Your Team Is Doing 'Random Acts of AI.' Here's How to Fix It.
85% of organizations offer AI training. Almost none of them are doing it well. Phyusion's Samantha Stark on what's actually blocking AI adoption, with a four-part strategy that changes it.
Quick summary: An AI adoption strategy isn't a tool rollout. It's a change management campaign. The organizations winning at AI have four things in common:
- The right tools for their specific workflows
- A leadership narrative that gets repeated
- Governance that gives people permission to experiment
- A measurement framework that proves it's working.
Everything else is random acts of AI.
Every B2B marketing leader has been in a meeting when someone shares a stat about AI productivity gains. Heads nod. Someone else mentions a tool their team has been experimenting with. More nodding. Then everyone goes back to their desks and does exactly what they were doing before.
Samantha Stark has a name for this: Random acts of AI. This is what happens when organizations have the urgency but not the AI adoption strategy to back it up. Every team member is poking at a different tool, pilots get launched without a roadmap, and training happens but nothing actually changes.
Stark is an AI adoption strategist who works with enterprises on the full arc of AI transformation, from skills training to multi-agent deployment. She's seen inside hundreds of organizations. Her finding: The technology is rarely the problem. Instead, it's the strategy, the culture, and the change management at the top.
Based on her own research, Stark has a pointed take on where organizations are falling short:
"85% of organizations are offering AI training. They're not training well. If it's not discipline-specific. If your marketer isn't seeing how to build an agent that writes Facebook ad copy, they just don't get it."
–Samantha Stark, AI adoption strategist, Phyusion
The Real Barrier Isn't the Tool, It's the Change Management
The single most consistent mistake Stark sees: Organizations treat AI adoption as a technology rollout. In reality, it's actually a change management campaign. The tools are the easy part. Changing how people work, what they believe is possible, and what they're expected to produce is the hard part.
"This is a massive change management campaign," Stark says flatly. "You can start playing around with the tools, but actually changing people's behavior is significant."
The barriers vary by organization, and Stark emphasizes diagnosing yours before prescribing solutions. Some teams are stuck because of governance gaps where people don't know what they're allowed to do. As a result, they do nothing.
Others have a culture problem: In risk-averse environments, people try a tool once, get a result they don't trust, and revert back immediately. Some organizations simply have terrible training, generic and off-the-shelf, not connected to anyone's actual daily workflow.
The teams that excel share one thing: A leader with a clear narrative about why the organization is transforming. Not a memo or a mandate, but a genuine, repeated, publicly modeled story about where AI fits in the team's future.
⚠ If your leader doesn't understand the tools themselves or use them, then the team will see through it immediately. Stark is direct: Role modeling isn't optional. It's the primary adoption driver.
The Four Components of an AI Adoption Strategy That Works
Four themes kept surfacing as the components of an AI strategy that actually produces behavior change rather than activity reports:
- Tools and capabilities. This isn't about finding the shiniest new platform. It's about ensuring your team has access to the right tools for their specific workflows. You need enterprise-grade versions in place where data sensitivity requires, and someone on the team who understands the genuine capability differences between platforms.
Microsoft Copilot, Claude, ChatGPT, and Gemini are not interchangeable. Treating all of them as equivalent produces mediocre results across all of them. - Culture and narrative. Someone on the leadership team needs to own the AI transformation narrative, not as a one-time announcement but as a sustained drumbeat. What are we trying to accomplish? Why does this matter for our team's careers, not just the company's productivity metrics? What does success look like in six months?
The teams making the most progress are the ones where leadership talks about AI the way they talk about any other strategic priority: Consistently, specifically, and with accountability.
Stark pointed to one CCO who built a virtual Chief of Staff, used it to pressure-test his own vision for the department, and then had his entire leadership team do the same exercise. That’s role modeling at scale. - Governance and risk. Fear of hallucinations, IP exposure, and data leakage are consistently among the top barriers Stark sees in every organization. Ignoring those fears doesn't make them go away. Addressing them with clear guidelines (what's allowed, what's not, what tools are enterprise-safe for which data types) permits people to experiment without feeling like they're putting themselves at legal or professional risk.
