The Next Generation Of HR Software Has Arrived, Finally.

HR software makes up one of the largest technology markets in the world. By our estimate this is a $250 billion market, consisting of payroll systems, core HR platforms, recruiting, training, benefits, and hundreds of other applications. And over the last five years, as companies have adapted to hybrid work and the pandemic, the market has exploded. Just last week Workday, one of the most well known players in the market, announced a 20.1% growth rate in subscription revenue, now reaching $1.3 billion per quarter.

Over the last two decades these systems have radically changed. In the 1980s and 1990s these were mostly back office tools designed for payroll administrators, HR managers, and IT staff.  Starting in the mid 2000s these systems all moved to the cloud, opening up their interfaces to employees. And today, as we all interact with dozens of apps from our phones, HR platforms have crossed the line into “work” tools.

Everything we do at work, from scheduling a meeting to analyzing a spreadsheet, now touches HR systems. Your goals, feedback, team interactions, financial results, hiring, team leadership, and benefits administration are all part of the HR technology stack. And as we proliferate these tools (the average large company has 80 such applications), they are starting to merge together.

Big HCM players like Workday, SuccessFactors, ADP, and Oracle now have partner ecosystems to integrate and simplify all these tools. The adoption of cloud architectures made this easy: once these companies opened up their apps with APIs they each decided to become a “platform” not just an application. Even ServiceNow and Microsoft (Viva) have gotten into the game.

For many years I’ve noticed the aging and somewhat legacy nature of these systems. While modern systems like Workday and SuccessFactors are highly adapted, their architectures are aging. Workday was founded in 2005 and SuccessFactors first shipped in 2001. While these vendors have modernized and updated their data structures and architectures in many ways, their core system is still fairly rigid and brittle. Once you “implement” these systems you’re stuck with many workflow and hierarchy decisions, forcing companies to “reimplement” them periodically as their company grows and changes.

And consider this: according to JPM Chase research, more than 75% of companies go out of business or are acquired within 15 years. So any HCM systems you buy have to be adaptable, flexible, and easy to change. None have been very good at this.

A much trickier problem is collecting and analyzing data. Since most companies have many systems (dozens of recruiting, training, compliance, wellbeing, survey, benefits, and payroll systems), it’s virtually impossible to get a single view of all your employees, all their various HR data elements (training, job history, performance ratings, feedback, pay, benefits) so IT department have to do a lot of work to bring this together. New platforms like Visier have revolutionized this new “systemic analytics” solution, but most companies still struggle.

In the last several months I’ve met with McDonalds, GE, and many other companies struggling with these data issues. McDonald’s wants a single view of all its employees (and contractors) across dozens of payroll systems. GE is doing the opposite: splitting the company into three independent businesses. In both cases they believe they need a whole “reimplementation” of their HCM platforms, which costs tens of millions of dollars and takes several years.

And there’s more: how do we get integrated data (we call it “systemic analytics”) to look at turnover, retention drivers, pay equity, internal mobility, and skills in an integrated way?  These newer HR initiatives require a whole new look at the workforce, integrating even more data across systems. And remember, by the way, that almost a third of the global workforce lives as contractors, so their data is hardly even recognized in these systems at all.

How can a new architecture make this easier?

Well, while the evolution of this market takes time, let me suggest something big. AI, the darling of the media and investing community, may finally bring the “new architecture” we need. And while no vendors have built a full AI-centric HCM system yet, I believe it is likely to come.

Let me explain.

As I discuss in our new whitepaper on AI architectures (coming out at Irresistible), there are three families of AI solutions:  those that have AI “added on,” those who built “AI features within” their platform, and those that are “built on AI.”

The Three Generation of AI Vendors - HR Technology by Josh Bersin

As we detail in our paper, the third category of platforms are built on LLMs and Neural Networks at their core. Rather than storing transaction data in a traditional system and then adding machine learning to improve the experience, they are built on AI first. And after many discussions with technology leaders in these companies, I believe this may be the future.

Consider how an LLM (neural network) really works. These systems are “voracious data analyzers,” looking at tokens (words) or numbers and finding the relationship between them in profound and fascinating ways. While we don’t really know precisely why one person outperforms another person at work, AI will be able to give us clues we simply never saw before. And now that many commercial vendors sell LLMs as products and web services (Google, Microsoft, OpenAI, Nvidia, Anthropic, Amazon, Meta), these platforms are easier and easier to use.

You may say, “it’s not that simple” – existing transactional systems store hundreds of data elements with complex workflow management, security, user interfaces, and integrity checks today. And I surely agree: these HCM and talent management systems are important and essential in every company.

But the problem with these applications is that they’re not flexible. As your company grows and changes the system gets more and more “complex” and “deficient” over time. This is why large companies like Microsoft, Allianz, Nestle, and others have large IT teams focused on process harmonization, data integrity, and architecture to keep things in sync.

In a sense this is why ServiceNow has grown so fast. In an attempt to disrupt this massive market, Bill McDermott and his team position their workflow engine as “the platform of platforms,” able to magically create business rules and apps that “cross” these back end systems, moving the legacy design challenges to a new layer. And as their growth can attest, companies desperately want this new layer of abstraction.

