-
Sapient Bodhi ranked #1 for deep research.
-
Sapient Slingshot wins Technology Breakthrough of the Year.
-
3M lines of COBOL → clear specs in 8 weeks.
-
Meet Sapient Sustain: context-aware AI for complex IT operations.
73% of Companies Use AI. Only 10% Say It’s Core to How They Operate
New Publicis Sapient research maps the distance between AI adoption and real enterprise impact.
August 3, 2026
In this article
Generate AI Summary
Loading AI-generated summary...
Most enterprises have already proven that AI works. A pilot in marketing here, one in sales or on the production line there; a developer shipping code faster with an assistant at their side. The technology delivers. The trouble starts when a company tries to take those isolated wins and run them across the whole business.
That second step is far harder than the first, and the obstacle is usually organizational.
“AI doesn’t have an innovation problem. It has an execution problem,” says Nigel Vaz, CEO of Publicis Sapient. The models are ready and the capital is flowing, but Vaz argues the real constraint is the organization itself.
Our new research backs that up. In a global survey of 1,550 enterprise AI decision-makers, Publicis Sapient found that 73 percent of companies now use AI regularly or in most processes, while only 10 percent say it is core to how their business operates. AI has spread through everyday work while the operating model underneath it has barely changed.
Leaders are clear that the technology is ready. Nearly half (47 percent) say AI is already fully capable of meeting their current business needs. At the same time, 42 percent say their organization isn’t structured to capture the value AI can already create. Asked to name the single biggest constraint on AI success, leaders were twice as likely to blame the way their organization runs (22 percent) as the capability of AI itself (11 percent).
In Vaz’s words: “In order to actually scale AI, you have to think about how AI is going to fundamentally redesign how people and platforms in an organization interact to create value, not just very simply apply AI to either an existing workflow or an existing area of the business.”
Your AI pilots work. But can they scale?
Picture a single developer writing code with AI tools. It works beautifully. Now scale that to thousands of developers across an enterprise, each on a different team, with different permissions and their own slice of context. The consumer-grade tools that felt effortless for one person start to break. Whatever platform they share has to hold context for every role and every handoff, or the value never materializes. Vaz calls this a common execution failure point.
He sees a historical pattern. Electricity and the internet only paid off once companies redesigned how they operated around them; early in the last century, some executives were effectively chief electricity officers, rebuilding the entire factory floor around a new power source. AI sits at the same inflection point. Treating it as a side project bolted onto legacy systems all but guarantees the impact stays small.
The survey shows how common that mistake is. Even with AI usage widespread, fewer than one in five enterprises (18 percent) say AI is fully integrated across the business. When leaders name what most limits their ability to scale, they point to the connections between systems: integration across systems (38 percent) and fragmented data (36 percent) top the list, followed by governance and organizational silos.
“You have to be really intentional about redesigning your operating model around AI rather than simply deploying it,” Vaz says. “In the context of how things work today, you need to be thinking about this whole notion of how a digital workforce and a human workforce can start to work together.”
A vaccine rollout shows what scale really demands. Getting an approved vaccine to market across 150 countries means reconciling each country’s rules on what a company may say about it, a negotiation between legal and marketing teams that can stretch launch timelines into months or years. But with agentic AI, agents can ingest the marketing content on one side, and the regulations on the other. The system flags a conflict, regenerates compliant content and then routes it to a person for final sign-off. A process that once took months collapses to weeks.
It’s time to operationalize intelligence
This example highlights something Vaz considers non-negotiable as companies move toward agentic AI: clarity about infrastructure, governance, data integrity and where a human belongs in the loop. Sometimes a person is there only to approve. Sometimes the work is human-led with AI support. Sometimes it runs on its own. “If you aren’t clear about these components, AI ends up creating more ambiguity than acceleration,” he says.
The gap between ambition and readiness is wide. Nearly two-thirds of leaders (65 percent) expect significant progress scaling AI over the next 12–24 months, while just 15 percent say their organization is fully equipped to support that today. That’s a 50-point gap between where enterprises want to be and where they are.
Vaz’s advice to leaders is to start with value. He suggests asking what the business is actually trying to move—market share, growth, cost or speed—and reimagining the work around those goals, prioritizing use cases by the business impact they create.
What the companies pulling ahead do differently
Most enterprises now have access to the same models, cloud providers and copilots. The difference increasingly comes down to how they adapt around them. The organizations making real progress are the ones rebuilding how work moves through the business. Across the survey, three shifts set them apart:
- Modernization: they modernize legacy systems earlier and embed operational context, rules and institutional knowledge directly into their AI systems, so no new initiative starts from scratch
- Coordination: they connect workflows across functions so decisions move with fewer handoffs, and they set enterprise-wide strategy to coordinate AI across teams
- Resilience: they create visibility across AI activity and outcomes, reduce operational debt and remove the repetitive work that slows execution
That thinking has shaped how we approach the execution problem. Drawing on more than 30 years of experience in enterprise transformation, we’ve built three platforms aligned to these shifts from the start: Sapient Slingshot for software development and legacy modernization, Sapient Bodhi for agentic orchestration grounded in industry expertise and Sapient Sustain for AI-run IT operations. Built for the enterprise from day one, they carry what the early pilots never had: the permissions, security and operational context that scale demands.
The companies that win the next phase of AI, Vaz suggests, will be the ones that engineer intelligent systems and human judgment to operate as one. He reaches for Iron Man to make it concrete: the suit alone is just a suit, a person alone has limits and the power comes from bringing the two together.
Our survey of 1,550 enterprise AI decision-makers shows where the gaps are and how the companies pulling ahead are closing them.