Interview with Jeb Horton from Hitachi Vantara

1. Barely 7% of organizations feel data-ready to fully onboard new AI infrastructure. This has created a drift between AI aspiration and operational readiness, while also creating an awareness gap. Leaning into your experience, what is the unrealized cost of succumbing to this ‘fear’ of change?

The cost is real, and it compounds. But I don’t think the issue is simply fear of change. In many cases, it is uncertainty about where to start and how to modernize without disrupting the business.

We have seen a similar pattern before with the BI and data warehousing boom of the 1990s and early 2000s. Businesses often struggled to get the results they expected because they didn’t trust the data. The pattern is familiar. Organizations rush to adopt the latest technology without fixing their data foundation. The result is poor inputs, poor outputs, and AI is no different.

If your data isn’t clean, properly governed and accessible, you’ll get hallucinations, bad decisions, and wasted investment. The cost of that delay is falling behind while your competitors move faster.

2. 80% and higher, by some estimates, is how many AI projects fail to take off in their pilot phase. And 95% of the time, it all boils down to a poor foundation, especially messy and incomplete data readiness. Do you think this is due to most organizations being too focused on AI adoption before addressing the infrastructure and data foundations need to support it?

Absolutely, and I’d take it even further. It’s not just a problem with AI – it’s a theme throughout enterprise technology. There was a similar issue with data analytics 10 years ago. The entire world, it seemed, was focused on the algorithm, thinking that was the key.

Now we’ve moved to generative AI with higher compute and cost, and yet the underlying problems are the same. If the data isn’t right, you get bad outcomes. Data preparation, cleansing and proper classification through coding and tagging are key.

So are governance, infrastructure resiliency and making sure the right data is accessible at the right time. There’s a lot of noise coming in, and filtering that out is important. Once you have clean, curated and trusted data, your analysis is straightforward.

3. 57% of the organizations store data across at least two different platforms, creating a multi-cloud sprawl. How does a more unified approach like Hitachi EverFlex break through the ‘hybrid paradox’ of fragmented platforms, when only 34% of businesses usually know where all their data is located?

The fragmentation problem is exactly what a platform approach is designed to address. It’s about consolidating multiple siloed data services into one construct so you’re not managing discrete storage solutions independently. You’re aggregating them, bringing them together, and using a metadata-based management layer to operate on all that data at a higher level. That matters a lot when you don’t know where your data is, because the platform gives you visibility to see what you have and act on it.

On the Hitachi EverFlex side specifically, our flexible consumption portfolio gives customers more choice in how they buy, pay for and operate enterprise data infrastructure. It allows our customers to consume our storage in a cloud-like model, whether they’re on-prem, in the cloud, or in a hybrid environment.

We support all three major cloud providers: Azure, AWS, and Google. The reason we built it that way is that no customer wants lock-in. The goal is to give them the flexibility and choice to decide what goes where based on workload, cost, governance and business requirements.

4. Fragmented data platforms create compliance gaps and delays in AI projects, but that’s only scratching the surface. For customers concerned about egress costs and multi-cloud sprawl, is frictionless data mobility without a proprietary storage tax the only way out?

Egress and ingress costs are a real and costly consideration. The economics of hyperscalers are focused on moving things to the cloud and keeping them there. That can be expensive, especially for AI workloads with large data sets. The answer isn’t necessarily one model over the other, but about giving customers the choice of what goes where.

Some things make sense in the cloud, and some are more cost-effective on-prem.

What we’re aiming to do is provide a cloud-like experience for on-prem infrastructure, where you pay for what you need and use. That gives customers more flexibility while helping them maintain greater control over cost, performance, compliance and data location. And because compliance, data sovereignty, and access controls all live within the same platform, you aren’t moving data between silos just to apply those controls.

5. Disconnected SaaS/ERP/CRMs decrease time-to-value. It’s a known dilemma but rarely acted upon. How do Global Services heads cope when compliance risks balloon mid-pipeline, turning initial data woes into enterprise-wide liabilities?

