Evolution of HR software: from filing cabinet to agentic HR (2026) | Peepel
Michiel Bearelle · 5 min · 2026-06-29
HR software evolved from paper files to the HRIS, to workflow suites, to agentic HR. One thread: from organising work to executing it. With numbers from Gartner, Deloitte, SHRM, McKinsey and Bersin.
Key takeaways
- HR software has had four eras: paper, HRIS, workflow suites, agentic HR.
- Every generation organised work better; only the fourth executes it.
- An AI feature bolted onto an HRIS or workflow suite doesn't change your era - only who executes the task does.
- RPA tried execution through brittle scripts and broke: EY estimated 30 to 50% of RPA projects fail.
- Bersin, Deloitte and Gartner confirm agentic HR is tipping now; Gartner: 40% of enterprise apps with task-specific AI agents by 2026.
In one sentence: HR software evolved through four eras: from paper, to the HRIS, to workflow suites, to agentic HR. Every leap solved a piece of manual work, but only the last one actually takes the work away. The thread: from organising work to executing it.
Look at the history of HR software and one pattern keeps repeating: every generation made organising work better, better storage, better workflows, better reporting, but the work itself stayed with the human. Only now, with agentic AI, does that shift. This piece lines up the four eras, shows where we stand today, and explains why the next step is fundamentally different from the previous three.
The pattern: HR software got stuck on "organising"
Every era of HR software promised less work. And every time, the work did become neater, but rarely less. The reason is structural, and it's the core of this whole story: until recently, software could store data and route tasks, but not actually reason, decide and act. It could put a form in front of you, but not fill it in. It could route an approval, but not judge it. The system organised the work; the human performed it.
What's changing now is exactly that boundary. Modern AI agents can be given a goal, plan, call tools (generate a contract, push data to payroll, check an absence) and report back. For the first time, software can perform the task itself instead of just teeing it up. That isn't a new feature; it's a new kind of software.
This "system of X" framing isn't a marketing invention either. It goes back to Geoffrey Moore, the author of Crossing the Chasm, who in 2011 for AIIM drew the distinction between a system of record (transactions and storage) and a system of engagement (interaction, think social and mobile). Important: Moore's second layer was about how people interact with software, not who does the work. We borrow the naming but add a different axis: not storing, not interaction, but execution, a system of execution. That's the logic behind the four eras below.
Key takeaway: Don't judge HR software on its feature count, but on a single axis: does it organise your work, or execute it?
The hidden cost of "organising"
Why is this more than an academic distinction? Because "organising but not executing" has a real price, and in most organisations that price is invisible because it's spread across hundreds of small actions.
The burden is concrete. SHRM finds that 57% of HR professionals are working over capacity because of chronic understaffing, while a large share of HR time goes to admin rather than people and policy. McKinsey calculates that the current generation of AI can automate work activities that today consume 60 to 70% of working time. And manual work scales errors: EY research finds 1 in 5 payroll runs contain errors, at an average cost of 291 dollars per error. (These international numbers indicate the order of magnitude, not Belgian precision.)
The burden also lives in sprawl. According to Capterra's HR App Sprawl Survey (2023), companies use on average five HR systems, and HR says roughly half of them have overlapping functionality, with payroll (69%) and applicant tracking (68%) the biggest duplicates. Every link between those systems that doesn't run automatically is more manual work, and more than half of HR teams name data-migration complexity or cost as the reason they don't clean up that sprawl. Organising piles up instead of disappearing.
The four eras of HR software
These four eras are also four levels at which HR software still coexists today. For the side-by-side architecture comparison, see Agentic HR vs. traditional HRIS. For the market map, see the best HRIS in Belgium, and for the practical playbook, HR automation: the complete guide.
Era 1, Paper and spreadsheets (manual)
What it solved: a first form of structure, a place for files and lists.