Governance doesn’t have to be heavy. A Slack channel for AI questions, a simple intake process for new tool requests, a monthly share-out of what’s working across teams: These create the scaffold that lets curiosity compound rather than dissipate. - Measurement. This is the most neglected component and the one with the highest leverage. If you're not measuring AI adoption, then you're flying blind.
Stark points to metrics like these as what good measurement makes visible: Job satisfaction up 30 points once people feel genuinely equipped, 45% more time for strategic work once execution gets handed off. That kind of data changes the internal conversation from ‘should we do this’ to ‘how do we do more of it.’
"You don't want to plug AI into your existing workflows. You want to reinvent your workflow. That requires discipline."
— Samantha Stark
The Incentive Question Nobody Is Asking
Stark raises a point that stops most CMOs cold: Most organizations haven't set AI-specific incentives for their teams. In a moment when AI fluency is arguably the most career-defining skill a marketer can develop, the implicit message in most compensation structures is that it doesn't matter.
The organizations ahead of the curve are fixing this. They're:
- Adding AI use case development to performance criteria
- Recognizing internal champions publicly
- Building 'AI IRL' showcase sessions into team meetings, not to pressure people but to (a) give the curious ones an audience and (b) make the hesitant ones realize this isn't as intimidating as they thought
One CMO shared at the Strategy Lab that they'd added an explicit element to senior executive compensation tied to three AI use cases that demonstrably drove business value. Not tool adoption for its own sake, but use cases connected to outcomes. That single change, they reported, accelerated organizational momentum faster than any training program had.
A Word on 'Tinkering'
Stark is careful to distinguish between productive experimentation and what she calls going off and tinkering without direction: The pattern where team members spend hours in tools, fail repeatedly, get frustrated, and quietly give up.
The solution isn't to eliminate exploration, it's to give it structure. Protected hours for experimentation, framed explicitly as career development rather than productivity extraction, work best when paired with clear guardrails about what the team is trying to learn and why.
The organizations that get this right don’t just have better AI adoption metrics. They have teams that feel genuinely equipped for where the market is going. And that confidence shows up everywhere, from how people present in meetings to how they attract and retain talent.
Samantha Stark is an AI adoption strategist at Phyusion, working with enterprises on AI skills training, agent deployment, and organizational transformation.
Want more?
For a framework on where your team sits on the AI maturity curve, read: AI Maturity Model for Marketing Teams: The 6 Stages You Can’t Skip.
CMO Huddles brings together senior B2B marketing leaders for candid, peer-to-peer conversations on the challenges that matter most. Learn more.
FAQs about AI adoption strategy
Random acts of AI is the pattern where organizations have urgency around AI but no coherent adoption strategy to back it up.Every team member experiments with a different tool, pilots launch without a roadmap, and training happens without changing actual workflows or expectations. The result is activity without impact: Lots of movement, no momentum.
Stark identifies four consistent barriers: One, governance gaps where people don’t know what they’re allowed to do. Two, risk-averse cultures where a single bad result sends people back to old habits. Three, generic training that isn’t connected to specific workflows. And four, lack of measurement, which means nobody can prove what’s working. Fear of hallucinations, IP exposure, and data leakage consistently rank as top concerns across organizations
Stark recommends building a measurement framework from the start to tracks: Output quality, time savings, team satisfaction, and strategic work ratio. These metrics shift the internal conversation from whether to invest in AI adoption to how to accelerate it.
Stark sees the strongest results when it’s a shared initiative between marketing and IT. Marketing owns the narrative and change management work and IT handles the technical governance and tool infrastructure. The CMO or CCO is often better positioned to lead the communications around AI transformation than a purely technical function. Why? Because behavior change is fundamentally a culture challenge, not a technology challenge.