Teams have already made AI part of their daily routine. The operating model, however, is still catching up.
That’s the core story in Cella’s latest Expert Perspectives Report. We surveyed creative, marketing, UX, operations, and strategy teams across 17 industries to uncover what’s actually happening on the ground with AI adoption.
Here are four key findings from the data and what teams can do about them.
1. AI adoption is nearly universal, functioning primarily as a writing and thinking tool
With 82% of professionals using AI daily or more, it’s clear AI is no longer experimental; it has firmly embedded itself into the creative and marketing workflow.
When we dig into exactly how it’s being used with such prevalence (and where the surprising gaps lie in visual tools), the data shows it clusters heavily around words and ideas. Because teams are treating AI as a key partner in their workflows, the quality of the input dictates the quality of the output.
From insight to action:
- Moving past basic, one-size-fits-all prompting unlocks much better results. Teams see real gains when they match their prompting level to their output expectations.
- Prompting manually can hit a wall at the team level. The solution is context engineering: building tone guidelines, personas and project histories that the AI draws on automatically, ensuring everyone stays aligned without starting from scratch.
- AI agents are the next frontier, but the safest path is to start with copilots and gradually earn your way to autonomy.
2. Everything got faster, but quality control requires more focus
More than any other gain, 74% of respondents named speed as AI’s biggest impact. However, that acceleration requires a new approach to review: 57% cited quality control as the biggest workflow challenge AI has introduced. In other words, there are concerns that the work could suffer.
Faster output naturally means less control, more material to review and risks to counter. To maintain high standards at this new speed, intentional checkpoints are essential.
From insight to action:
- The most effective creative quality control happens before generation even starts. Structuring context upstream with clear creative leadership at the helm means significantly less cleanup downstream.
- Creative review cycles work best when they are deliberately designed. Decide which tasks can be automated, which are shared, and which remain human-only, and build a checkpoint for each.
- Layering AI over a broken workflow often just amplifies existing friction. Streamline the process first, then let AI accelerate it.
3. Measurement, governance and ownership are the missing pieces
Despite widespread daily use, 43% of teams say they don’t measure AI’s impact at all, and an identical 43% say there is no clear owner. These two gaps reinforce each other: without an owner, nobody is accountable for measuring, and without measurement, the case for an owner never gets made.
From insight to action:
- IT and Legal will naturally own security and privacy, but marketing and creative leaders need to own how AI touches the brand.
- Tallying up prompts and drafts doesn’t capture the full value of your AI. Looking at four distinct kinds of return—financial, operational, relational, and strategic—tells a much more accurate story.
- Governance works best on two tracks: creating safe sandboxes for internal teams (like sales and HR), and requiring firm human sign-off for anything public-facing.
4. The work has changed, but the titles haven’t
The skills professionals say matter most are judgment and strategy, not simply mastering new tech tools. Yet brand-new AI roles are still rare. Instead, teams are evolving through upskilling and training.
People are redefining their daily work without redefining the organizational chart, which helps explain why 41% say they want support in redesigning how their team operates.
From insight to action:
- As AI absorbs repetitive production tasks, roles naturally move up the value chain, and upskilling provides the bridge to get there.
- Fostering psychological safety and hosting low-stakes experiments like demo days do far more to encourage healthy adoption than strict mandates.
- Transitioning from ad-hoc experiments to a mature operating model involves five deliberate steps, starting with matching the right kind of AI to the specific task.
- Proactive teams that lead adoption get the opportunity to help design policies that genuinely support their workflows.
Conclusion
Adoption is no longer the hurdle. The questions that matter today are about ownership, effective measurement, and maintaining quality as velocity increases. Explore the full report to see the full breakdown, benchmarks, team sentiment data, and the three steps worth taking now.
If you’d like support turning these insights into action for your team, get in touch with our experts today.
