The New Operating Model: How Agents and Humans Win Together
Last month, a client said something that stuck with me.
“We know we need to take advantage of agentic AI. I’m just terrified we’ll automate away the thinking.”
She was expressing the right concern. But she was looking in the wrong direction.
The risk isn’t that AI agents replace your team’s thinking. The risk is that your organization adopts agentic tools without an operating model built to use them well – and then wonders why nothing changed.
Companies Are Asking the Wrong Question
Most organizations we talk to are asking: “Which AI tools should we buy?”
The better question is: “What operating model and governance do we need so agents can work effectively while humans stay focused on strategy, creativity, and values?”
This distinction matters more than it sounds. Old automation was rules-based. If this, then that. Rigid. Predetermined. An agentic system is different. An agent perceives its environment, makes autonomous decisions, and acts – without waiting for a human to approve every step. It monitors data, recognizes patterns, detects intent, and improves over time.
The practical difference: telling a system “Send this email to Segment A at 9am Tuesday” is automation. Telling a system “Increase conversion for high-intent accounts” – and then letting it identify the accounts, choose the channels, personalize the content, measure results, and iterate – is agentics.
That shift requires a fundamentally different organizational structure. And most organizations don’t have it yet.
Where are you on the maturity curve? We use a four-stage framework to orient clients:
- Stage 1 – Human-led: AI assists at discrete steps. Humans own every decision and all execution.
- Stage 2 – Human-in-the-loop: AI orchestrates and executes within defined parameters. Humans govern, review exceptions, and hold accountability for outcomes.
- Stage 3 – AI-led: AI makes most operational decisions autonomously. Humans set strategy, define guardrails, and intervene on exceptions only.
- Stage 4 – AI-native: Workflows are designed from the ground up for AI execution. Human roles are fully redefined around oversight, creativity, and strategic judgment.
Most marketing organizations today are somewhere between Stage 1 and Stage 2 – and the transition between them is where the operating model either holds or breaks. Critically, not all marketing systems move at the same pace. Your content supply chain may be at Stage 2 while your campaign orchestration is still at Stage 1. That’s not a failure, but rather the norm. Forcing uniform maturity across the organization is one of the most common and costly mistakes in AI transformation.
Why Your Current Model Won’t Support This
Most marketing operations were designed for sequential handoffs. Strategy. Creative. Execution. Approval. Launch. Work moves through departments. Silos are expected. Delays are tolerated.
Agentic orchestration requires continuous cross-functional collaboration. Decisions happen in real time, across boundaries, involving brand, performance, data, and compliance simultaneously. Agents need to act in the moment – and the moment can’t wait for a five-day approval cycle.
We’re in what we’d call the “messy middle.” Old and new processes coexist. Conflicting workflows collide. Technology investments sit underutilized because the organization around them wasn’t redesigned first.
Here’s the specific bottleneck shift most organizations aren’t prepared for: at Stage 1, your constraint is production capacity – how fast can we make things? At Stage 2, the constraint becomes judgment capacity – how fast can the right humans make the right calls on the right content? The volume of content produced increases significantly. The number of human touches per piece decreases significantly. Your org design must anticipate this shift before Stage 2 is deployed, not after.
We see this pattern constantly. Teams move fast. Spend heavily. Still lack clarity on what would actually move the business forward. The tools aren’t the problem. The model is.
What an Adaptive Operating Model Actually Looks Like
An operating model built for agents has three core pieces.
1. Standardized Work: Perfect for Agentics
Agents own the repeatable, data-driven work. But not all content is the same, and being precise about this distinction is critical for org design.
| Content Type | Examples | Execution Owner |
| High-volume variants & adaptations | Banner resizing, email subject lines, social copy adaptations | AI generates at scale |
| Campaign content – mid-tier | Campaign copy, social posts, promotional emails, blog content | AI generates options; human selects, edits, approves |
| Brand-defining / hero content | Brand manifesto, hero campaign copy, tone-of-voice-defining work | Human authors |
| Culturally sensitive / high-risk | Crisis communications, regulatory content, leadership voice | Human owns entirely |
When work is truly procedural – clear rules, consistent inputs, measurable outputs – agents execute it with consistency and scale that humans can’t match. When content carries brand equity, cultural weight, or strategic judgment, humans remain the authors. Organizations that blur this line don’t just create quality risks – they create accountability gaps.
2. Human Judgment Work: Where Humans Own Everything That Matters
Humans own strategy. Creative direction. Brand voice. The work that requires judgment – especially when you need to break a rule because the moment demands it. A journey agent can orchestrate 10,000 personalized experiences simultaneously. It can’t decide when to abandon the playbook entirely.