What it left behind: everything. Employee files in folders, leave in an Excel, contracts in an email. Software, where present at all, mostly stored files. Workable for a handful of people, but every change is manual and every mistake stays invisible until it surfaces. And this era is far from over: estimates suggest around 90% of organisations still rely on spreadsheets for part of their most important data.
Era 2, The HRIS (system of record)
What it solved: fragmentation. The first truly integrated HRIS, PeopleSoft, arrived in 1988; the early 90s brought the big HR ERPs (Oracle, JD Edwards), the late 90s moved the software from client-server to the web, and from 2005 onward to the cloud with SaaS players like Workday (which IPO'd in 2012 at 9.5 billion dollars and today serves more than half of the Fortune 500). Employee self-service portals shifted some input work to the employee. The HRIS centralised personnel data into one source of truth: contracts, leave balances, documents, neatly versioned and auditable.
What it left behind: the work itself. The HRIS does nothing with that data; it remembers the result, you still perform every task. Many organisations sit here: a nice, orderly system, but as much work as before.
Era 3, Workflow suites (system of workflow)
What it solved: the chore of routing things around. Modern HR platforms added automation of the routing: requests, reminders, approval chains, status updates.
What it left behind: execution. The human still clicks through every step; the software organises the work more efficiently but doesn't perform it. This is where most HR suites sit today, often with an AI feature on top that doesn't change the level (see the next section). The automation is also bounded by the rules you set up in advance: the system doesn't act beyond those configurations and doesn't surface insight you didn't ask for.
Era 4, Agentic HR (system of execution)
What it solves: the work itself. AI agents execute the full workflow, across data, documents, calendar and payroll, and only pause for a human decision where policy or risk requires it. Not "AI helps you fill in the form", but "the form is filled in, sent on and followed up, just approve it". For the first time, work isn't organised better, it's taken away, and the human is left with judgment, policy and the exceptions.
Key takeaway: The first three eras organised work ever more efficiently. Era 4, agentic HR, is the first that executes it.
Why an AI feature doesn't lift you into the next era
This is the most common confusion. AI adoption in HR is growing fast, SHRM reports that 43% of organisations use AI in HR tasks, up from 26% in 2024, so at first glance everyone is already in era 4. But look at the use cases: the weight sits on recruiting and text generation (job posts, CV screening). That is augmentation: AI that helps you with work you still do yourself.
A chatbot or copilot bolted on top of a system of record stays in era 2. PwC puts the difference sharply: agentic AI "acts autonomously, making real-time decisions", whereas classical automation "relies on static rules and requires human intervention to adapt". The era doesn't change by adding a feature, but by who performs the task: the human (augmentation) or the agent (execution).
Key takeaway: An AI feature speeds up a sub-task; it doesn't shift you into a new era. Only when the software runs the full workflow and you only approve are you in era 4.
The false start: RPA, and why it broke
Agentic HR isn't the first attempt to make HR work execute rather than organise. The previous decade RPA (robotic process automation) tried exactly that: software robots mimicking screens and mouse clicks to handle tasks automatically. Execution in theory, brittle in practice.
RPA works with rigid, rule- and screen-position-based scripts, without understanding of context. Change a screen, a field or a process variant, and the bot breaks. EY estimated in its Get ready for robots report that 30 to 50% of RPA projects fail, with the bulk of cost in implementation and maintenance rather than licences. In HR that's especially visible: an onboarding bot that doesn't know an exec needs different access or clauses than a blue-collar worker quietly misses steps. RPA automates the rule, not the judgment, and at an exception it mostly does the same wrong work faster.
That's the exact difference with era 4. An agent doesn't follow a fixed script; it gets a goal, reasons about the situation, picks the right tools and brings a human in where judgment is needed. Where RPA broke at every deviation, agents are built precisely to handle deviation. Era 4 isn't "more RPA", but a different kind of execution: adaptive instead of brittle.
Key takeaway: RPA tried execution through rigid scripts and broke at every change (EY: 30 to 50% of projects fail). Agentic HR executes by reasoning and adapting, not by mimicking rules.
