A short post on what I'm currently speaking about, why these three topics in particular, and how the talks have evolved over the last year as my own thinking has sharpened.
If you're an event organiser thinking about programming, or a corporate L&D lead looking for someone to bring perspective into a leadership offsite, this is the most direct version of what I'm currently equipped to deliver.
Thesis 1 — GEO is structurally different from SEO
The talk most agencies and conference audiences want from me right now. The argument: AI search is not a sub-discipline of SEO. The retrieval mechanism is different, the measurement frame is different, and the team capability required to win in it is different.
The 30-minute version walks through specific examples from our own work — how share-of-voice in ChatGPT correlates poorly with classical organic visibility, how some brands with strong SEO are weak in AI search and vice versa, and what the operational implications are for in-house marketing teams and the agencies that serve them.
This talk has evolved as the data set has grown. Eighteen months ago, the argument was speculative. Now it's grounded in roughly a year of observed data across hundreds of brand-and-prompt combinations. The numbers carry the argument now in a way they couldn't earlier.
Audiences this works best for: marketing leadership conferences, agency leadership summits, in-house marketing offsites, and SEO-focused events where attendees are starting to wonder whether their existing operating model needs updating.
Thesis 2 — Agentic AI is workflow, not interaction
The argument that's gotten the strongest response in the last year. Most marketing AI investment has been on chatbots and assistants — interaction automation. The actual leverage is at the workflow level — agents owning multi-step processes end-to-end. Same underlying technology, dramatically different return profile.
The talk uses the amivisible.co pipeline as the working example, walks through what the agentic version of an audit workflow looks like, what failure modes we encountered, and what marketing teams could be doing in their own contexts. Often runs alongside a Q&A or workshop component because the audience consistently wants to apply the framework to their own workflows.
The talk has gotten harder as the audience has gotten more sophisticated. Two years ago it could be a fairly broad introduction to agentic systems. Now most rooms have at least one person who's tried to build something agentic and has questions about specific failure modes — observability, hallucination, cost management. The talk has had to develop more depth in those areas.
Audiences this works best for: corporate workshops, agency leadership groups, conferences with a strong technical audience, and innovation-themed events where the question isn't "is AI useful" but "what are we missing in how we're applying it".
Thesis 3 — The four-day week as hiring strategy
The newer argument I've been developing publicly — the four-day week as a hiring strategy. The wellbeing case for a four-day week is well-rehearsed. The hiring case — that the model functions as a filter that brings senior, output-oriented, self-directed operators into the building — is less so, and I think it's the stronger argument.
The talk draws on roughly four years of operating Marketing Signals on a four-day week, with specific data on how the hiring funnel changed, what kinds of senior people self-selected in and out, and how the operational discipline required to make the model work also produces an organisational shape that's better-suited to senior knowledge work generally.
This talk is positioned slightly differently from the other two. It's less technically dense and more strategic. Audiences are usually leadership groups thinking about how to attract and retain senior talent in tight labour markets — a question that's been more acute in 2025-26 as the senior layer of most agencies has been harder to recruit into.
Audiences this works best for: agency leadership summits, services-business CEO groups, leadership offsites at scale-up companies, and HR/talent-focused events.
How I prefer to structure these
A few preferences worth sharing for organisers.
30-45 minutes works best. Most of the talks have a 30-minute keynote version and a 45-minute extended version with more material and audience interaction. Anything shorter than 30 minutes loses the depth that makes the argument credible. Anything longer than 60 minutes is usually better split into a talk plus a workshop session.
Q&A is genuinely useful and I'm comfortable with it being unstructured. The questions audiences ask shape how the next version of the talk evolves. I prefer programming where Q&A is built in rather than tacked on as an afterthought.
Workshop formats work best for the agentic AI material. The conceptual framework lands cleanly in a 30-minute talk. The actual application lands in a 90-minute workshop where attendees work through the framework against their own workflows. If you can offer the workshop format, you'll get more value than from the keynote alone.
I prefer to have access to slides until shortly before the talk. The talks are evolving as my own thinking does, and I usually update the deck the week before the event to reflect the most current data or arguments. Organisers asking for finalised slides four weeks out get an older version of the talk than they would otherwise have had.
Booking and selection
I'm selective. The reason: the talks each take a meaningful amount of preparation, and I'd rather give a smaller number of talks well than a larger number badly. I lean toward events where the audience is genuinely the right fit for the material — typically marketing or agency leadership — rather than general business audiences where the depth gets lost.
For organisers: the easiest way to start the conversation is via the contact form on this site, with a sense of the audience, the format, the date, and what you're hoping the talk would unlock for the audience.
For corporate L&D leads: the workshops on agentic AI have been a strong fit for senior leadership offsites where the question is "how do we operationalise this in our own organisation". The format works particularly well for groups of 8-25 senior leaders who can engage with the workflow-mapping exercise alongside the conceptual material.
I'd rather decline a talk that doesn't fit than do it badly. If we're not the right match, I'll often suggest someone else who'd be better positioned.