AI in Marketing

Stan Berteloot is the Chief Innovation Officer at Nytro Marketing. The thirteen agents he built aren't a replacement for the firm. They're a research project running inside it, and a personal productivity experiment at the same time.
The question he's working on is one the whole industry is circling without quite landing on: what does a marketing agency actually look like when AI can handle the specialized, repeatable work? Stan is building toward an answer rather than waiting for one.
His setup runs on AWS Lightsail. Each agent is scoped to a specific discipline: cold email sequencing, LinkedIn content strategy, campaign analysis, and others. Stella Chen, a Claude-powered agent, functions as team lead, coordinating the specialists and reporting directly to Stan. The structure mirrors how a real team works, which is partly the point.
Pierre Dumont, his AI executive assistant, monitors six email accounts on a two-hour cycle, triages the volume, replies when pre-authorized, books meetings, and escalates anything that actually needs Stan. For a CIO doing research and client work simultaneously, that kind of triage isn't a convenience. It's what makes the rest of the work possible.
Governance runs through what Stan calls a Constitution: a tiered permission system that classifies every agent action by risk level. Reading a record is automatic. Sending a client email requires confirmation. Launching a campaign requires a full stop. The system connects Apollo, HubSpot, Toggl, and Google Workspace, and audits its own health weekly without being prompted.
If you want a sense of what this looks like in practice, Pierre is reachable directly. Call or text +1 (551) 290-5742. He'll give you a firsthand preview of what's feasible.
What Stan is building is an argument, made in working software, about how a firm like Nytro operates going forward. Not smaller. Not cheaper. More capable per person, with humans focused where judgment actually matters.

Next.

Meta is advancing its AI-driven advertising capabilities, aiming for a future where ad creation is fully automated. However, marketers express skepticism about the timeline and effectiveness of these tools. Meta's Andromeda ad retrieval system and the Advantage+ AI tool suite are central to this push, offering features for creative, targeting, and budget optimization. These tools require marketers to relinquish some control over ad targeting and creative processes, leading to mixed reviews. Some advertisers appreciate the potential for increased efficiency, while others are wary of the "black-box" nature of these systems. The Andromeda system encourages a broader audience reach by using diverse creative content, which has led to a significant increase in the volume of creative assets needed. Despite the potential benefits, many brands remain cautious about using AI-generated creative due to legal concerns and a desire to maintain brand integrity. Meta has introduced features to allow advertisers to opt out of certain AI tools, but some feel pressured to use them to avoid performance penalties. While some marketers report performance gains from Meta's automation tools, others find them less effective than third-party systems. Meta continues to refine its AI tools, with recent updates reportedly improving ad quality on Facebook by 14%. Nonetheless, human oversight remains crucial for many brands to ensure quality and engagement.

That said.

The FBI successfully extracted copies of incoming Signal messages from a defendant’s iPhone, even after the app was deleted, by accessing the device’s push notification database. This occurred during a trial involving individuals accused of setting off fireworks and vandalizing property at the ICE Prairieland Detention Facility in Alvarado, Texas, with one person shooting a police officer in the neck. The incident illustrates how forensic extraction, which involves physical access to a device and the use of specialized software, can uncover sensitive data from secure messaging apps in unexpected locations. Signal offers a feature that prevents message content from appearing in push notifications, underscoring its potential importance for users concerned about privacy. This case highlights the need for users to be aware of how their data might be stored and accessed, even when using apps designed for secure communication. The information was reported by 404 Media, with further details available to paid subscribers.

On a different note.

Former Apple engineers Chris Nolet and Ryan Burgoyne have developed an AI wearable device called Button, designed to prioritize privacy and immediacy. The device, which resembles an iPod Shuffle, activates only when pressed, ensuring it does not listen passively. This feature addresses privacy concerns, as Nolet was unsettled by a past experience where he was unknowingly recorded. The Button, priced at $179 and available for preorder, is set to ship in December. It features a generative AI chatbot that responds quickly to queries, unlike previous devices such as the Humane Ai Pin, which failed due to slow response times and unmet promises. The Button can connect to earbuds or smart glasses via Bluetooth and is intended to complement smartphones rather than replace them. Nolet emphasizes that the Button is designed to be fashionable and versatile, usable as a wearable or stored in a pocket or bag. The device aims to be the hardware solution for generative AI, offering a new way to interact with AI technology beyond traditional smartphones and computers.

Meanwhile.

Explosive Media, a group of young Iranian activists, has gained attention for its AI-generated Lego-style videos mocking Donald Trump and the US amid the Iran war. Since February, the group has released over a dozen videos, including one where a Trump mini-figure interacts with Gulf leaders and Iranian officials. These videos, which have gone viral on platforms like TikTok and Instagram, are noted for their sophistication and cultural insight. Critics suggest the group may have ties to the Iranian government, given their internet access in a restricted country. Despite this, Explosive Media claims independence, emphasizing their understanding of American culture and internet dynamics to engage audiences effectively. Their content contrasts with the Trump administration's narrower meme strategies.

After that.

Google has introduced an integrated Notebooks feature to its Gemini platform, enhancing organizational capabilities for chats and files. This feature syncs with NotebookLM, allowing users to manage personal knowledge bases across Google products. Notebooks can be used to perform specific tasks by organizing content and providing context-specific instructions. They also sync between Gemini and NotebookLM, enabling access to features like Video Overviews and Infographics. The number of notebooks and sources available depends on the user's Google AI subscription plan, with varying limits for free, Plus, Pro, and Ultra users. The rollout begins with web access for Ultra, Pro, and Plus subscribers, with plans to extend to mobile and free users. Google aims to add more features to Notebooks in the future.

In other news.

ChatGPT Ads has achieved $100 million in annualized revenue within six weeks of its pilot launch, as confirmed by OpenAI. The platform, which began on February 9, 2026, now hosts over 600 advertisers, including major brands like Best Buy and Ford, targeting its 900 million weekly active users. A self-serve advertising platform will launch in April 2026, eliminating the $200,000 minimum commitment, thus opening the platform to mid-market and growth-stage advertisers. Despite a lower click-through rate (CTR) of 1.3% compared to Google's 29.2%, ChatGPT ads reportedly deliver higher conversion rates, with users converting at 1.5 times the rate of other channels. This is attributed to the high-intent nature of the audience engaging with AI-driven conversations. The platform's $60 CPM pricing reflects this premium audience engagement. International expansion is planned for Canada, Australia, and New Zealand, offering new opportunities for advertisers. The upcoming self-serve launch is expected to increase competition, but early adopters may benefit from lower competition and costs. Advertisers are advised to prepare conversion tracking and identify high-intent topic clusters to optimize their campaigns in this emerging advertising landscape.

Finally.

Meta's AI app surged to No. 5 on the U.S. App Store following the launch of its new AI model, Muse Spark. This marks a significant leap from its previous position at No. 57, driven by increased consumer interest. Muse Spark, Meta's first model under Alexandr Wang, offers multimodal input capabilities, including voice, text, and images, and is designed to assist with tasks such as health education and complex problem-solving in science and math. It also supports visual coding for creating websites and mini-games. Despite this growth, Meta AI still trails behind leading AI chatbots like OpenAI's ChatGPT, Anthropic's Claude, and Google's Gemini. The app has been downloaded 60.5 million times globally, with 25 million downloads this year alone. India leads in downloads, followed by the U.S., Brazil, Pakistan, and Mexico. Meta plans to expand Muse Spark to platforms like WhatsApp, Instagram, and Facebook in the coming weeks.