Definitions of AI concepts in a marketing context, as used and developed in day-to-day practice by Florian Michael Fischer, Chief AI Officer at SlopeLift Vienna. As of 2026.
The guest commentaries linked here were originally published in German; the versions on this site are English.
Agentic Commerce is the use of autonomous AI agents in e-commerce that make purchasing decisions, run campaigns and execute processes on behalf of users or companies.
In a performance marketing context, Agentic Commerce means campaign agents optimise budgets and creatives without manual intervention, attribution agents connect touchpoints in real time, and shopping agents buy on behalf of consumers. Florian Michael Fischer has been working with Agentic Commerce systems at SlopeLift since 2025, for clients including BAWAG, dm and Austrian Airlines.
Further reading: The Funnel We Forget, a guest commentary in medianet on Agentic Commerce and new purchasing processes. And: The Agent That Wasn’t, on why most AI agents are automation in disguise.
Agentic Workflows
Agentic Workflows are automated process chains in which AI agents plan and execute multi-step tasks on their own, guided by defined goals rather than fixed step-by-step instructions.
In marketing: an Agentic Workflow for campaign reporting analyses data on its own, identifies anomalies, formulates recommendations and produces the report. The human reviews the result. SlopeLift has used Agentic Workflows for reporting, campaign optimisation and content processes since 2025.
Agentic Workflows are the technical implementation of the Human-on-the-Loop model: the AI handles execution, the human handles oversight.
AI-Leverage
AI-Leverage is the ability to deploy AI systems so that one person or a small team produces output that previously required many times the headcount.
AI-Leverage is a way of organising and thinking: which tasks are systematically delegated to AI agents? Where is human judgement concentrated on the points where it has irreversible consequences? Florian Michael Fischer observes at SlopeLift that AI-Leverage is becoming the decisive dividing line in the labour market, particularly for mid-career people whose domain knowledge AI can replicate. The gap that an AI-Leverage lead creates can no longer be closed by hiring alone.
Further reading: The Job Crisis Is Already Here. A guest commentary in medianet on AI-Sorting in the labour market and what it means for marketing teams. And: Sea of Sameness, on why high AI-Leverage is the only way out of AI-driven uniformity.
AI-Sorting
AI-Sorting is the quiet sorting of the labour market by AI: the transformation is not driven by waves of layoffs but by a widening gap between those who have built AI-Leverage and those whose skill profile has become cheaper and more replaceable.
Florian Michael Fischer analyses this effect at SlopeLift and beyond. When 10 per cent of a workforce produces three times as much with AI, the headline figures barely move. The sorting happens anyway: through projects that get smaller, roles that are advertised differently next time, briefings that order less external capacity. The most exposed group is mid-career people with ten to twenty years of experience: precisely the domain knowledge AI can replicate cheaply.
Further reading: The Job Crisis Is Already Here. A guest commentary in medianet on AI-Sorting in the labour market. And: Sea of Sameness, the output side of the same effect: where uniformity comes from, and where differentiation remains.
Chief AI Officer (CAIO)
The Chief AI Officer (CAIO) owns the company-wide AI strategy and its operational implementation, with decision-making authority and budget control over AI projects.
The role grew out of the realisation that AI projects fail when they run without C-level ownership. The CAIO sits between strategy (the board) and implementation (tech and marketing). Florian Michael Fischer has been CAIO at SlopeLift since October 2025, one of the first people in this role in Austria.
Further reading: Being a CAIO in Austria. The role, its responsibilities and day-to-day practice in an Austrian agency context.
Delegation Matrix
A delegation matrix records, for every AI-supported workflow, what an agent decides alone, what requires human approval and what stays off limits entirely. It has to exist before an agent goes live.
The concept comes out of Florian Michael Fischer's practice as Chief AI Officer: without explicit delegation the opposite happens: agents decide where they should not, and humans approve where automation gains nothing. The matrix is not a technical document but a strategic decision: it names what may belong to a human only. Together with Human-on-the-Loop it is the operational counterpart to the governing philosophy: the philosophy says where the human stands, the matrix says when they step in.
Further reading: The Human-on-the-Loop Principle, the guest commentary in horizont.at, where the delegation matrix is described for the first time.
FOMA (Focus On Maximising Advantage)
FOMA (Focus On Maximising Advantage) is a decision framework that prioritises AI investments by the competitive advantage they create. How new a tool is, or how widely it is covered, plays no part.
Florian Michael Fischer developed the concept as a counterweight to AI strategies driven by tool hype. FOMA asks: where does AI create the largest measurable lead over the competition in this specific company? That question governs every AI investment decision at SlopeLift.
Not every new AI tool is relevant. FOMA filters by actual effect and stops resources flowing into demos instead of results.
Human-on-the-Loop
Human-on-the-Loop is a governance model for AI systems in which AI agents work autonomously while humans monitor the process and intervene at defined thresholds.
The model differs from Human-in-the-Loop (where a human approves every step) through a structural shift in emphasis: with in-the-loop, the human does the work and the AI assists. With on-the-loop, the AI handles execution. The human sets the goals, supplies the context and intervenes at defined thresholds. Florian Michael Fischer uses Human-on-the-Loop at SlopeLift as the standard for Agentic Workflows.
In practice the model forces an explicit delegation matrix: what does the agent decide alone, what needs human approval, what stays off limits? That matrix has to exist before the agent goes live. A bad process plus AI produces more bad output, faster. Human-on-the-Loop is the discipline that prevents this.
Sea of Sameness describes the homogenising effect of identical AI tools on marketing output and company differentiation: when every team uses the same models with similar prompts, the output converges, and differentiating through tool access becomes impossible.
The term comes from Raja Rajamannar (former Global CMO of Mastercard), who observed a „massive Sea of Sameness“ in marketing. Florian Michael Fischer's analysis adds the empirical dimension: AI assistants generate under 0.5 % of all website visits but deliver around 11 % of B2C revenue, at three times the conversion rate of paid search. Whoever is the brand that AI systems recommend wins disproportionately. Standard output becomes a commodity.
The answer to the Sea of Sameness is not better prompts but three factors that cannot be copied: proprietary data, judgement (→ AI-Leverage) and leadership willing to make real bets. The Sea of Sameness makes visible what AI-Sorting describes, seen from the other side: uniformity of output at company level, a widening skills gap at the level of individuals.
Further reading: The Sea of Sameness, a guest commentary in medianet and SwissCognitive on differentiation in the age of AI.
Two-Funnel Model
The two-funnel model describes running two purchasing processes in parallel: a human funnel (emotion, story, brand) and an agent funnel (structure, specs, machine-readable signals). Both run at the same time, with the weight shifting towards the agent funnel.
The idea comes from the Agentic Manifest (Florian Michael Fischer, 2026): optimise marketing for humans only (attention, emotion, story) and the second funnel sits empty. Build for agents only (structured data, machine-readable feeds, clean APIs) and the brand is lost. Do both, and you win both: the human purchasing decision and the agentic one.
In the human funnel, differentiation comes from experience, trust and narrative. In the agent funnel, what decides is feed quality, availability data, return rates and checkouts a machine can walk through. The infrastructure for the agent funnel is being built right now, while the human funnel has been understood for 18 years. The weight shifts, but neither disappears.