IBM made itself AI's Client Zero: 3,000 digital workers, $3.5B productivity gains

IBM spent the late 2010s and early 2020s stagnating while Deep Blue and Watson kept its AI reputation alive. CEO Arvind Krishna judged generative AI's rise an existential moment demanding more than incremental change. The answer was Client Zero: become IBM's own most demanding customer, transforming operations with AI before selling to others. 'If we can't be the best user of this technology, we shouldn't be in the AI business,' Mohamad Ali, senior vice president of IBM Consulting, told Newsweek. Ali, who left in 2009, returned in 2023 to help lead the transformation.

IBM decomposed its operations into 490 workflows, picked 70 as low-hanging fruit, and created more than 3,000 digital workers across HR, finance and supply chain. Every employee was trained in AI tools and encouraged to build their own assistants; 150,000 joined AI hackathons and 119,000 completed agentic AI education. The push ran from the top: the CEO and executive team met weekly, backed by a transformation steering committee, a project management office and a productivity discovery team, with the company controller verifying the numbers each quarter.

The reported results: HR cut time-to-fill 47% and its operating budget 40%; supply chain automated 90% of purchase order processing, saving $315 million over three years; AI support addressed 86% of top IT issues. IBM says it has unlocked $3.5 billion in productivity gains and counting, revenue growth swung from negative 3% to positive 5% over three years, R&D rose from 9% to 12% of revenue, and the stock rose 36% over the past year, more than double the S&P 500. 'We didn't just go buy a whole lot of Copilot or Gemini licenses,' Ali said.

$3.5 billion in productivity gains (IBM-reported); revenue growth swung from -3% to +5% over three years; stock up 36% over the past year

Krishna judged generative AI existential: incremental change would leave IBM selling a technology it had never proven on itself.

Decomposing work into 490 named workflows made benefits measurable — 70 quick wins first, then scale to 3,000 digital workers.

Training every employee and letting them build their own assistants turned adoption into a grassroots competition rather than a mandate.

Weekly executive reviews, a steering committee and quarterly controller-verified numbers kept claimed productivity gains honest.

Before selling AI, become its most demanding customer: break work into named workflows, redesign them for human-plus-digital teams, and let employees build their own assistants.

By late 2025 IBM reported $3.5 billion in productivity gains and counting: hiring time down 47%, 90% of purchase order processing automated with $315 million saved over three years, and 86% of top IT issues resolved by AI support. Ali's advice to clients mirrors the internal playbook — stop celebrating isolated experiments ('I have 1,900 AI pilots and not a dime of benefit to my bottom line,' one telco CFO told him) and commit top-down to redesigning how work gets done.