When organizations are clear about this boundary, something shifts: your best people stop executing and start thinking.
But here’s a risk most organizations aren’t talking about yet: craft atrophy. As AI handles more production, humans reviewing output see less of the creative process and more of the finished product. Over time, without deliberate maintenance, they can lose the ability to judge quality independently of what the AI produces. The feedback loop becomes dangerous: humans approve AI output because it looks right – but their standard of “right” has been calibrated by AI output, not independent creative judgment.
The mitigation isn’t complex, but it is deliberate: maintain a human-only creative practice for brand-defining work. Not because AI can’t do it. Because keeping human judgment sharp and independent of AI influence is itself a competitive asset.
3. Active Governance: The Connective Tissue
Governance isn’t a quarterly compliance review. It’s continuous work at the campaign, audience, and channel level.
One of the most consequential governance decisions in an AI-augmented content system is one that’s rarely framed as a governance decision at all: who defines what counts as “low-risk” content for tiered approval? At Stage 2, AI classifies content as low-risk (queued for batch human approval) or high-risk (escalated for individual approval). That threshold definition is a brand and marketing leadership decision – not a technology configuration. If the AI team sets it, the brand has quietly lost control of a critical editorial decision. This threshold must be reviewed quarterly and updated as AI capability and brand risk evolve.
As AI does more of the work, all roles evolve toward management and governance. Your team isn’t smaller. They’re different.
The Questions You Can’t Skip
Building this model requires conversations most organizations haven’t had:
- Decision Rights: When an agent recommends something that optimizes performance but sacrifices brand consistency, who decides?
- Incentive Alignment: If an agent optimizes for efficiency but damages brand equity, how does compensation reflect that trade-off?
- Tiered Approval Ownership: Who sets and governs the threshold between AI-auto-approved and human-reviewed content – and how often is it revisited?
- New Roles: Have you identified whether you need:
- An AI Content Workflow Director – who owns the AI content production pipeline end-to-end, including exception protocols, override governance, and parameter quality standards?
- An AI Creative Excellence Lead – who deliberately protects human creative judgment from atrophy by running a human-only creative practice and benchmarking AI output against an independent human standard?
- An AI Content Performance Analyst – who monitors how your brand is cited, represented, and recommended in AI-generated environments (LLMs, generative search), and feeds those signals back into content creation?
- Governance Structure: Does governance actually exist today, or are compliance conversations happening informally?
- Literacy and Translation: Who translates between “what the business needs” and “what the agent needs” as instructions? This is a real role gap.
- Accountability: When agents fail or bias emerges, whose problem is it? Marketing? Data? Tech? Legal?
Notice that none of these are technology questions. They’re organizational questions. The real competitive advantage won’t come from the smartest algorithms. It will come from companies that figure out their people and organizational strategy first, then layer in the technology.
Where to Start
The organizations that win don’t start with the tool. They start with three questions:
- What problem are we actually trying to solve? Not “we need to use AI,” but the specific operational pain – speed, volume, personalization at scale, measurement gaps.
- How will we know if we’ve succeeded? What outcomes matter and what must be measurable?
- What combination of people, process, and technology will get us there – and in what order?
Once those answers are clear, we recommend an iterative, hypothesis-driven design cycle rather than a big-bang transformation:
- Design the hypothesis: define a minimum viable operating structure, provisional org design, workflow rules, and human/AI boundaries for one specific process.
- Run a pilot: test the hypothesis with a real team on real work. Generate honest signal, not proof that you were right.
- Learn: synthesize what you observe about structure, behavior, and governance together. Distinguish design failures from implementation failures.
- Redesign: refine the hypothesis based on evidence, then expand.
Map your current workflow. Get clear on governance before you deploy, not after. Identify what’s truly standardizable. Design for the partnership. Then start small: one process, one team, one proof of concept. Learn. Scale.
The Bottom Line
The future of marketing is humans and AI working in concert. Each doing what it does best. Scaling together.
Getting there requires designing your operating model before you implement the technology. Not the other way around. It requires being precise about which content humans own and which AI handles. It requires naming the governance decisions that masquerade as technical configurations. And it requires actively protecting human judgment – not just assuming it will remain sharp as AI takes on more of the work.
That’s the move that wins.
What’s your biggest bottleneck right now? Drop it in the comments.
If you’re ready to think through what an agentic operating model could look like for your organization, that’s exactly what our Strategic Roadmapping services are built for. We assess your current state, identify your real pain points, and build a prioritized path forward. Reach out – let’s talk about what’s possible for your team.
Authored by:
David Iscove, Account Director, Omnicom Adobe Practice & Kathy Haven, Managing Director of Strategy, Omnicom Adobe Practice
