Over the last year, we've spent a lot of time rebuilding how our marketing organization works with AI. What surprised me most is that the biggest changes were not creative, they were operational. Most conversations about AI in marketing still focus on tools: copy generation, image generation, campaign automation. Those things matter, but they are only part of the story.
The larger shift is structural. AI changes how information moves through an organization, how decisions get made, and how teams coordinate work. The marketing teams getting the most leverage from AI are not simply layering it onto existing workflows, they are rethinking the systems underneath them.
Across our team, four areas consistently mattered most: context, coordination, creativity, and control. None of these are entirely new ideas. What changes with AI is the importance of getting them right and how to bring them to life.
1. Context: Making Information Usable
The biggest bottleneck inside most marketing organizations is not creativity, it is fragmented information. AI exposes this quickly. If strategy lives in one place, reporting in another, and institutional knowledge inside Slack threads and meetings, outputs become inconsistent fast.
We started by standardizing context.
Our team built a Marketing Context Library: structured documentation covering strategy, audiences, positioning, workflows, active initiatives, and brand standards. The system is organized into four layers: Foundation, Channel, Standards, and Initiative. Both people and AI systems pull from the same source.
Documentation became infrastructure. One example is our Weekly Marketing Rollup, a company-wide summary of performance, launches, priorities, and issues across the organization. Every function contributes in a consistent format. Beyond helping with alignment, it has become a primary reference point for AI-assisted reporting and analysis across the company.
Data accessibility mattered too. Our team can query systems like Mixpanel, Meltwater, Cision, Braze, and Google Analytics without needing SQL expertise because we built standardized access layers and clear source hierarchies underneath them. Without that structure, systems produce conflicting answers and waste time searching fragmented sources.
All of these revised inputs allowed us to drastically simplify measurement and definitions of success. Instead of dozens of disconnected KPIs across teams, we aligned around a smaller shared scorecard focused on revenue contribution, conversions, and sentiment.
A useful test for context is simple: could a new marketer understand how your organization works on his or her second day? If not, the foundation probably needs work.
2. Coordination: Moving Beyond Automation
Once context is reliable, the next challenge is coordination.
For most teams, this starts with automation. Simple workflows create a surprising amount of leverage: notifications triggered from launches, routing systems for approvals, AI-assisted drafts waiting for human review before publishing. Our team usually ships these kinds of workflows within days of identifying the need. The gains are incremental individually, but meaningful in aggregate. Less manual coordination. Faster execution. Fewer operational bottlenecks.
The next layer is systems that can operate across multiple steps. At MetaMask, we use a small network of systems to help support our market-focused editorial platform: MetaMask Alpha. These systems gather market signals, synthesize inputs, and help draft content for editorial review. Humans remain deeply involved, but the workflow significantly expands the team's capacity. All in, it allows us to respond to market trends and inform our users with information that matters to them in a timely manner.
We have also built an internal brand agent that can answer questions about tone, messaging, and positioning directly inside Slack. That may sound experimental, but reducing friction around brand execution has had a measurable impact on speed and consistency.
A common mistake is trying to jump directly to more advanced systems before the underlying workflows are stable. Most organizations still have meaningful gains available in basic coordination work alone.
3. Creativity: Increasing Iteration Speed
The largest creative impact from AI has not been idea generation, it has been iteration speed. Teams can explore more directions, prototype faster, and refine work against live feedback loops much earlier in the process.
We have seen this most clearly in briefing. Better structured inputs consistently produce better work. When briefs incorporate live performance context, audience signals, and historical learnings, the quality of creative discussions improves substantially.
Prototyping speed is another major shift. Teams can now pressure test multiple campaign directions in days instead of weeks. The output is rarely final creative. The value is expanding the range of possibilities worth evaluating.
Continuous feedback loops also change how creative teams operate. Sentiment and audience reactions that once appeared quarterly now arrive continuously, through tools like Brandwatch and Meltwater, feeding directly into planning and iteration cycles.
What has not changed is the importance of judgment. Taste still matters. Positioning still matters. The ability to recognize what feels culturally relevant, emotionally resonant, or distinct remains deeply human work. The strongest teams are using AI to expand creative possibilities.
4. Control: The Difference Between Experimentation and Trust
Control is probably the most overlooked part of this transition. In practice, it determines whether AI systems become trusted or quietly create organizational resistance.
For us, this starts with trust thresholds.
Every AI-assisted workflow is assigned a level of autonomy before launch. Some systems can operate independently. Others require approval from a channel lead or senior review before anything ships. We also measure performance rigorously. AI-assisted outputs are tested against human-only baselines across major workflows. If the system is not improving outcomes, we deprioritize the use case and revisit it later.
Finally, every workflow maintains an audit trail. Inputs, prompts, outputs, and decisions are logged so issues can be traced and improved over time. Most AI programs do not fail because of one catastrophic mistake. They fail because small inconsistencies gradually erode trust across the organization.
Control is what prevents that from happening.
Most organizations today are still uneven across these areas. Context is incomplete. Coordination is fragmented. Creative workflows are evolving. Governance often lags behind, that is normal. But the shift underway is becoming clearer: AI is no longer just a tooling decision for marketing organizations. It is increasingly an operational and organizational design decision. The teams that recognize that early will have a meaningful advantage over the next few years.
A special thanks to Rosalie Forman Samuels, Head of Growth and Nick Nelson Head of Brand & Content for their contributions to this article and to our excellent marketing team at Consensys for their contributions to the thinking and systems outlined here. And yes, this post was drafted with the help of LLMs.