At an industry conference, Satyen Sangani, CEO and Co-Founder of Alation, sat down with the CIO of a leading manufacturer for one of the most candid conversations of the event. With well over a million automobiles on the road, a commitment to restore them to service within 24 hours (as the CIO described it), and operations spanning more than a hundred locations across multiple countries, this is not a company that can afford AI experiments that don't deliver.
The result was a frank, ground-level look at what it genuinely takes to move enterprise AI from pilot to production, and what gets in the way.¹
Cutting through the AI hype: an enterprise perspective
The CIO opened with a perspective that might surprise an audience of data, business and AI leaders: he still calls AI a hype. He isn't dismissing it. He's running multiple AI initiatives at a global manufacturer, but he has learned to separate noise from value.
"I still consider it a hype. I think there is more on paper than it is in real world."
That skepticism has a number. He estimates that 60–70% of the AI use cases he has reviewed in recent months don't pass his value test. "A lot of the use cases which I've seen over the last couple of months, I would say 60–70% of them I would not start now because they are simply to realize them might be not so difficult, but to run them is just too expensive for the value they deliver."
The CIO's caution echoes the analyst view: Gartner predicted in June 2025 that more than 40% of agentic AI projects will be canceled by the end of 2027, citing escalating costs and unclear business value (as reported by Reuters).²
The takeaway for data and business leaders: be honest about total cost of ownership, not just implementation cost. Token consumption and inference costs add up fast at enterprise scale, and the ROI math must work in production, not just on a whiteboard.
The use case that changed everything: AI agents for aftermarket parts
The first production AI deployment centered on a very concrete operational problem. A small team of material planners was responsible for managing more than a million aftermarket parts, coordinating with hundreds of suppliers, and serving hundreds of dealer locations, all while honoring a same-day parts delivery promise to customers.
"I don't have [hundreds of] material planners, so... at the moment it's [only a handful]."
The volume-to-headcount ratio made the problem obvious. Planners were spending the bulk of their days on what the CIO calls "operational archaeology": manually tracking whether a part was ordered, shipped, received, and available at the right distribution center. An AI agent was built to handle exactly this workflow, monitoring the entire parts supply chain, flagging shortages before they happened, and surfacing a 7-day forward-looking prediction of parts availability.
What makes this use case notable is how it differs from classical machine learning. Rather than simply predicting parts decay curves, the agent does deep research across multiple data sources (ERP master data, legacy warehouse systems, dealer inventory management systems), synthesizing them the way a skilled planner would. This is AI as a force multiplier.
Why data accuracy is the make-or-break factor for enterprise AI
Getting the agent to work required solving a problem that trips up most enterprise AI programs: data accuracy across fragmented, legacy source systems. The company's parts data lives in its ERP, in disparate warehouse management systems across each parts delivery center, and in dealer management platforms, none of which were built to talk to each other.
Alation's AIOS played a central role here. By establishing consistent metadata structure and data context across these sources, the team was able to give the agent the trusted, well-defined inputs it needed to produce reliable results. Importantly, it wasn't just a technical win. It was a trust win.
"The best moment was, of course, when they were coming and said, 'Yes, this is correct data. This is exactly what I would have come out with as well.'"
That moment of validation from the material planners themselves, people who were initially skeptical of handing anything to a system, was the turning point. It highlights a critical lesson: the social chain matters as much as the technical chain. Getting end users to validate accuracy early is how you build the trust that sustains adoption.
Maintaining that accuracy is ongoing work. "You still need to work on the accuracy. I think that it's always depending on the data you're feeding in," the CIO noted. "The data work will never end."
AI as task replacement, not job replacement
One of the most important reframes in the conversation was the CIO's distinction between AI replacing jobs and AI replacing tasks. The material planners are not gone. They are doing more valuable work:
"It's not taking the job, it's basically taking tasks out of their daily operations... they are doing now more value-adding topics for the company. They can talk with the suppliers, they can look into are we using—can we actually bundle suppliers?"
This "abundance" framing, a term Sangani used during the conversation, is a more productive way to position AI internally. Rather than asking "what work will the machine take?" ask "what higher-value work does this unlock?" For business leaders trying to drive adoption across cautious teams, that reframe is essential.
Governance, education, and change management: the hard work of AI transformation
The CIO was candid about his own evolution on governance, a topic he once resisted. "I hate, and I always hated the term governance. I never wanted to be governance because I said this is you know police and whatever. I don't want that. I need to—we need to do that, we need to—but this topic requires a strong governance and requires also a strong education."
