Pick the one closest to your event and I'll tailor it from there.
01
Custom AI Assistants for Every Employee
Stop having your team start every AI conversation from scratch.
How MMA deployed MAVEN (a custom AI assistant pre-loaded with org context, brand standards, tool integrations, and data-handling rules) to every employee across a 90-person global organization. We cover the architecture decisions, how the governance rules ship inside the assistant so staff comply by default, and how we migrated 88 seats between AI platforms without disrupting workflows. The session includes live demos, and attendees take home the MAVEN template.
proof: Deployed across 90 staff in 19 countries
audience: AI workflow integration, team enablement format: 30-min keynote · 60-min talk · 90-min workshop
02
What's Actually Working: AI Across the Ad Funnel
Results from 29 experiments with 18 brands, averaging a +160% lift.
Findings from 29 controlled experiments across 18 global brands including Kroger, AT&T, GM, Shell, and Home Depot, testing AI applications at every stage of the programmatic advertising funnel: AI-generated video creative, pre-launch creative scoring, audience discovery models, contextual targeting using LLM embeddings, and full-funnel optimization. The talk separates the applications that improved results from the ones that aren't ready yet.
proof: +160% avg lift, top performers >+250%
audience: data, measurement, AI analytics, case studies format: 30-min keynote · 45-min talk
03
Building an AI-Capable Workforce
What to teach people once they already know how to prompt.
The workforce training model MMA runs across its global staff: a gamified self-paced AI Mastery program, a tiered certification track (Foundations → Agent Development → Agent Management), and the AI Agent Foundry, a mandatory program where every employee dedicates 10% of their capacity to building automations. We show what people built, including meeting transcripts that auto-generate CRM updates, lead discovery pipelines, weekly reporting automations, and branded content drafts. Attendees get our validated impact templates.
proof: 100% staff trained, 40+ live production workflows
audience: team training, enablement, change management format: 45-min talk · 90-min workshop · AI Academy session
04
Embedding AI Governance Into Your Stack
Governance your team follows because the tools enforce it.
How to build governance into the tools, approval processes, and system architecture your team already uses, so the responsible choice is also the easiest one. The material comes from MMA's Responsible AI Innovation Lab (RAIL), which has published 8 governance frameworks adopted by Fortune 500 marketers, and from my Cambridge research on multi-agent governance. The session walks through automated risk tiering, AI usage policies that travel inside the assistant, self-enforcing data classification rules, and approval workflows that route by risk level without creating bottlenecks.
proof: 8 published RAIL frameworks, Fortune 500 adopted
audience: ethics, legal, governance, martech format: 45-min talk · 90-min workshop
05
Who Controls Your AI Agents?
The protocols connecting your agents carry design choices that favor the companies that wrote them.
Original research from the University of Cambridge analyzing the dominant agent communication protocols (A2A, MCP) that govern how AI agents coordinate across enterprise systems. Standards marketed as "open" contain architectural choices that concentrate control with platform owners, transform agents into auditable subjects, and quietly reposition human workers as managers of AI teams. Marketing leaders adopting agentic AI need to know whether their agents end up working for them or for the platform.
proof: From my Cambridge dissertation, Protocols of Power. The executive briefing; 07 is the same research as a lecture
audience: ethics, legal, vision, executive briefing format: 45-min keynote · 60-min deep dive
06
The Marketing Org in 2030
Which marketing roles stay human, which become agentic, and what that does to the org chart.
Three plausible versions of the 2030 marketing org, each with different assumptions about agent autonomy, headcount, and where trust breaks. We work through which roles compress, which expand, which new ones appear, and what to hire and reorganize for now. Runs as a keynote or as a facilitated scenario session where your leadership team argues its way to a position. The scenarios draw on MMA's research and the AI Leadership Think Tank's work with CMOs who are partway through the change.
proof: Built on the 300+ ALTT leader community
audience: executive leadership, future of work, scenario planning format: 30-min keynote · 45-min keynote · 90-min facilitated executive session
07
Protocols of Power: A Foucauldian Reading
A protocol specification is a political text. Read it as one.
My Cambridge dissertation treats Google's Agent2Agent protocol as what Foucault called a diagram of power: an architecture that governs conduct at a distance. Working from Foucault, Bowker and Star's classification theory, and Galloway's protocological control, I read A2A's technical and governance documents as governmental texts and trace how the Agent Card makes agents legible, rigid message formats discipline them, mandatory logging builds a panopticon, developers learn to think like a protocol, and a founders-only committee launders corporate authority as open governance. What emerges is a new subject, the docile digital worker, caught at the moment before this infrastructure turns invisible.
proof: 15,287-word M.St. dissertation, University of Cambridge, 2025. The lecture; 05 is the same research as an executive briefing
audience: AI ethics and STS scholars, policy researchers, governance leads, university and think-tank audiences format: 45-min lecture · 60-min seminar
08
The Data Dilemma: Privacy Against Fairness
Detecting bias needs the demographic data that privacy law tells you to delete.
Privacy and fairness are treated as rival values in AI governance, and the tension is real: GDPR's data minimization and special-category rules strip out the attributes a bias audit needs, while fairness pursued without limit invites surveillance. Reading Barocas and Selbst on proxy discrimination against Viljoen's relational theory of data governance, I argue the binary is false. Amazon's hiring tool, the Apple Card, and IBM's exit from facial recognition show what each regime hides. The question underneath is who decides which sensitive data gets used, and under whose authority.
proof: From my Cambridge essay The Data Dilemma
audience: privacy, legal, ethics, policy format: 45-min talk · 60-min seminar
09
Beyond Principles: Why Responsible AI Stalls Inside the Firm
AI ethics has an oversupply of principles and a shortage of mechanisms.
Rakova and colleagues' interviews with 26 practitioners show responsible AI work stalling on the same failures: no authority, no aligned incentive, no metric that competes with revenue. Marchant asks whether professional societies like the ACM and IEEE can fill the vacuum with soft law. I read both against two earlier experiments in industry self-regulation, the Fair Labor Association and the Interactive Advertising Bureau, and argue that without structural separation between standard-setting and member advocacy, and without a credible regulatory threat behind it, self-governance reproduces ethics washing at industry scale.
proof: From my Cambridge essay Beyond Principles
audience: governance, policy, trade associations, AI ethics researchers format: 45-min talk · 90-min seminar
