The shape of AI adoption

It’s been hard to miss the flood of new data around enterprise AI adoption lately. Analyst reports, customer stories, and product announcements all point to the same conclusion: we’re past the stage of isolated pilots. AI, and increasingly, agentic AI, is becoming embedded into the way organisations actually work.

At Nativevideo, we’ve been watching this evolution closely. Between customer conversations, industry research, and what we’re building with Vinton, one theme keeps coming through: successful AI adoption is not about chasing the latest model or feature. It’s about building systems that combine intelligence, trust, and structure so that AI doesn’t just automate, but augments work.

Let’s look at what some recent research reveals about this shift, and how it connects to the future of AI agents in the enterprise.

According to Capgemini’s Rise of Agentic AI report, nearly one-third of enterprise functions will have AI agents performing daily processes within the next 12 months. Within three years, that figure jumps to 58%.

Perhaps the most interesting part is where these agents are already taking hold. The front-runners are Customer Service and Support (87% expected adoption within three years), IT (84%), and Sales (78%). These are teams where conversational and operational AI has already proven value: automating repetitive work, improving speed to response, and providing better insights.

Tools like Vinton are now embedded in many of these departments, transcribing and summarising meetings automatically, and feeding structured, searchable data straight into Salesforce or other CRMs. What used to be hidden in call recordings or meeting notes becomes usable intelligence, ready for follow-up, reporting, or even training future AI models.

AI adoption is moving from experimentation to execution. What started as innovation projects or pilot programs is now part of daily operations.

Of course, adoption alone doesn’t guarantee success. The Salesforce State of the Connected Customer report highlights a critical insight: users are far more comfortable with AI that assists them than AI that acts for them.

People like AI that helps with scheduling appointments, generating summaries, making recommendations. But when asked how comfortable they’d feel letting AI respond or make financial decisions on their behalf, comfort levels drop sharply.

That distinction matters. Trust isn’t just a matter of accuracy; it’s about transparency, control, and context. The more visible and explainable an AI system is, the more likely people are to trust it.

At Vinton, this philosophy is at the core of how we’ve designed Vinton. It’s not a “black box” that takes over your meeting. It’s a transparent assistant that listens, supports, and captures conversations, all while keeping data safely stored and structured inside Salesforce. It’s not replacing people, but giving them tools that make their work smarter, faster, and more valuable.

The third pattern we’re seeing is organisational. McKinsey’s recent research shows that most enterprises are settling on a hybrid model for AI deployment.

Here’s what that looks like in practice:

• Centralised governance and risk management: — ensuring data quality, compliance, and security stay tightly controlled.
• Distributed innovation at the edges: empowering business functions and local teams to build and deploy their own use cases.

This structure strikes the right balance. It creates guardrails at the core, while letting creativity and speed thrive across teams.

For AI to scale, that model makes sense. It’s exactly how we see customers adopting Vinton: the data and security principles sit in Salesforce’s trusted platform, but the use cases vary widely, from financial services to manufacturing, recruitment to higher education. Each department tailors Vinton to its workflows, but the organisation as a whole benefits from consistent, contextual data.

The next phase of AI adoption isn’t about doing more automation, but about doing it right. That means:

• Grounding AI in context so that it understands the data and environment in which it operates.
• Building trust through visibility and human-in-the-loop collaboration.
• Maintaining clarity and control through thoughtful governance and structure.

At Dreamforce this year, and in countless conversations since, we’ve heard a clear shift in tone. The questions are no longer “Should we try AI?” but “How do we make it reliable, secure, and meaningful for our teams?”

Because in the end, the goal isn’t to have more AI. It’s to have better work.