Human-agent work is already reshaping marketing. The question is who leads it.

Enterprises are shifting to a new operating model where agents participate in work with marketers to complete tasks and return with final results instead of suggestions.

We call this human-agent work. Rather than a copilot that gets consulted, agents act as true participants and operate in the same systems, draw on the same context and are held to the same standards as your team. However, few enterprises — and even fewer marketing organisations — have the infrastructure ready to adopt this new model.

A capable agent dropped into a marketing operation, without knowing how you operate, what's worked before or who needs to approve what, is essentially a brilliant stranger. It can't absorb organisational knowledge the way a new hire does, so workflows break and every output requires a second look.

BCG found that turning an agent from a brilliant stranger into a trusted partner is what separates enterprises that are seeing returns on their AI investments from those that aren't. Organisations that gave agents the shared context they need saw roughly a threefold increase in productivity and cut cycle times by up to 80%.¹ Skipping this step leaves enterprises stuck spending more time checking an agent’s work rather than reducing the overall time spent.

Implementing AI initiatives can feel overwhelming and easily lead to never-ending proof of concept projects that offer little end-to-end value. Instead, marketing organisations should focus on key improvement areas, prioritising them based on internal needs and potential value.

Adobe customer Moria Fredrickson, VP of Brand and Digital Experience and her team at Lumen took this approach. Rather than chasing every AI use case at once, they ran their AI initiatives against a disciplined, goal-first plan. And the results followed: Lumen now saves roughly 70 hours per campaign execution and $5 million a year on paid media production, as well as launches social campaigns 65% faster — about 15x its pre-transformation pace.

Moria's advice for escaping the proof of concept loop:

• Start with the goal in mind. Decide which projects to progress or phase out only when the KPIs for each of them have been defined before they start.
• Time-box the work into four short phases. Explore, prove, implement and monitor projects over 4-8 weeks, ending in a clear go/no-go decision.
• Begin where the data already exists and in the tools people already use. Point growth-minded people toward what to explore and prove when it comes to AI, then hand off implementation and monitoring to operationally strong teammates to scale a proven POC into end-to-end value.

When teams are ready to ready to scale their AI agents, ensuring the underlying systems are connected and governed will prevent pilots from stalling at the same wall.

The shift towards human-agent work isn’t something enterprises are easing into. According to IDC, 90% of large enterprises are already deploying agents through all three approaches available to them — third-party tools, agents embedded in existing platforms and custom-built agents — and expecting productivity gains north of 30% within the year.³

But dropping an agent into a workflow as-is and expecting it to catch on like a new hire won't work — no amount of enthusiasm gives an agent the context that a person can just pick up. The marketing organisations getting real value out of human-agent work won't be the ones with the most agents. They'll be the ones that gave their agents what they needed to act with confidence.

Thats’s exactly what Adobe Workfront delivers. Most agent deployments today ask enterprises to piece context and governance together on their own. Workfront already has both built in, so agents can be trusted with real work from day one. Combined with AI Collaborators, the leap to a human-agent model can be as simple as assigning tasks to agents your team already trusts, Adobe-built or otherwise. Those agents draw on the same knowledge of the business that any person on the team would, reaching across Adobe and third-party systems to get the job done, with the right people checking in exactly where it matters. The result is a workflow that's fully connected and governed, from the first task to the final decision.

Enterprise context, unified visibility and end-to-end governance helps agents move from being brilliant strangers to productivity-boosting teammates.