Will AI Make IBM the World’s Most Productive Company?

IBM didn’t have to look far to find one of its most ambitious AI clients. The company, with more than 275,000 employees worldwide, is deploying AI to transform itself, acting as its own “Client Zero.”

CEO Arvind Krishna kicked off the transformation in early 2023 with the ambitious goal of becoming the world’s most productive company.

By automating repetitive tasks, AI has saved IBM employees more than 3.9 million hours of time in 2024 alone. Today, IBM is sharing with its enterprise clients not only the technologies it has refined internally, but also the business lessons it has learned from the transformation.

“We’re a hardware, software and services business, and we operate in a lot of different countries,” says Joanne Wright, Senior Vice President of Transformation and Operations at IBM. “We feel like if we can do this within IBM, you can really see a vision of doing it across every industry, across every client.”

The company’s strong results come at a time of increasing C-suite skepticism about AI. A global survey of CEOs by the IBM Institute for Business Value found that while most expected their AI investments to at least double in the next two years, only 25% of their AI initiatives have so far delivered their expected ROI. Only 16% of those initiatives have scaled enterprise-wide. Why aren’t AI initiatives living up to their potential?

Consider the guiding principles that Krishna has laid down for IBM’s transformation: “Eliminate, simplify, automate.” That means eliminating operational complexity and simplifying end-to-end workflows before automating tasks and embedding AI across operations. “Before you apply any new technology, you have to decide what to stop doing and then redesign the workflows,” Wright says. “Otherwise, you risk automating bad processes.”

IBM didn’t settle for scattered pilot projects. With a steering committee providing leadership, company teams used external benchmarks to target the workflows most likely to deliver a high ROI when redesigned or automated.

IBM had 100 professionals dedicated to supporting Americas IT issues for employees around the clock at 1-888-IBMhelp, costing about $40 million a year.

As an experiment to see how fast an enterprise-scale AI application could be built, the AskIT assistant was created in a 100-day sprint. AskIT was trained on 80% of the IT issues the company faces most frequently and can handle inquiries in multiple languages, with humans handling the rest via multi-lingual chat.

AskIT has saved $18 million in support costs, while preserving 24/7 human chat for complex queries.

IBM has staged companywide challenges in which every employee has had the opportunity to suggest ways to simplify their work. About 178,000 employees registered for the 2024 IBMer watsonx Challenge, and 70% of teams who submitted projects said they had created improvements that they plan to apply to their daily work.

But these challenges do more than merely tap into innovation. They also embed AI transformation principles and tools into IBM’s culture, while helping to allay employees’ inevitable fears that AI is coming for their jobs.

“The challenges have helped to demystify [the technology] and helped people focus on what it could do for them personally,” Wright says.

More than 50 AI projects across IBM have transformed many roles: AI is enabling customer support teams to resolve the majority of inquiries, assisting salespeople in drafting prospecting communications and helping consultants develop evidence-based strategies.

In a typical project, teams start by consolidating data from multiple systems into a single source of trusted data on the company’s enterprise management platform, housed in the hybrid cloud. These teams streamline tasks and then build domain-specific digital assistants.

Many of these assistants are based on relatively small language models and trained on internal datasets, making them more customized than the large language models (LLMs) that run consumer AI services. If unsure of an answer on a key topic, they can be instructed to refer to human experts, minimizing “hallucinations” by AI systems. These assistants can be used to help reduce the tedium of finding and organizing information, while making that information more accessible to authorized users.

AskHR can help employees and managers carry out more than 100 types of transactions, from employee verification letters and address changes to salary increases and employee transfers, says IBM Senior Vice President and Chief Human Resources Officer Nickle LaMoreaux.

IBM employees transfer to a new manager an average of once every three years—that’s more than 90,000 transfers a year. Using the previous online process, managers would make errors on one of every seven transfers. And 14.1% of the time, they needed someone from HR to complete the process.

Now, IBM’s AskHR agent guides managers through the process, asks questions and checks their work. Last year, approximately 84% of employee transfers were routed through the agent. The current error rate? Zero.

The change gives HR managers more time for higher-value work such as manager coaching.

Many of the domain-specific AI assistants now available throughout IBM departments will soon be subsumed into a single agent, AskIBM. The next frontier at IBM is agentic AI, as digital assistants and agents evolve from answering queries to executing complex tasks with a degree of autonomy.

Agentic AI is already implemented in less sensitive applications, and “you’re going to progress to more sensitive functions as trust increases,” says Matt Lyteson, Chief Information Officer, Vice President, Technology Platforms Transformation at IBM. IT teams are building out identity and access management systems so that AI agents can be given discrete identities and managed as they interact independently.

Watsonx Orchestrate contributed to Dun & Bradstreet enhancing its solutions so that its customers could see an estimated 10-20% reduction in time for procurement tasks, leading to significant cost savings. Avid Solutions now takes 25% less time to onboard clients, thanks to watsonx Orchestrate. Brazilian financial giant Sicredi increased its customer retention by at least 10% using watsonx.ai and other IBM solutions in a pilot.

Meanwhile, the technology is helping even IBM’s less technical employees transform their respective domains. Chief Procurement Officer Lorie Meola has established an informal “procurement lab” that is piloting new solutions for her function.

“All this stuff I’m building, I’m going to need people who understand it and how to maintain it. … I’ve said in my team meetings, ‘If you can eliminate your job with these technologies, I will find another job for you to go work your way out of,’” Meola says.

Getting a holistic view of a supplier used to require spending hours talking with various colleagues, and hours more conducting research.

Now, an assistant called AskProcurement connects with digital assistants from the procurement, risk, finance and sales functions, and compiles data from internal applications and external sources to produce a presentation-ready brief.

The brief is ready in seconds, saving an estimated 26,000 hours annually. Procurement staff now have more time to work on business challenges and build relationships with suppliers.

Unlike other AI companies, IBM has more than a century of experience in applying innovative technologies to solve business problems—from mainframe computers to PCs, from the internet to the cloud. Its Client Zero transformation—and the continuous improvement feedback loop that drives it—is just the latest chapter of the company adapting its own operations.

“AI has made us much more effective and much faster,” says Wright. “Our employees and our clients are seeing a new IBM.”

Read about our agents