Zuora CIO Karthik Chakkarapani is an AI-driven transformation leader with deep experience modernizing global enterprises.getty​Despite the rapid adoption of AI tools, there remains a debate about how useful AI is to employees. ​​Imagine you need to prepare for a customer meeting. How much time does it take to assemble and synthesize account information, develop context, and get insights and actions? How many different systems do you have to use to get that done? You log in, export files, switch between tabs and manually combine everything into a useful document. Maybe you asked a trusted colleague to help, but that still means going back and forth to get it right.​For many organizations, ​fit-for-purpose enterprise AI solutions are helping to remove that overhead, making old processes obsolete rather than just improving them bit by bit. Rather than focusing on which systems you use, many of today's solutions focus on what you want to achieve. This means moving away from manual work and toward smart automation, as the new way of working is increasingly built around user intent. ​How have enterprise AI models evolved?​For years, enterprise software promised powerful systems, and users were expected to learn how to use them. Every new platform meant training, change management and time for employees to get used to new tools so they could do their jobs.​ The result was compounded across two decades of SaaS adoption in an enterprise technology environment of both extraordinary capability and extraordinary friction.​The problem is that these systems, while very powerful on their own, weren't always great at talking to each other, and even a single business decision required data from multiple systems, including CRMs, ERPs, analytics platforms, customer support systems, collaboration tools and the spreadsheets that stitched them all together.​I’ve called this the “plumbing” problem. Enterprise software was built to give organizations the ability to manage complex operations consistently across distributed teams and geographies, and it largely succeeded at that. What it didn’t optimize for was the human experience of working within it.​Employees became, in effect, operators of the pipes, required to spend significant portions of their working day on the overhead of navigating the system rather than on the substance of the decisions and outcomes they were hired to drive. ​How does enterprise AI change workflows?​Enterprise AI solutions are changing how people interact with technology, which is an exciting development.​ In this new model, work starts with asking, "What outcome do I need?" instead of "Which system should I use?" Finance executives forecasting the impact of a 5% revenue slowdown no longer build the scenario manually. Instead, they describe it, and the AI analysis runs independently.​Properly designed enterprise AI solutions remove the interface of the underlying SaaS ecosystem, replacing system navigation with natural language and application logic with outcome-based interaction.​How is modern SaaS evolving?​I believe that all the hysteria around the SaaSpocalypse misses the point. SaaS isn’t going away. SaaS will move from user-facing to unseen infrastructure.The systems of record aren’t going anywhere, either. CRM platforms hold the customer relationships and pipeline data that revenue teams depend on, while ERP systems enforce the financial controls and operational processes that regulated enterprises can’t operate without. Data infrastructure underpins every meaningful AI output. Put even more simply, these are the plumbing pipes carrying water and they’re not optional.​What’s changing is how employees access these systems. AI acts like a faucet, giving people what they need without interacting with the complex infrastructure underneath. The real advantage comes from how well the intelligence layer connects everything.​Why do data foundations matter now more than ever?​If AI is the faucet, then data quality is like water pressure. Unfortunately, in most companies, that pressure isn’t steady.​ If your data is scattered, your AI solutions will be too, leading to incomplete or unreliable insights. Poor data quality erodes trust in AI workflows. If employees get a bad recommendation from AI, they’ll go back to manual work. ​Moreover, APIs are the pipe connections that serve as the integration points that allow AI to move fluidly across the SaaS ecosystem without hitting dead ends. Interoperability standards are the pressure regulators, ensuring data moves between systems at the right volume, in the right format and with the right permissions. And governance is the local building code that provides the framework to ensure everything is connected correctly, access is controlled appropriately and that the system as a whole can be trusted.​If any part of the “behind-the-wall” plumbing is weak, then the “delivery layer” of the faucet sputters. It’s clear that the investment in data foundations isn’t glamorous, but it’s the prerequisite for enterprise AI to work.​What will separate successful companies in the AI era?​The most common mistake I’ve seen in enterprise AI adoption is simply adding AI to old workflows without redesigning them.​ This is what’s known as AI theater: It looks impressive in demos but doesn’t change much in practice. The gap between promises and real results grows, and employees get tired of trying tools that aren’t much better than what they already use, so they go back to old habits.​The organizations that I believe will succeed are the ones that simplify their SaaS environments rather than continuing to accumulate tools. They’ll also embed enterprise AI solutions directly into workflows, at the decision points and handoffs where intelligence changes outcomes. Crucially, they’ll define success by measuring against business outcomes, like revenue and margins, rather than softer metrics, like usage and adoption rates.​In an AI-first company, more employees will judge tools by how easily their intent turns into action and results that help them do their jobs better. Companies that make this process smooth will be more likely to compete at a level others can’t match.​​Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?