The Dashboard Era of Marketing Is Ending
This week in AI & marketing
Marketing has been running on the same operational model for at least a decade. You log into a dashboard. You find a problem. You export the data. You build a report. You meet about the report. Eventually, somebody logs back in and makes a change. AI has been helping with pieces of that loop, but the loop itself hasn’t moved.
This week, several of the biggest AI companies made it very clear that the loop is on its way out. Reporting, analysis, planning, even execution are being pulled into a conversational layer where the marketer talks and the system does. The dashboard turns into a chat window, the export into a question, the Friday performance review into a thread.
We said last week that AI would own the customer’s decision cycle. This week, it came for the marketer’s work cycle too. Both sides of the marketing function are being pulled into the same conversational layer simultaneously.
The reflex will be to wait. To ask IT. To commission an audit. To pilot it in Q3. That reflex is exactly the problem. By the time the audit lands, the teams that skipped it and jumped straight into the new workflow will have six months of operating knowledge no consultant can sell you, because none of this behaves the way the old systems did.
Same answer as last week. Skip the audit. Start the work.
— Torsten and Peter
AdRoll Shipped Something More Interesting Than Another Dashboard
Most marketing workflows still look surprisingly similar to how they looked five years ago. You open a dashboard. You find a problem. You export data. You build a report. You discuss the report. You decide what to do. Then somebody eventually logs into the platform and makes the change. AI has been helping with pieces of that process, but the workflow itself hasn’t really changed.
AdRoll’s new MCP server is a small glimpse of what the next version might look like.
The company now allows tools like ChatGPT, Claude, Microsoft Copilot Studio, Cursor, and n8n to directly access campaign data and campaign workflows. In practical terms, that means you can ask an AI assistant something like: “Which of my EMEA retargeting campaigns underperformed this month, what do you think caused it, and where would you reallocate budget?” The assistant can pull the data, analyze it, identify patterns, and prepare draft campaign changes without ever leaving the conversation.
That’s a much bigger shift than it sounds.
For years, marketers have been told AI will help them create content faster. That’s true, but content generation was always the easy part. The more valuable opportunity is operational. Most marketing teams spend an enormous amount of time moving information between systems, building reports, preparing recommendations, and translating analysis into action. AdRoll is essentially trying to remove that translation layer.
The company was smart enough to keep humans in control. Every change is draft-first and requires approval before going live. Good. Nobody wants a hallucinating chatbot managing seven-figure media budgets.
Still, imagine where this goes in two or three years. A weekly performance review becomes a conversation. Campaign analysis becomes a conversation. Budget planning becomes a conversation. (“Markets are conversations,” for those of us old enough to remember The Cluetrain Manifesto.)
The AI already has access to the campaign data, attribution reports, CRM information, creative performance, and revenue metrics. Instead of spending half a day assembling the information, you spend your time deciding what to do about it.
OpenAI Wants Codex to Be The Universal Tool For Knowledge Work
If you’ve never touched Codex, you’d be forgiven for thinking this story has nothing to do with you.
Codex launched as a coding product and, to many people, that’s still what it is: a tool for software engineers. But OpenAI’s latest update makes it increasingly difficult to maintain that distinction. Alongside new capabilities like Sites and Annotations, the company also introduced role-specific workflows for sales teams, analysts, researchers, designers, and marketers.
According to OpenAI, knowledge workers now make up roughly 20% of Codex users and are adopting it three times faster than developers.
For years, we’ve talked about AI as something that helps people do their jobs. Increasingly, the major AI labs seem to be building products that want to become the place where those jobs happen. The Sites feature can turn spreadsheets into interactive applications. Sales workflows connect directly into Salesforce and HubSpot. Creative workflows connect to tools like Figma and Canva. The goal is no longer simply helping you write a better email or create a better chart but becoming the environment where the work itself takes place.
This neatly fits into the rumors surrounding OpenAI’s broader product strategy. Multiple reports suggest that ChatGPT, Codex, and Atlas could eventually merge into a single application, a “superapp”. Add Office integrations, Google Workspace integrations, CRM integrations, and dozens of SaaS connections, and you start to see what OpenAI may be building: a universal interface for knowledge work.