What we’re seeing is that disconnected SaaS, ERP, and CRM environments don’t just slow down digital transformation anymore, they create operational and compliance liabilities that compound over time. Every new AI workflow, customer interaction, or business process is generating more data across more environments. The challenge is that most enterprises still operate with fragmented visibility, fragmented governance, and fragmented accountability.

That’s where Global Services increasingly steps in.

The conversation is no longer just about integrating systems. It’s about helping customers regain operational control over sprawling data estates before compliance exposure, cyber risk, or AI governance gaps become business problems. Enterprises are realizing that bad data architecture creates downstream liability. If your data lifecycle policies, retention rules, security posture, or sovereignty requirements aren’t consistently enforced across environments, AI simply amplifies the problem.

The traditional approach of buying disconnected tools and trying to stitch them together later is becoming unsustainable. Customers want outcome-based operating models where infrastructure, governance, resiliency, and lifecycle management are designed together from the beginning. That’s one reason we’re seeing growing interest in flexible consumption models like Hitachi EverFlex.

Organizations want the ability to modernize infrastructure and data operations without locking themselves into rigid procurement cycles or overbuilding capacity ahead of demand. Flexibility matters because the compliance environment, AI workloads, and business priorities are all changing simultaneously.

For services leaders, the priority is to move that conversation earlier. You have to help customers understand where risk is being created, which systems are introducing the most complexity, and what operating model is needed before those issues become harder and more expensive to unwind.

6. With internal breach concerns rising to 41% from 31% and data security topping 56% of all AI worries, leaders are starting to believe AI aids attackers even more. Can you walk our readers through the urgency of balancing shadow AI governance risks with real-time compliance, especially across the cloud infra sprawl already ripe for attack?

Cyber resiliency is one of the most important conversations happening at the moment.

AI can now flag threats that used to require hours of manual analysis. The ability to use AI to flag potential threats faster is extremely valuable. But that same speed creates risk if organizations do not have the right governance and compliance controls in place, especially as shadow AI, or the use of AI tools outside approved enterprise policies, expands across the business.

The balance is not between innovation and control; it is about making sure the controls move at the same speed as the innovation.

However, when data is spread across multiple environments and platforms, the attack surface expands. You need to know who has access to what, what data needs protection and at what level, and whether those controls are being applied consistently. That’s not possible when you’re managing discrete silos. A consolidated platform lets you set access attributes, apply compliance controls, and get the visibility you need to be proactive rather than always chasing the last problem.

If teams are using AI tools without clear governance, they may expose sensitive data, create compliance issues or rely on outputs that have not been properly validated. That is why security, compliance and AI governance need to be built into the operating model from the start. The goal is not to slow AI adoption down, but to make sure it scales safely, with the same level of control, visibility and accountability organizations expect from the rest of their enterprise infrastructure.

7. Despite heavy investment in AI data prep, only 1 in 5 businesses has these models moved into production. For those waiting to realize the promised ROI, how can leaders scale AI initiatives through flexible consumption and managed services models like Hitachi EverFlex without eroding margins or risking the five nines of reliability they’ve spent decades building?

ROI in a consumption model works differently than a lot of people might expect. It’s not about getting return on a big investment upfront but about lowering the upfront cost and better aligning spend with value realization over time.

Instead of overcommitting before the business case is proven, customers can start smaller, scale as value is demonstrated and manage margins more effectively as costs and benefits come into better alignment. When you lower the ongoing cost of storage, customers have more flexibility to invest in other priorities, and that’s when the actual value shows up.

Regarding reliability and pricing, we charge per gigabyte, so customers have a predictable structure and can scale usage up or down depending on their needs. We also support heterogeneous environments and run alongside competitors’ equipment, which many vendors won’t do. That flexibility matters to customers who want to avoid vendor lock-in.

Managed services also play an important role here.

Many organizations want to scale AI, but they do not want to add more operational burden to teams that are already stretched. A flexible consumption model can help align cost to actual usage, while managed services can help maintain the reliability, resiliency and operational discipline customers need for mission-critical environments.