Julianne DeVincenzoFive dispatches from Opticon NYC: Seth Godin's uncomfortable obvious, 154,000 experiments, AI that needs to be lazier, and what Alicia Tillman left in the room.
Every marketing conference in 2026 has an AI story. The useful part is seeing where that story meets the work: technology moving faster than teams, operating models, and our willingness to make harder decisions about how the work gets done.
September 1 at the Javits. Five breakout tracks. A main stage. Hundreds of marketers trying to answer the same question: what does good work look like now, and what are we willing to change to get there?
A few postcards from the day.
Alex opened with five words that set the standard for everything that followed: humanity is marketing's competitive edge. The weight was in what came next. For leaders, deciding where humans belong in the loop is now an operating decision with consequences for speed, quality, and accountability.
Robert Chatwani from DocuSign and Lauren Boyman from KPMG brought that idea down to actual choices inside large organizations. They spoke from the middle of the work, with the constraints and unfinished questions still visible. That honesty is rare on a main stage.
The session named five human qualities that become more valuable as AI becomes more capable: judgment, creativity, empathy, instinct, and accountability. With the right tools and instincts, one marketer can now reach an audience that once required a much larger team. That changes who gets to compete. It also raises the standard for every decision behind the output.
AI creates abundance. Human judgment decides whether any of it matters.
Julianne DeVincenzo|Head of Content, Optimizely
Shaf opened with a set of numbers that made the scale tangible: 500,000 agent executions, a path toward one trillion tokens, and customer workflows reduced by as much as 97%. Campaigns are responding to real-time news in minutes. Agents are completing first-pass brand and legal reviews before a draft reaches a person. This is already happening inside customer environments.
The title said it plainly: One Platform. Three Trails. Infinite Scale. In 35 minutes, the keynote delivered a year's worth of announcements. Each launch moved the platform closer to how modern content teams actually operate.
Mark AI is where that story starts. It introduces purpose-built models post-trained for marketing work and evaluated across 285 marketing functions using expert-written criteria and independent judges. Shaf also announced the AI Benchmark for Marketing, an open-source benchmark built to measure marketing AI objectively. Optimizely's Agent Platform scored 58% all-pass against Claude Code's 52%, with 2.1 times the efficiency.
Lauren Hammarstedt introduced Virtual Teammates, which may change the daily rhythm of content teams faster than anything else announced. The demo featured Paige, a Product Marketer Virtual Teammate. Paige flagged analyst results, found underperforming pages through GA4, drafted updates, routed approvals, and published. The work kept moving without another status chase in Slack. Five Virtual Teammates are available now: Chief of Staff, SEO and AI Search Analyst, Marketing Analyst, Personalization Strategist, and CRO Manager.
Content teams should pay close attention to the SEO and AI Search Analyst because discovery is changing faster than most editorial workflows. Naz framed the shift with one figure: AI referral traffic and crawlers now account for 60% of web visits. Agent Visibility, built with Conductor, shows how a brand appears in AI-generated answers, including citations, visibility, and category competitors. Agent-Ready Experiences extends that work through WebMCP so agents can find, understand, and act on content.
The content team's Q4 calendar got old before Q4 started.
the efficiency of Optimizely's Agent Platform against Claude Code's according to the AI Benchmark for Marketing, an open-source benchmark built to measure marketing AI objectively.
Chuck Gahun from Forrester and I called our session Humanity Is the Premium Layer. The thesis was on the slide but the room did the real work.
I arrived with a structured argument and the audience pushed it somewhere more honest. People have heard enough reassurance that humans still matter. They want to know what that belief changes on Monday: who owns the final call, where taste enters, when speed stops helping, and who is accountable when the output is wrong.
A deeper fear surfaced in the discussion. Marketers are afraid of losing the part of the job that asks them to understand another person, the part that made the work worth choosing in the first place. The premium layer is protected through deliberate decisions about what the machine holds and what the human owns. Capability alone will never make those decisions for us.
Nikita Bokil had fifteen minutes and one sharp argument: AI needs to be lazier. She argued for restraint as a design choice, leaving room for judgment and keeping a person accountable for the decisions that carry weight.
