Moving from AI ambition to governed, operational workflows faster
TL;DR Summary (click to expand)
Most enterprises are still early in their AI transformation journey, and the biggest barrier is no longer interest in AI, but operational execution. According to Notion’s Great Renovation report on AI transformation, 88% of organizations remain in the early stages of AI maturity, and only 12% have embedded AI into recurring workflows or autonomous operations. In my experience, organizations accelerate agentic AI transformation when they embed technical talent close to the business, focus on a few high-value workflows first, and build with governance and measurable outcomes from the start.
Marketing teams are seeing the same execution gap in real time. As EMARKETER reports, 76% of global marketing leaders spend at least 3 hours per week editing or correcting AI outputs, and only 4% say AI saves time at every stage. This tells me that the problem is not access to AI. It is the absence of connected systems, clear workflow design, and the operating model needed to turn experimentation into reliable business value. Forward Deployed Engineers help close the gap by embedding execution, integrations, and governance into the workflows that matter most.
Most enterprises are still early in their AI transformation journey, and the biggest obstacle is no longer interest. It is execution. In my experience, organizations accelerate agentic AI transformation when they embed technical talent close to the business, focus on a few high-value workflows first, and move from experimentation to operational reality with clear governance and measurable outcomes.
Enterprise AI strategy is not in short supply. Almost every leadership team I speak with has a point of view on AI, a list of use cases, and some level of urgency around doing more with it. What is much harder to find is the operational muscle to turn this ambition into working systems that people actually use. The gap is especially visible in the current wave of agentic AI. The potential is easy to see. Teams can imagine AI agents orchestrating workflows, connecting systems, surfacing insights, handling repetitive tasks, and supporting better decisions across the business. However, getting from this vision to something reliable, governed, and useful inside a real enterprise is where most organizations slow down.
According to Notion’s Great Renovation report on AI transformation, 88% of organizations are still at level 1 or 2 of their AI transformation journey, where AI is used more as a personal productivity tool than as part of recurring workflows or autonomous business operations. Only 12% have reached levels 3 or 4, where AI is embedded into workflows or running parts of the business autonomously. In other words, most companies are still much closer to experimentation than transformation.
This does not mean the ambition is wrong. It means the operating model usually is.
AI Adoption Headwinds Are Real, and They Are Familiar
When agentic AI efforts stall, the root cause is rarely lack of ideas. More often, it is a collision of technical, organizational, and operational headwinds.
Organizations are trying to navigate a fragmented ecosystem of vendors and frameworks. They are managing skill shortages at the same time the technology stack is evolving in public. They are asking internal teams to connect models, systems, security requirements, knowledge sources, and workflow logic, often without enough embedded technical capacity to do all of it well.
The same report highlights another important tension: 55% of AI decision-makers say their organization is investing in AI faster than employees can learn it. At more advanced organizations, the number climbs to 62%. I think that statistic matters because it exposes the real bottleneck. AI transformation is not just about buying access to models or approving innovation budgets. It is also about helping the organization absorb, operationalize, and trust new ways of working.
Confidence gaps make this harder. The report found that 49% of AI decision-makers are very or extremely confident in their organization’s AI capability, while only 23% of AI users say the same. This disconnect matters. Leaders and operators often experience different realities, and any AI strategy built only from the executive view misses the friction inside day-to-day workflows.
You can see the same pattern in marketing organizations. As EMARKETER recently highlighted, 76% of global marketing leaders spend at least 3 hours per week editing, fact-checking, or correcting AI outputs, and only 4% say AI saves them time at every stage. The biggest time killers are not surprising: hallucinations and fact-checking at 48%, copying and pasting between disconnected tools at 40%, and compliance checking at 37%. Even more telling, 65% say they would pause or adjust their company’s AI implementation plans if they had the chance to improve guardrails and governance or rethink their operating model. To me, this is the clearest sign that adoption is still outpacing operational maturity.
This matters for marketing teams in particular because speed alone is not the win. If AI just shifts work into more review cycles, more tool friction, and more brand risk, then the organization has not really become more efficient. It has simply moved the labor around. This is exactly where adtech and marketing operations experience matters. Teams need connected systems, clearer workflow design, and stronger operational guardrails if they want AI to deliver measurable business value.
