Trending Articles
Trending Topics
Trending
Jon Lester, Vice President IBM HR Technology, Data & AI, shares IBM’s HR digital transformation journey that started in 2017
The past few years have signified a move from AI as an exciting technology concept to a tangible tool for business process – notably in the field of HR where AI’s ability to manage data quickly and accurately has transformed how many HR business leaders work today. No more so has this evolution been apparent than in our recent news announced at THINK, where we launched our most comprehensive set of agent capabilities to date, helping companies build, run and manage agents – delivered via IBM watsonx Orchestrate.
As experts in Generative AI (GenAI), we know we need to practice what we preach, which is why, for almost a decade, we have been evolving our own HR capabilities with AI, culminating in a new wave of GenAI-driven innovations. This year, we have launched AskHR Agents: GenAI tools that deliver hyper-personalised user experiences that help HR professionals work more productively and effectively.
Getting to this stage hasn’t been easy. It has required agility and a willingness to fail and learn fast. Here, I’ll take you through the events that led to this important milestone in our HR journey, and share some lessons for businesses targeting a similar transformation. After all, many companies are excited by the potential of GenAI across multiple business domains, but few have been able to leverage the technology to unlock real value for employees, customers or business leaders.
IBM’s HR digital transformation journey started in 2017, back when tools like ChatGPT were still in the realm of science fiction.
Our HR processes, policies and technologies varied significantly from region to region. Naturally, this meant that running HR was fairly complex, time-consuming and prone to discrepancies. Many HR professionals were performing low-value repetitive tasks regularly, which was neither fulfilling nor rewarding.
Cloud solutions were one key driver of change, where advances helped us eliminate silos and start building a truly global HR function with policies, processes, data, and technologies aligned across the globe.
Fast forward roughly three years, and we witnessed the emergence of data fabric: data architectures that enable end-to-end integration of multiple data sources. Giving rise to Platform as a service (PaaS) and extensive API integrations, data fabric enabled us to build traditional AI solutions with machine learning and Python that could surface insights from rich data sets directly to HR leaders.
At the same time, a new engagement layer for employees emerged with the creation of HR Digital Assistants using Natural Language Processing, transforming HR support as well as the experience of getting help through a chat interface. Alongside responding to over 2,500 questions and pulling information from 8,000 policy pages, our new digital assistant, “AskHR,” enabled employees to “transact” via chat into multiple HR platforms in a quarter of the time.
Over the next few years, we kept pushing forward and started experimenting with digital twins: an integration of multiple processes and workflows across programmes, that is a digital assistant that had both context and a memory. Digital twins enabled us to further automate low-value, repetitive HR tasks, for example data processing, and create solutions that were domain experts in their HR field, from recruitment and retention to training and promotions.
Then in 2023, GenAI stepped out of the realm of fiction and into the world of business. At the time, I remember being repeatedly asked if we would have done anything differently if we had GenAI capabilities back in 2017. My answer was always “no,” but we would have done it much quicker! All the work we did simplifying processes, unifying data siloes, and automating information retrieval and updates was necessary for unleashing the full power of GenAI in HR.
Every three years in HR technology, we set a new vision for ourselves – our goal for 2022-24 was to help humans focus on high-value work by moving repetitive, low-value tasks to AI assistants. We can confidently say that we have achieved this goal with, for example, our HR support SMEs moving up two band “levels” due to their upskilling and moving to higher-value work. Our next three-year vision is to create a hyper-personalised experience for every IBMer using GenAI and agentic frameworks, which will further increase employee productivity.
Imagine asking an AI Agent to provide operational insights for you with voice prompts, for instance, or to automatically identify areas for learning and career enhancement based on your current performance, skills and aptitude. You could even have an AI Agent reach out to you to recommend a new set of benefits based on an upcoming life event. That’s the next stop for HR at IBM. There’s still more work to do, but our destination is just around the corner, and we believe these five lessons will keep us on track:
At each stage of our AI and GenAI journeys, we took time to think about the real-world possibilities that new technologies provided. While it may sound simple, this crucial step is key to ensuring that any further development of GenAI solutions are aligned with actual user needs and the strategy of the business. In an age where companies are rushing to get to market with GenAI solutions, taking time to think through your business case carefully will save a lot of pain in later stages.
From the very start of our AI journey, and before GenAI emerged as a viable technology, we had been working on reducing our technical debt. This means that we replaced disparate legacy systems that supported HR functions at a regional level only, with solutions that drive globally standardised processes. Eliminating technical debt reduces development complexity and helps to ensure that GenAI models are trained with clean, quality data. Based on our starting point of 2016 we had reduced HR technical debt by 85% by the end of 2024.
Large Language Models (LLMs) are very powerful, but in our experience, the wider the scope of an LLM, the more prone it is to respond with hallucinations. Plus, training a single LLM can take a tremendous amount of time, resources and energy. To mitigate these risks and challenges, we developed domain-specific models that will eventually morph into small language models (SLMs) which support career guidance, rewards, recruitment and more. By breaking our development efforts into smaller chunks and using relevant data specific to distinct use cases, we’ve achieved high accuracy levels and reduced the ‘cleaning’ and automated ingestion of content into a LLM plus basic levels of tuning into a two-week sprint.
Leaders often feel tempted to try and build the perfect solution before deploying it into production, but this will only delay return on investment. We learned early on that you get the best results when you start with minimum viable solutions and tweak them along the way.
In 2023, for example, we received 42,000 pieces of feedback from the user base of our digital assistant, AskHR. These provided us with valuable insights into how we can make AskHR even more relevant to IBMers so they can supercharge their productivity. What’s more, responding to user feedback in this way helps every employee become an integral part of our HR transformation initiative, which ultimately helps to boost user adoption of the tools we develop.
In the same way that all previous deployments of AI followed IBM’s AI Ethical principles, i.e. the purpose of AI is to augment – not replace – human intelligence; data and insights belong to their creator; and new technology, including AI systems, must be transparent and explainable. GenAI solutions must do the same. There is sometimes a reluctance for HR leaders to invest in GenAI solutions due to the risk of hallucination around highly compliant content. The adherence to AI Ethics at every step can mitigate much of this reluctance.
We’ve learnt a lot throughout this process, and even in the early stages, have seen strong results. For instance, we processed over a million HR-related transactions in 2024 alone; each transaction improved productivity in domain-specific tasks by up to 75%.
Productivity gains like this are driving tremendous improvements in efficiency, work-life balance and employee satisfaction. We’re excited for what the future holds for our GenAI transformation, and we hope that by sharing our journey we can inspire many more companies to leverage AI in HR and beyond.
