Before the main story, here are some important happenings this week.
AI frontier companies can’t reliably contain the systems they’re building. A new assessment found significant gaps in how OpenAI, Anthropic, Google, xAI and Meta contain and monitor advanced AI. As companies give AI agents greater autonomy, permissions, containment and monitoring are quickly becoming core cybersecurity and risk-management priorities. (Reuters)
AI agents could break marketing’s traditional attribution model. The IAB says the rise of agentic commerce is making conventional attribution harder because purchases can increasingly happen without a traceable impression or click. Its new AI initiatives are focused on creating shared ways to classify and credit AI-influenced outcomes as machines play a larger role in discovery, evaluation and purchasing. (IAB)
ChatGPT is rapidly becoming a global advertising channel. OpenAI has expanded ChatGPT Ads internationally and is building out ways for marketers to buy, manage and measure campaigns inside the platform. The strategic implication is significant: if consumers increasingly use conversational AI to explore options and make decisions, ChatGPT could aggressively compete with search, social and retail media for marketing budgets. (OpenAI)
Creator marketing is becoming part of brands’ AI search strategy. Brands and agencies are starting to examine whether creator content surfaces in AI-generated answers and which creators are already cited by AI engines. The shift could blur the lines between influencer marketing, PR and SEO as marketers try to influence not just what people see online, but what AI systems say about their brands. (Digiday)
Registration is now open for the AI Trailblazers Intelligence Summit, taking place on September 29 at Rockefeller Center in New York City. Join fellow senior marketing, technology, and business leaders for practical strategies to move from AI experimentation to measurable business impact, with a strong focus on honest lessons learned, real-world case studies, and actionable insights. Learn more at fall26.aitrailblazers.io.
Taken together, these stories suggest that the next AI bottleneck will not be access to intelligence. It will be the speed at which companies can absorb new signals, decide what they mean and act while the opportunity or risk still matters.
The Decision Latency Advantage
Back at PepsiCo years ago, alongside a few other marketing leaders, I helped pioneer real-time marketing. The ambition was to deliver marketing at the pace of culture. Or in other words to sense what was happening in the world and give brands the ability to respond while the moment still mattered.
The hardest problem was not identifying the right moments or producing the content. The problem was the company itself.
Who could say yes at eleven at night? How many people had to see the work before it went out? What happened when legal, brand and the regional team disagreed and the moment closed while they were still on the thread? It is fair to say that we were forced to spend far more time redesigning approvals, ownership and escalation than we did on the creative itself.
What we were really building was not a marketing capability. It was a learning loop that happened to live inside marketing, because marketing sits closest to the customer and often feels organizational lag first. This matters much more now than it probably did then.
The Expensive Misunderstanding
Most companies still define marketing as messaging, content, media and campaigns. It is the function that explains the experience after the important product and business decisions have already been made.
Sarah Friar, CFO at OpenAI, recently wrote on LinkedIn that trust “isn’t built with marketing.” It comes from doing the right things, keeping your word and improving people’s lives. She is right that trust is earned through behavior. No campaign has ever rescued a company that consistently disappoints its customers.
But trust is not something separate from marketing. Marketing should help a company understand what it has promised, whether customers experience that promise and what must change when they do not. Advertising is one output of that system. It is not the system itself.
For years, misunderstanding this was mostly a status problem for marketers. Now it is an economic problem, because how a company defines marketing determines what AI does to it in the years ahead. Let me explain.
McKinsey’s latest State of AI survey finds that 88% of organizations use AI in at least one function, yet only 39% can attribute any EBIT impact to it. Of the thirty-one organizational practices McKinsey tested, fundamental workflow redesign was among those most strongly associated with bottom-line impact. Not model choice. Not budget. Not talent.
Putting an intelligent system inside a slow organization does not create a fast organization. It creates a faster way to encounter the same bottlenecks and to do so repeatedly. That’s even more painful!
Marketing is living this in an unusually vivid way. In BCG’s 2026 survey of 300 global CMOs, roughly half say marketing now owns AI investment decisions inside the function, yet 42% still use generative AI only to assist humans with discrete tasks, and just 8% run campaigns where multiple agents operate autonomously. Marketing has been handed the mandate and the budget, and has largely spent both making the existing work faster. That’s a problem.
In other words if marketing is treated as a simple advertising department, AI gives it more content, produced faster, at lower cost and in greater volume. That is throughput. It is measurable, budget-friendly and potentially useless.
The real prize is collapsing the distance between a customer changing their mind and the company changing what it does for them.
The Metric That Matters
I have started asking leadership teams to measure something most have never considered - decision latency.
Decision latency is the elapsed time between a meaningful customer signal arriving somewhere in the business and a decision changing because of it. In many large organizations, the first honest answer is somewhere between a quarter and never. The goal is not simply faster decisions. It is reducing the time required to make a decision that is at least as good, and ideally better, because it incorporates more current evidence.