Most users will never know whether Codex is writing code behind the scenes to complete a task. Nor should they care. Nobody opens Excel thinking about how it executes formulas. They care about the spreadsheet.
Microsoft Thinks AI Should Stop Waiting For Instructions
While OpenAI is turning Codex into a platform for work, Microsoft is pushing a slightly more ambitious idea.
The company announced Scout this week, an always-on agent that lives across Outlook, Teams, OneDrive, SharePoint, calendars, contacts, and documents. Unlike Copilot, which largely waits for prompts, Scout is designed to operate continuously in the background. It can identify upcoming deliverables, coordinate meetings, block focus time, surface risks, and prepare materials without needing to be asked each time.
The easiest way to think about Scout is as a digital chief of staff.
Most AI assistants today have a strange limitation: they’re incredibly knowledgeable, but they have almost no awareness of your actual work. Every conversation starts from scratch. They know a lot about the world and very little about your world. Scout attempts to solve that problem by maintaining persistent awareness of the systems where work happens.
Whether that sounds useful or slightly unsettling probably depends on your personality.
Increasingly, the big AI companies seem to be betting that the next phase looks less like software and more like colleagues. Not human colleagues, obviously. But systems that understand context, remember objectives, coordinate work, and occasionally take initiative.
Google Finally Built AI Visibility Into The Platform
For the past year, an entire industry has emerged around measuring AI visibility.
New startups appear every month promising to tell you whether your brand shows up in ChatGPT, Gemini, Perplexity, AI Overviews, AI Mode, or whatever new interface launched last week. Some are excellent. Some are questionable. All of them are trying to solve the same problem: marketers have almost no visibility into how AI systems discover and recommend products.
Google just took a meaningful step toward changing that.
Merchant Center is getting a new set of AI Performance Insights, including Share of Voice reporting, conversational query analysis, shopping funnel metrics, and attribute gap detection. For ecommerce marketers, this is probably the most important announcement Google made outside of AI Mode advertising itself.
The most interesting feature is probably Product Term Insights. Traditional search taught us to think in keywords, now AI shopping is teaching us to think in conversations. People don’t search for “running shoes.” They ask for “comfortable running shoes for flat feet that can survive standing all day at work.” Those are fundamentally different discovery mechanisms.
The Merchant Center has traditionally been treated as a feed management tool. Upload products, fix errors, move on. Google’s new reporting suite effectively turns it into an AI visibility platform. Product descriptions, structured attributes, and feed completeness are becoming discoverability assets rather than administrative tasks.
Some ecommerce teams will adapt quickly. Others will continue treating product feeds as a logistics problem and wonder why competitors keep appearing in AI-powered shopping experiences while they don’t.
Ahrefs Just Published One Of The Most Useful AI Search Research Summaries We’ve Seen
Speaking of AI visibility, Ahrefs has probably earned the right to be taken seriously on this topic.
Over the last six months, the company analyzed more than a billion data points across a dozen of separate studies looking at AI search behavior. This week, CMO Tim Soulo published a summary of the most important findings, and several of them challenge assumptions that have become surprisingly common in AI SEO circles.
The first finding is probably the least surprising: “Best X” listicles remain extraordinarily influential, with nearly 44% of pages cited by ChatGPT fall into that category. Love them or hate them, the format continues to dominate AI citations.
The second finding is much more interesting. Roughly 28% of ChatGPT’s most-cited pages have zero Google organic visibility. Not poor rankings. Not page two. Zero visibility. That suggests AI search is increasingly becoming its own discovery layer rather than a simple extension of traditional search results.
Then there’s schema markup. If you’ve spent any time on LinkedIn recently, you’ve probably encountered someone claiming that schema is the secret to AI visibility. Ahrefs found essentially no impact. AI Overviews slightly decreased, AI Mode slightly increased, ChatGPT slightly increased, but none of the changes were statistically meaningful.
Out of every factor Ahrefs studied—including backlinks, domain authority, page count, and traditional SEO metrics—YouTube mentions showed the strongest correlation with AI brand visibility.
We’re still very early in understanding how these systems work. Anyone claiming to have fully solved AI search optimization is probably selling something.
But we’re finally moving beyond theories and anecdotes. Read it yourself:
Share with a colleague. Earn paid membership months.