He's not alone in his distaste for data governance, but the stakes are higher than ever. Data readiness is where many enterprise AI efforts stall: Gartner predicted in February 2025 that organizations would abandon 60% of AI projects not supported by AI-ready data through 2026.³ In the same release, Gartner reported that 63% of organizations either lack the right data management practices for AI or are unsure they have them, based on a July 2024 survey of 1,203 data management leaders.³
Alation's view is that the rest follows from there: when data is poorly governed, context goes missing and the humans in the loop stop trusting the output. His investment in data infrastructure, made years before the current AI wave, paid dividends precisely because it built the data mindset across the organization that AI now requires.
He was equally frank about the human side of transformation. The CIO said that in manufacturing, finance, and HR, many workers see AI as a threat.¹ "You need to slowly bring them to a point that they work with it, and that they are experiencing it." Leadership pressure and frontline anxiety often pull in opposite directions, and navigating that gap is one of the defining challenges of the CIO role today.
Building the agentic enterprise: what comes next
Looking ahead, the CIO's vision goes well beyond individual use cases. He is working toward a "plug-and-play" agentic orchestration platform, a framework where best-of-breed agents can be deployed and governed centrally, with clear guardrails on which data each agent can access.
"The dream is to have this agentic orchestration platform and all the agents can run there... I have a data governance framework to make sure that I'm not having agents in place which are using data which I do not allow that they use this data."
This vision of humans and "digital workers" operating as teammates on the same value chain requires precisely the kind of enterprise-wide data intelligence infrastructure that Alation was built to provide. As he summarized: "We will have humans in a certain sense and we will have digital workers, and we have them in one team working together on topics and delivering value for the company."
Key takeaways for data and business leaders
This story offers a clear blueprint for any organization serious about moving AI into production:
Start with a concrete, high-value problem, not a technology solution looking for a use case.
Invest in data quality and metadata governance before deploying agents. Accuracy is a trust problem as much as a technical one.
Bring end users into the validation loop early. Their buy-in determines whether the agent actually gets used.
Measure ROI rigorously. If a use case is expensive to run relative to the value it delivers, it doesn't pass the enterprise test.
Frame AI as task replacement, not job replacement, to build the organizational will to change.
As Sangani put it near the close of the session, getting to production requires "clear problem identification, a large multiplicity of data sources, a social chain, feedback loops, and constant evolutionary change." There are no shortcuts, but for organizations that have done the foundational data work, the opportunity is very real.
Alation is the data intelligence platform powering trusted enterprise AI. Start a conversation with our team.
Sources & notes
Every external claim on this page is independently verifiable. The public sources are listed here.
This story is drawn from a conversation at a public industry event. At the customer's request, the company and executive are not named. Customer-specific figures and quotations are as described by the speaker during the session.
Gartner predicted that more than 40% of agentic AI projects will be canceled by the end of 2027 due to escalating costs and unclear business value. — Reuters, via The Star, 25 June 2025 ↗ https://www.thestar.com.my/tech/tech-news/2025/06/25/over-40-of-agentic-ai-projects-will-be-scrapped-by-2027-gartner-says
Gartner predicted that through 2026, organizations will abandon 60% of AI projects unsupported by AI-ready data. A July 2024 survey of 1,203 data management leaders found that 63% of organizations either do not have, or are unsure they have, the right data management practices for AI. — Gartner, 26 February 2025 ↗ https://www.gartner.com/en/newsroom/press-releases/2025-02-26-lack-of-ai-ready-data-puts-ai-projects-at-risk
ANALYST ATTRIBUTIONS & DISCLAIMERS
Gartner, Lack of AI-Ready Data Puts AI Projects at Risk, Q&A with Roxane Edjlali, 26 February 2025.
Gartner, Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027, 25 June 2025.
Gartner does not endorse any vendor, product or service depicted in its research publications and does not advise technology users to select only those vendors with the highest ratings or other designation. Gartner research publications consist of the opinions of Gartner's research organization and should not be construed as statements of fact. Gartner disclaims all warranties, expressed or implied, with respect to this research, including any warranties of merchantability or fitness for a particular purpose. GARTNER and Magic Quadrant are registered trademarks and service marks of Gartner, Inc. and/or its affiliates in the U.S. and internationally and are used herein with permission. All rights reserved.
- AI
- Customer Stories
- Data Governance
- Digital Transformation
Keep reading
More from the data desk
