside so a forgotten loop never turns into a billing surprise.
The full guide goes deeper on verification design and the tools beyond Claude Code. 👉 Read the full article here.
That's a wrap for this week in AI.
See you next week,
Rananjay
P.S. If you found this valuable, share it with a fellow marketer who's trying to make sense of AI. And hit reply to let me know—what's the one AI workflow you want to automate this quarter?