What 21 senior leaders across nine industries revealed about the real barriers to agentic AI — and why experience design is the discipline that determines whether investment delivers
Worldwide spending on AI is forecast to total $2.52 trillion in 2026, a 44% increase year-over-year. For many organizations, a significant portion of this investment will be put toward agentic AI, a transformative technology that is reshaping traditional roles, structures and ways of working.
Budgets are growing. Boards are aligned. And yet, across conversations with 21 senior leaders in industries such as financial services, healthcare, retail, energy and pharmaceuticals, one pattern emerged with striking consistency: the gap between what organizations are building and what they need to build is widening.
The problem isn’t infrastructure. It isn’t even the data. The problem is that using AI to deliver experiences that customers and employees trust, use and return to is a design problem. While most organizations are building the rails, few are designing the ride.
We call this the experience gap. Based on what we heard, it will be a defining challenge of the AI era.
Finding 1: The data problem isn’t one problem. It’s five.
Ask any executive what’s blocking their agentic AI ambitions, and “data” will be part of the answer. Yet the data problem is not a single challenge. It involves five distinct complications that require different solutions.
Data exists, but lives in disconnected legacy systems that were never designed to communicate. The assets are there; the architecture isn’t.
Data was structured for human consumption, not for machine learning. One research leader described an unsuccessful 18-month effort that involved multiple specialist consultants and a significant investment to make a client’s historical data AI-accessible. The data often exists, but the problem is architectural.
Compliance, security and legal requirements limit which systems and data AI can access or act upon. A senior digital leader in commercial banking told us that at least half of their three-year AI transformation timeline is dedicated to navigating governance versus deploying technology.
Documents, audio, field notes, historical contracts — much of the richest data in organizations exist in formats AI cannot efficiently access without new architectural approaches. Several leaders identified unstructured data as their largest untapped opportunity.
Some of the most valuable behavioral signals for agentic personalization were simply never captured.
The organizations making the most progress have diagnosed which specific problems they have. The ones stalled are treating data as a monolithic wall, when what they’re actually facing is five different doors, each requiring a different key.
Digital Experience, health insurance sector
The biggest opportunity we’re seeing is optimizing unstructured data — documents, images, audio and customer interactions. The challenge isn’t AI capability anymore; it’s storing and retrieving that information so it can actually contribute to outcomes.
Finding 2: AI is only as good as the people working alongside it.
Every organization we spoke with has a policy for human involvement in agentic AI processes. Healthcare providers cite regulatory guidance. Banks cite risk thresholds. Most others cite trust and accountability. The positions are clear and often carefully considered.
What almost no one has done is design what it means for employees to work alongside AI — and how that directly shapes the customer experience.
Think about a relationship manager who uses AI to surface a customer’s full interaction history, flag relevant context and prepare for a conversation. In addition to the manager being better informed, the customer interacts with someone who is fully present. Rather than replacing the manager, AI helps them succeed in the moments that matter most.
This model only works when the employee experience has been designed with the same care as the customer experience it is meant to improve. Without thoughtful workflow design, employees are slow to adopt AI tools, they don’t trust the outputs, and they default back to the manual processes that were intended to be replaced.
A healthcare leader captured the gap directly, reflecting on their own program:
Chief Transformation Officer, healthcare sector
We put immense care into the patient-facing agent but almost none into the employee-facing UX. Seems silly now that I think about it.
Neglecting the employee experience is the norm across organizations – exacerbating the challenges for employees whose roles are being fundamentally reshaped by AI deployment. One organization addressed this by building trust incrementally through evidence rather than argument. The organization created a mechanism that tracked AI recommendation accuracy over time and surfaced the data directly to skeptical employees. Sure enough, adoption followed, thanks to an experience design intervention disguised as a technical solution.
Finding 3: Your brand is no longer a guideline. It’s code.
If an AI agent writes to a dissatisfied customer, makes a recommendation to a worried patient or declines a request, what does your brand sound like in that moment?
The most mature organization we spoke with, a Tier 1 bank, encountered this challenge when deploying agentic AI-generated customer communications at scale. Leaders learned they needed to define their brand with a new level of precision. To do so, they asked questions like, ‘What words do we use?’ ‘What reading age?’ ‘What emotional register?’ ‘What do we never say?’ Their answers informed operational parameters enforced through machine-executable code that governed AI behavior in ways a brand guideline cannot.
Head of Data and Analytics, banking sector
In a world of AI, you need to double down on understanding what your brand means if you want the AI to represent it.
For the bank, the process forced a level of brand clarity described as more valuable than anything prior brand strategy had produced. And when AI-generated complaint letters were blind-tested against human-written ones, the AI-authored versions consistently came out ahead.
This example extends beyond communications. As AI agents interact and respond at scale, the gap between intention and experience is determined by the quality of AI design.
Finding 4: A pivot separates leaders from laggards.
Across 21 conversations and nine industries, the organizations delivering the most measurable impact from agentic AI share almost nothing in terms of sectoral priorities, technology stacks, geography or budget. But they do share a single pivot.
At some point in their AI journey, each stopped asking what technology can do and started asking what the experience should be.
The distinction sounds simple. The implications are not. Organizations that begin with technology reach for existing processes and automate them. Organizations that begin with experience redesign the process before they touch technology.
COO, technology platform sector
We don’t want to automate what we have today. That’s not anywhere near best practice of where we want to go.
This pivot changes how use cases are selected, and also changes how success is measured. Organizations shift from asking where volume is highest to where a better experience will create the most value. The most sophisticated metric in this study wasn’t a Net Promoter Score (NPS) or call deflection rate signaling efficiency. It was tracking whether customers accepted an AI-generated recommendation or edited it out — a direct signal of trust.
Most importantly, this pivot changes what gets built. The organizations that have made the pivot are building experiences rather than automations. The difference compounds over time.
The clearest example of this pivot came from one of the most advanced organizations we encountered:
Head of Data and Analytics, banking sector
AI is just an enabler that sits within our change portfolio. They’re not AI projects. They are business-driven projects that would have existed with or without AI.
That is what it looks like when the pivot has been fully made. AI becomes a tool in the service of an experience strategy. And across customer satisfaction, operational performance and commercial results, the outcomes reflect exactly that.
What this means for today’s leaders
Technology is not the differentiator. Every organization in this study — regardless of size, sector or investment level — has access to comparable infrastructure, models and vendor ecosystems.
But the experience gap is widening. The organizations pulling ahead are doing so because they asked the right design-centered question first and built the organizational capability to answer it. Organizations that embedded design-thinking into how they select use cases, structure human-AI interaction and encode brand into systems are delivering on the promise of agentic AI.
On the other end of the spectrum, the organizations that treated experience design as optional suffered more failed pilots, employee adoption crises and brand moments that AI handled in ways no guideline had anticipated. These challenges left them struggling to keep up with competitors and increased their exposure to costly mistakes.
As organizations look to close the experience gap, they must ask whether each agentic encounter is truly optimized, humanized and holistically designed across its full lifecycle. But a more fundamental question precedes any framework: what are customers and employees actually experiencing every day?
