Hitachi's CIO says the Japanese conglomerate's enterprise AI strategy isn't one-size-fits-all

Bala Krishnapillai has a simple line he keeps repeating: there is no one AI solution for a company like Hitachi. That might sound like corporate hedging, but for an outfit with nearly 290,000 employees spread across 607 subsidiaries in 190 countries, it's closer to necessity than caution.

As Hitachi's Americas CIO, Krishnapillai has split AI adoption into three lanes. Everyday stuff — Microsoft Copilot, Google Gemini — handles meeting notes and translation, the latter a genuine operational need given how many languages Hitachi does business in. Job-specific tools come next, picked by IT working alongside business leaders, whether that's content generation for creative teams or competitive analysis for sales. Coding assistants sit in a third lane, built around a close relationship with Anthropic. Nothing here is exotic. What's notable is the restraint: three and a half years after ChatGPT, Hitachi still hasn't rolled out one blanket tool for the whole company.

That restraint has a price tag attached to it. Krishnapillai openly says token consumption is now a real concern, and controls are going in. R&D gets a pass because staying competitive there matters enough to justify open usage, but elsewhere division managers own their AI spending and are expected to track it. It's a quiet acknowledgment of something an EY survey backs up hard: 98% of companies using token-based AI tools say the costs have forced them to rethink their approach, though only 64% actually have budget guardrails in place yet.

The messier problem, though, was data — not tokens. Over a century of acquisitions, including an $11 billion ABB power grids purchase and a $9.6 billion GlobalLogic deal, left Hitachi with customer data scattered across more than 150 different CRM systems. Building a sales proposal used to take weeks just to pull together an accurate picture. Rather than force everything into one database, Krishnapillai brought in Appian to layer connectivity on top of the existing mess. Hitachi and Appian claim that's produced a 40% efficiency gain in sales and marketing and cut operating costs by 20%.

What comes next is the part everyone's watching nervously. Appian CEO Matt Calkins frames agentic AI as the natural payoff of that data work — autonomous agents wandering through Hitachi's systems, asking questions nobody scripted for them. Krishnapillai isn't rushing there. Given a report this week that models from Anthropic and OpenAI have already taken unsanctioned actions on their own, he's putting security and governance first, and is shopping for software that can track every AI agent Hitachi spins up before letting any of them roam free. His own words: early phase, not there yet.