But isn’t that precisely what large scale AI systems really do? Absolutely yes. So we can expect AI-core systems built on Neural Nets and large language models to slowly but surely replace these legacy systems. Core HCM vendors will probably go slowly, but for the most part they see this coming, so they’re making their way as fast as they can.

SuccessFactors, for example, is building a new HXM Graph system based on a graph database, designed to model a highly decentralized, agile company like most of us are quickly becoming. While they have to move slowly because of their many integrations with SAP, they see the future clearly, and they’re experimenting with it now. And they’ve already built Copilots right into the application, adding Generative AI for recruiters, HR staff, and others.

SAP Generative AI strategy

I had a long talk with the head of machine learning at Workday and they see this as well. While Workday believes that their architecture is sound, they see AI models as essential extensions to the Workday architecture. So their engineers look at many possible LLMs and AI models as they built new features, in an attempt to incrementally add value to their massive application.

One of the first use-cases they mention is the ability to “ignore” certain security or workflow rules for employees who are highly trusted. So the Workday UI would simple be different for each user, based on that user’s history, usage pattern, and history in the company.

Workday AI and Machine Learning Strategy

While these are exciting efforts and innovations, I think it will go further.  When I talk with Eightfold, Beamery, Gloat, Phenom, Seekout, or other “AI at their core” vendors they see an even more expansive future. Why wouldn’t the HR system predict and recommend all our learning, developmental activities, job moves, and even day to day activities? If you think about the intelligence in the Microsoft Graph, coupled with the massive data sets managed by Eightfold, Beamery, Gloat, Phenom, Seekout, and others, you can imagine these systems doing much more than Workday does today.

And then we look at specialized vendors like Cornerstone and ServiceNow. Cornerstone’s new AI-fabric is designed to look at all the learning and development activities across 7,000 customers and give your company prescriptive recommendations on content, mobility, skills, and more. While their implementation is still new, the demos I saw most recently are starting to deliver on this vision. Cornerstone customers can see what industry skills, content, careers, and mobility are “recommended” by others in their industry, similar to the way Eightfold does this for recruiting and talent management.

Cornerstone AI architecture

ServiceNow also sees this as a disruptor. After listening to Bill McDermott and his team describe their future, they want to be “the AI platform for the enterprise,” and their recent announcement of Employee Development And Growth (through acquisition of Hitch Works) clearly shows where they’re going. They want to be your “enterprise-wide intelligence platform,” eventually replacing HCM systems in a whole new platform architecture.

ServiceNow Generative AI Strategy

As with all these major architectural shifts (the move to cloud, the move to mobile), the move to AI will seem confusing at first. The large vendors will move slowly and incrementally while new startups will move at lightning speed. But some will likely disrupt. I was particularly interested in the LLM-based business system launched by Palantir. It’s essentially an ERP (financial system) designed around a specially built large language model.

I recently discussed this trend with Ashutosh Garg, the Co-Founder of Eightfold. He, like me, also believes this AI-centric architecture will win out over time. Their system can integrate and aggregate almost any data because its architecture is so open. (Traditional ERP systems are not designed for this.) He believes “transactional integrity” will be an add-on to the AI core, essentially the opposite approach to incumbent vendors.


Where This Architecture Will Go

While I never try to predict the innovations of smart pioneers, I see a massive new future ahead. When we first described “Talent Intelligence” five years ago, we were essentially pointing out the need (and opportunity) to use AI to understand jobs, roles, skills, and employee fit across a much larger data set than the existing employees in your company. This same idea will take over all the HCM applications over time, simply because LLMs and GPUs now make it possible.

As I see it, most of the essentials of business management come down to human judgement. Who to hire, who to promote, and who to move into a new position are each difficult decisions we make based on personal judgement. When you add other decisions like how big to make the team, how to improve productivity, and how to build new skills faster – even more “guessing” is required. And then there are issues like “how much to pay someone” or “who should be the general manager” which are both tricky and consequential.

Will this be a vendor-provided LLM or one of your own? Both options will likely be available. Small companies will use off-the-shelf LLMs from vendors; large companies will run their own, using internal data in a protected, fully secure environment.

If we use an LLM to help with these talent and HR decisions our companies will just run better. Yes, the AI will never fully replace human judgement. But imagine if you could see the statistical proof for a given decision and THEN apply judgement?  I know this will dramatically improve our decision-making.

These Generative AI architectures really are the next big thing. Jensen Huang, Co-Founder and President of Nvidia, calls LLMs “the biggest computer industry transition of our generation.” And I have to agree.

Our new AI whitepaper describes many examples of these “AI-informed” decisions and you’ll be surprised how powerful they can be. While I cannot predict how fast this transformation will occur, I believe we now have the “next-gen” HCM architecture we have been waiting for.

Come to the Irresistible conference to hear more.

Additional Information

The Rise Of The Copilots: Microsoft Steps Up The Pace

How AI Is Disrupting The HR Tech Marketplace

What Is A Neural Network? Fantastic Overview Of How AI Systems Work.

Redesigning HR: An Operating System, Not An Operating Model.

Why Is The World Afraid Of AI? The Fears Are Unfounded, And Here’s Why.