To make this operational, choose three to five recurring decisions where delay has a visible economic cost such as responding to a change in customer behavior, shifting media investment, correcting a deteriorating customer experience, adjusting an offer or changing a product priority. For each one, record when the relevant signal became available, when the organization recognized it, when the evidence needed to act was assembled, when someone with authority made the decision and when the resulting change reached the customer. That separates sensing latency, analysis latency, approval latency and activation latency. A simple decision log will quickly reveal where the time actually disappears.
Decision latency should not be reported as a flattering average. Leadership teams should track median signal-to-decision time, the 90th-percentile time that exposes decisions becoming trapped in the organization, and the number of handoffs and approvals involved. They should also define a relevance window for each type of decision and track the decision expiration rate, the percentage of signals that fail to produce a decision before that window closes. Those measures should be paired with a rework or reversal rate, because the goal is not to make bad decisions faster. If latency falls while reversals, errors or customer harm rise, the organization has accelerated haste rather than improved its intelligence.
Once that baseline exists, AI’s contribution becomes much easier to evaluate. Is it detecting the signal sooner? Is it assembling the evidence faster? Is it retrieving relevant precedent, exposing missing information or generating better alternatives? Is it allowing decisions to be made closer to the work without adding unacceptable risk? If AI has not materially shortened any of those intervals while maintaining or improving decision quality, it has accelerated production rather than decision-making.
A field experiment involving 776 professionals at Procter & Gamble offers a glimpse of what is possible. Researchers found that individuals working with AI produced solutions comparable in quality to two-person teams working without it. More importantly, AI helped dissolve functional boundaries. Technical people produced more commercially balanced solutions, while commercial people produced more technically grounded ones.
The handoff between functions, where much of the elapsed time inside a large company disappears, began to collapse for the first time. And that’s what was most surprising and impactful.
But faster does not automatically mean better. Researchers at BetterUp Labs and the Stanford Social Media Lab, writing in the Harvard Business Review, gave the failure mode a name: workslop, work that looks credible but lacks the substance to move the task forward. Around 40% of the desk workers they surveyed had received workslop in the previous month, and took close to two hours to repair each time. The damage is not the bad output. It is the transfer of thinking from the person who produced it to the colleagues who have to fix it. A function that measures itself on volume will quietly turn an intelligence advantage into yet another review bottleneck.
That is why every AI-enabled workflow needs an evaluation loop. Three questions are enough to begin, and they are the three I open most of my advisory conversations with:
What outcome are we trying to improve?
What evidence would change our mind?
How quickly would that evidence alter the next decision?
The third question exposes the maturity of the organization in the AI era.
The Real-Time Intelligent Enterprise
The shift from real-time marketing to the real-time intelligent enterprise will not come from buying more tools or simply putting your employees through more training. It requires redesigning the handful of processes where delay costs the company most.
Marketing has a central role because it should be one of the company’s most sensitive nerve endings. It detects shifts in customer behavior, language and expectations. Product interprets them. Technology changes the experience. The company measures the response and decides again. That is the loop which matters to every CEO and every board.
The large language model itself will not be the moat. Over time, every competitor will have access to increasingly similar intelligence at a falling price. The advantage will come from how quickly an organization can turn that intelligence into learning, and learning into a better decision. That is why decision latency matters as it measures not how sophisticated your AI is, but how much faster your company is becoming because of it.
At PepsiCo, the constraint was outside the building. Could a large company see culture moving and act before the opportunity disappeared? The constraint has now moved inside. Signals are abundant, synthesis is cheap and generation is nearly free. What remains scarce is an organization capable of acting while the decision is still about the present. If your AI investment has not made decisions dramatically faster and the work demonstrably better, it has probably been installed somewhere safe, away from the decisions that matter.
The winners will not be the companies that use AI. Nearly every company will. They will be the companies that learn faster because of it in places where customer signals change decisions, where decisions change the experience, and where the results feed the next decision before competitors have finished their meeting. The winners will be the companies whose learning curves bend upward before their competitors notice the ground has moved.
Where I’ll be this fall
Join us at the AI Trailblazers Fall Summit at Rockefeller Center in New York on September 29 where you will learn more about what it takes to be an Intelligent Enterprise. You can find more details and register here.
Here’s a snapshot of some of the public conferences where I’ll be speaking in the coming months, in addition to the company-specific training and consulting engagements that I do.
What I’ve written lately
Companies Are Missing the Point of AI (August 2026)
Taste Is No Longer a Human Advantage (July 2026)
When AI Becomes the Referee (July 2026)
The Front Door Moves (July 2026)
The Execution Layer is Collapsing (June 2026)
Shiv Singh is the CEO of Savvy Matters, which helps business teams translate AI disruption into practical business and marketing strategies, organizational design, executive-ready roadmaps, and bespoke education programs. He is also the Co-Founder of AI Trailblazers, a vibrant community uniting marketers, technologists, entrepreneurs, and venture capitalists at the forefront of AI.
A former two-time Chief Marketing & Customer Experience Officer and author of Marketing with AI for Dummies (4th print run, translated into five languages), Shiv built his career at LendingTree, Visa, PepsiCo, and The Expedia Group, and serves as a public-company board member of a Fortune 300 company and private investor.