At the November MarTech Conference, Greg Kihlstrom, principal at The Agile Brand, moderated “What AI and agents mean for marketing teams… Now and in the future,” a forward-looking conversation about how agent-powered organizations will reshape roles, workflows and leadership.
He was joined by Jiaxi Zhu, head of analytics at Google; Noah Dinkin, founder and CEO of Stensul; and Robin Ross, digital analytics and loyalty executive at Activate Insight.
A quick poll set the session’s baseline: A handful of teams are far along with agents, but most are either just starting or piloting a few use cases. That context framed the conversation around practical steps — how to define “agent,” where they belong in marketing work, which skills managers need and how to measure value without getting trapped by vanity metrics.
What counts as an “agent” (and what doesn’t)
Level-setting the vocabulary came first. Dinkin drew a clear line between workflows and agents. If a system follows a predefined path — even with branching and orchestration — it’s still a workflow.
An agent is different: It reasons about goals, adapts to context, uses tools and decomposes problems into steps it plans and executes. That distinction matters, he argued, because calling every automation an “agent” muddies expectations about autonomy, governance and risk.
Asked how agents will change the division of labor, Ross focused on acceleration and proximity to the customer. The emerging pattern, he said, is about shrinking the distance between creative and consumer — removing layers that slow execution and measurement.
Agents can compress cycles, but “the human is more relevant than ever” to set strategy, apply domain expertise, and make judgment calls. Zhu added that marketers will lean on agents to surface and synthesize more data with higher accuracy, yet budget decisions, tradeoff logic and cross-channel context still require humans who can connect dots and build trust with decision-makers.
On ethics and brand governance, Dinkin described a near-term operating model many enterprises are adopting: Agents generate first drafts, humans review. The catch is that highly capable systems can ingest more context than any one reviewer, creating tension when human preference contradicts data-backed performance. That reality, he said, intensifies the need for strong governance — guardrails that protect brand equity while letting workflows transform step by step.
Ross pressed on data custody: If teams can’t clearly explain how information is handled end-to-end, they risk reputational damage. The allure of easy wins shouldn’t eclipse obligations to customers, partners and IP owners; privacy, security and legal teams must be embedded from the start.
How should leaders prepare their organizations?
Dinkin’s advice starts with a deceptively simple task most teams skip: Map the actual workflow. Stensul’s “SPEED” framework — Scan the process, Pinpoint bottlenecks, assess Economic impact, identify Experiments and Define guardrails to avoid pushing a bottleneck downstream — forces teams to document who does what, with which tools and where decisions happen.
Zhu urged leaders not to pilot in silos; sales, service and marketing agents all touch the same customer. Data, language and suppression rules must stay consistent across departments, even if the agents differ by journey stage.
Ross cautioned that agents merely let you “create problems faster” if you’re not centered on end-to-end customer experience. Cross-functional alignment is the antidote.
Skills for managers will evolve accordingly. Zhu highlighted workflow design, role clarity and trust-building — deciding what an agent should systematically produce versus what people must own, and adjusting jobs, training and hiring around that split.
Ross put a sharper point on the human moat: Domain expertise, critical thinking and judgment. Tools are widely accessible; advantage accrues to teams that know what “good” looks like, can interrogate outputs and choose when not to automate. Those same skills mitigate the risk of handing off too much, too soon.
When the discussion turned to ROI, the panel pushed back on simplistic narratives. Dinkin warned that leaders often fixate on cost reduction when they should ask, “What’s our cost per outcome?” Whether the outcome is revenue, pipeline, sign-ups or qualified opportunities, value comes from producing more of it at equal or lower aggregate cost — not merely shrinking a tooling line item.
Zhu agreed and pointed to a common pitfall: Stopping at intermediate metrics like agent usage or page visits. Engagement can be a leading indicator, but the program must ultimately tie to the company’s North Star —revenue growth, margin, customer experience or cost to serve. He recommended designing pilots with attribution in mind and borrowing A/B logic from media to isolate lift. Ross added a cultural nudge: rather than treating AI as a pretext for cuts, progressive teams reinvest efficiency to capture share and elevate people into higher-impact work.
If there was a single throughline to the session, it’s this: Agent-powered marketing doesn’t diminish the human role — it elevates it. Agents expand the surface area of what’s possible; people provide the context, courage and constraints that turn possibility into performance.
The winners won’t merely adopt agents; they’ll redesign how work gets done so outcomes improve — and integrity scales with speed.
Fuel up with free marketing insights.
Contributing authors are invited to create content for MarTech and are chosen for their expertise and contribution to the martech community. Our contributors work under the oversight of the editorial staff and contributions are checked for quality and relevance to our readers. MarTech is owned by Semrush. Contributor was not asked to make any direct or indirect mentions of Semrush. The opinions they express are their own.