While companies pour millions into developing sophisticated AI capabilities, most relegate the user interface to an afterthought, causing technically impressive AI platforms to sit unused while simpler, more intuitive tools see widespread adoption. In practice, it’s often not technical limitations but poor UX choices that prevent enterprise AI from delivering real business value.
Shilpi Bhattacharyya, a product strategy leader at IBM and recipient of the TITAN Innovation Award for AI-Driven Product Management, has seen this dynamic play out repeatedly. Over the past several years, she’s led work on flagship platforms like IBM’s Cloud Pak for Data and its AI-driven infrastructure services, helping bridge the gap between what AI can do and what users are willing, or able, to adopt. Her experience offers a window into how companies can turn AI from a sunk cost into a powerful, practical asset.
When AI Is Built—But Not Used
One of the largest risks in enterprise AI is abandonment. Bhattacharyya points to a troubling but consistent trend: There’s roughly a 30% chance that any large-scale AI deployment will be shelved after rollout. And it’s not just a matter of technical failures. According to McKinsey, more than 80% of organizations aren’t seeing tangible impact from their use of gen AI. In many cases, the problem lies in usability.
These challenges intensify in complex business settings where AI must fit into existing processes and win over staff with varying degrees of expertise. Financial analysts need risk tools that don’t require specialized knowledge. Healthcare providers need diagnostic support fitting naturally into patient consultations, while manufacturing staff rely on maintenance alerts they can trust without extensive training. The result is a widespread adoption crisis, where AI initiatives that should be successful on paper fail to take root.
Prioritizing the User Experience
Companies that succeed with enterprise AI are increasingly those that treat user experience as a strategic lever rather than a finishing touch.
“AI adoption is much more of a human puzzle than a technical one,” Bhattacharyya notes. “We spend years refining our algorithms at IBM. But what moved the needle was rethinking how people actually interact with these systems, and rebuilding the experience around that.”
Instead of layering design on top of completed engineering work, Bhattacharyya’s teams embedded designers within development groups from the start of the project. They spent time with users across industries, from financial analysts to plant technicians, to understand the real-world context in which AI tools are used. That input shaped stripped-down interfaces that made powerful capabilities more discoverable and usable.
A redesign of IBM’s Cloud Pak for Data platform led to a 40% increase in Net Promoter Score and an 8% boost in conversion rates. More importantly, license renewals jumped 28%, translating directly into retained revenue.
“The eye-opener was seeing those improvements flow straight to our bottom line,” Bhattacharyya says. “For every 10-point increase in NPS, we saw a 7% bump in renewals. That makes a case for UX investment much stronger than anecdotal feedback can.”
This improvement also confronts a broader truth in enterprise software: most tools are only partially used. Studies show that up to 80% of features in cloud-based platforms go untouched—a sign that users are often overwhelmed or simply unaware of what’s available to them. For companies paying full price for underused software, that can lead to unnecessary overhead costs.
Solving for Trust & Usability
In AI infrastructure tools, where systems monitor massive volumes of backend data and flag potential issues, the design problem is even more acute. “These systems were originally designed by and for engineers,” Bhattacharyya explains. “You would get dense dashboards full of raw, difficult-to-interpret metrics. It was like asking someone to read their own MRI.”
Her team adopted a different framework: treat system alerts like clinical diagnostics. Instead of presenting all the data, they highlighted critical issues, suggested likely causes, and offered actionable next steps. The focus was on clarity and confidence. “We moved away from showing everything to showing what mattered,” she says. “The goal was to make insights legible to the people responsible for acting on them.”
When AI Is a Commodity, UX Becomes the Edge
Looking ahead, Bhattacharyya predicts that core AI capabilities will become increasingly commoditized. As generative models and predictive engines converge in performance, the question will be who offers the most usable product.
That shift puts a premium on user experience. “The companies that do well will treat design as part of product strategy,” she says. “It means bringing UX professionals to the table with data scientists and engineers from day one. It also means redefining success metrics to include how well users understand and trust the tools they’re given.”
If AI is to become more than a technical showcase, enterprises need to prioritize usability as much as scalability. The platforms that gain traction will be the ones people can easily integrate into their work, not the ones that simply check a box on a capabilities list. In an era of growing AI parity, the human experience is what becomes the competitive edge.