Most teams measure maturity by how much AI can do. Nikita asked whether the stronger signal is knowing when to leave it alone. That requires saying no to capability, something many organizations still mistake for a lack of ambition.
Fifteen minutes. One clean idea. Then she left space for people to do something with it.
We built the breakout tracks around a simple premise: the format shapes the conversation. The Trailhead was our room for customer stories and real metrics. The Outpost gave one speaker one position to defend, with no Q&A to soften the edges. Across both rooms, the strongest sessions dealt with the space between capability and application.
Verndale opened the Trailhead by showing an AI-native content workflow from brief to published page. The agent handled the structural work. People made the judgment calls. What struck me was how ordinary it looked, in the best way. The question for most teams is how long they wait to build toward it.
Horizontal Digital's The Work Nobody Budgets For named the invisible labor eating content teams alive: coordination, status updates, version management, and the endless hunt for ownership. The industry spends a lot of time speeding up the draft while the connective tissue absorbs the day. That is where relief could be felt first.
OneMagnify's Revenue Doesn't Disappear Overnight made the commercial case for patience. A cut to content, editorial frequency, or distribution may take six to eighteen months to appear in revenue. By then, the organization has moved on and the original decision has disappeared from view. Leaders have to hold short-term efficiency and long-term consequence in the same frame.
In The Outpost, Peter Ten Eyck put 154,000 experiments on the screen and asked what winning looks like when AI changes the rules. The number carried its own argument. It represented 154,000 decisions to test an assumption instead of trusting it. AI will raise the standard for experimentation. Teams still need to know what they are testing, why it matters, and what they will do when the result surprises them.
Across both rooms, the tool mattered less than the decision behind it.
The afternoon panel started with an uncomfortable premise: the AI playbook is already outdated. Suzy Davidkhanian from EMARKETER moderated Wei Zheng from Conductor, Alex David from G2, and Jeff Barg from Clay. Their vantage points were different enough to create useful disagreement about where human judgment belongs and where teams should place their investment.
Wei treated AI-generated answers as a distribution channel. Brands that structure content for an answer engine are pulling ahead of teams still publishing for a search index alone. Alex focused on quality. Faster production can create more content that nobody reads when editorial judgment loses ownership. Jeff focused on sequence. AI creates more value during research and structure than it does when asked to put a glossy finish on a thin draft.
Together, they made a practical case for specificity. The right approach depends on what you are making, who needs it, and how your own evidence changes the next decision. Put the tokens into research and structure. Keep human judgment closest to the finish.
Alicia Tillman walked to the center of the stage and gave the room more than twenty-five years of choices, pressure, mistakes, and consequence. She stayed with the specific moments and trusted people to understand what they meant.
With Alicia, the room changed. People stopped taking notes. The silence after certain lines came from recognition, the feeling of hearing someone else name part of your own experience. She understood that a story can carry its own proof when the details are true enough. At a marketing conference crowded with frameworks and proof points, one of the day's most persuasive moments was someone telling the truth cleanly and leaving the interpretation with us.
Seth Godin ended the day by naming what had been underneath it: 'You're not stuck. Something invisible is holding it.' He called it the uncomfortable obvious.
The invisible constraint is often our attachment to existing logic, existing metrics, and an old definition of what the work is for. A new platform can arrive quickly. Redesigning the work asks someone to question what the organization has been optimizing toward and say it out loud when agreement would be easier.
If failure is not an option, than neither is success.
Seth Godin|21 bestsellers, founding editor of the Carbon Almanac, blogger and entrepreneur.
Across the day, speakers gave that constraint different names: operating model, workflow, sequencing, human judgment, psychology. The point held. Capability moves quickly. Organizations change when someone is willing to challenge the shape of the work itself.
The room went still. It was the quiet that arrives when people recognize something they already knew and finally have language for it.
I spend a lot of time building toward that two-second beat after a sentence lands. A room gets quiet. People recognize something they had been carrying without a name. Six months later, they may forget the slide and remember the shift.
That is the content worth making.