That is why so many teams get trapped in what I think of as expensive experimentation. They run pilots, produce demos, and identify promising use cases. But they are unable to create a repeatable path from concept validation to operational deployment.
Governance is another major reason the path breaks down. Survey results in Deloitte’s 2026 State of AI in the Enterprise make the point clearly: 73% of respondents cite data privacy and security as a top AI concern; 50% point to legal, IP, and regulatory compliance; 46% note governance and oversight; and model quality, consistency, and explainability account for another 46%. These findings show that governance is not a late-stage review step. It is part of what determines whether an AI initiative can move from interesting pilot to trusted operational system.
In practice, this means teams need more than enthusiasm and tooling. They need a way to build guardrails, clarify ownership, and design workflows that can stand up to security, compliance, and operating scrutiny from the start. This is where Forward Deployed Engineers can help. A good FDE does not just help build the workflow. They help shape the operating model around it so the work can actually move forward.
Why Forward Deployed Engineers Change the Pace
In my experience, the fastest path through the gap is embedded execution.
Forward Deployed Engineers (often shortened to FDEs) are hands-on technical operators who embed closely with a client team to help design, build, integrate, and operationalize complex solutions inside the real environment where the work happens. In practice, this means they sit between strategy and delivery. They are not just outside advisors writing recommendations or isolated engineers building in a vacuum. They work shoulder-to-shoulder with business and technical stakeholders to turn agentic AI from an abstract initiative into a live operating capability.
When organizations rely only on internal engineering teams, AI work often gets pushed behind competing priorities. When they rely only on consultants, they may get strategy artifacts without working systems. Forward Deployed Engineers sit in the middle of this gap. They bring engineering-grade execution but stay anchored to business outcomes.
That is especially important in multisystem enterprise environments. Agentic workflows rarely live inside one tool. They depend on data access, model behavior, permissions, integrations, guardrails, and handoffs across teams. The work is not just building an agent. It is orchestrating an environment where the agent can operate reliably.
This is where an AI orchestration layer (aka AI harness) such as StationOne is foundational as part of the execution layer. In the right environment, it gives teams a structured way to connect models, skills, knowledge, guardrails, and workflows without starting from scratch every time. But even with the right platform, somebody still has to translate business intent into operational design. This translation layer is where embedded Forward Deployed Engineers create outsized value.
The most successful transformations I have seen do not begin with a broad declaration that the organization is becoming AI-native. They begin with a narrow workflow that matters.
Start With Workflows, not Abstractions
The most successful transformations I have seen do not begin with a broad declaration that the organization is becoming AI-native. They begin with a narrow workflow that matters.
This could be a media operations process that consumes too many manual hours. It could be a campaign management task that depends on too many disconnected systems. It could be a reporting motion that delays decisions because nobody trusts the data flow. The point is not to start with maximum scope. It’s to start with a workflow where the value can be made visible quickly.
Notion’s Great Renovation report on AI transformation also found that the most advanced organizations separate themselves by integrating AI into existing systems, building governance and oversight, and tracking impact with defined metrics. These are not side considerations. They are the mechanics of successful adoption. When teams start with a real workflow, they are forced to answer the questions that matter early: What systems need to connect? What guardrails are required? What role should the human operator still play? How will we measure whether this is working?
That is why I believe Forward Deployed Engineers can accelerate transformation so effectively. They help organizations rally around concrete use cases, reduce the distance from insight to value, and build the operating discipline that makes scale possible later.
Agentic AI transformation does not need more broad enthusiasm. It needs more working examples, more embedded builders, more operational follow-through.
For organizations trying to close that gap, the opportunity is real. So is the complexity. But in my experience, the teams that move fastest are not the ones with the boldest AI messaging. They are the ones that get the right people close to the right workflows early, build something useful, and learn by operationalizing.
If your organization is working through that challenge now, I believe it is worth a conversation. Contact us for a huddle.
About the Author
Seth Samuels is General Manager of Kochava Foundry, where he helps enterprise organizations turn AI ambition into operational reality. His work focuses on embedded technical execution, agentic workflow deployment, and designing practical systems that create measurable business value.