Why it's tipping now
The leap to era 4 is no longer theoretical, and it's the biggest independent HR authorities that confirm it.
Josh Bersin, the best-known independent HR analyst, names the era explicitly as "Agentic HR" and describes the path "from assistants to agents to superagents". He predicts AI "superagents" will automate up to 30% of core HR roles by 2026 (Josh Bersin Company).
Deloitte states that by 2027 half of the companies using generative AI will have launched agentic AI applications that perform complex work with limited oversight (Deloitte, Global Human Capital Trends 2025).
Gartner expects 40% of enterprise apps to contain task-specific AI agents by 2026, up from less than 5% in 2025, and agentic AI to be present in 33% of enterprise software by 2028 (vs less than 1% in 2024). In HR specifically, a Gartner survey suggests around 82% of HR leaders plan some form of agentic AI within twelve months.
But the direction isn't a given. Gartner also warns that more than 40% of agentic AI projects will be cancelled by the end of 2027, due to rising costs, unclear value or insufficient risk controls. In other words: the technology makes era 4 possible, but only when it's well governed. And understanding lags adoption: an Everest Group poll found that only 22% of HR leaders fully understand the difference between agentic AI and classical AI. The technology is accelerating faster than the insight, which is exactly why this is the moment to grasp the evolution clearly.
What era 4 demands of HR
Taking work away sounds like less responsibility for HR. In practice, the role shifts, it doesn't disappear. Three things change:
HR becomes a supervisor, not an executor. As Josh Bersin puts it, we're "not losing the HR department" to agentic AI; the role shifts from performing tasks to supervising, reviewing and approving agents. That's higher, not less, work.
Business rules need to be made explicit. An agent only performs as well as the rules that are codified. If you want it to handle leave or a pay change correctly according to your own priority rules, that knowledge has to come out of someone's head and into the system. Making it explicit is the real work behind era 4, and the layer that's hardest to copy.
You want the outcome, not a new role as an operator. A trap with many "AI in HR" and build-it-yourself automation: they turn the buyer into an operator who has to orchestrate agents and maintain flows themselves. That's shifting work, not removing it. True era 4 means the software carries the orchestration, with approval gates for sensitive actions, scoped permissions, a trail of every action, and, for European employers, EU hosting, GDPR compliance and a certification like ISO 27001.
Which era is your HR in?
The four eras coexist today, and most organisations sit in era 2 or 3. The relevant question isn't which features you're still missing, but in which era you want to operate.
Quick diagnosis, the first "yes" is your era:
- Do you mainly run HR in Excel, email and folders? → Era 1.
- Do you have one central system, but still perform tasks manually? → Era 2.
- Does your system automate requests and approvals, but you still click through every step? → Era 3.
- Do you give an instruction and the system runs the workflow, with approval where needed? → Era 4.
A useful second question: where does your last hour of HR work go? If the answer is "re-typing data, chasing statuses, preparing exports", you're in era 2 or 3, no matter how modern your tool looks.
At Peepel, era 4 isn't theory. Today, the platform already executes concrete HR actions end-to-end instead of merely teeing them up: payroll pushes to the social secretariat, approved leave requests, generated contracts and onboardings, and AI-generated surveys. That's exactly the difference between era 3 and era 4 - the work gets done, you only approve.
Ready for the next era of HR software?
Peepel is the agentic HR platform for European companies, era 4: it executes HR work instead of organising it. Book a 30-minute demo and we'll first look together at which era your HR is in today.
FAQ
- What are the eras in the evolution of HR software?
- (1) Paper and spreadsheets; (2) the HRIS / system of record; (3) workflow suites / system of workflow; (4) agentic HR / system of execution. Each step organised work better; only the last one executes it.
- What's the difference between an HRIS and agentic HR?
- An HRIS (system of record) stores and organises your HR data; you execute the tasks. Agentic HR executes the tasks itself and only asks for approval where it has to: from organising to executing.
- Which era are most companies in today?
- Era 2 or 3, a system of record or a workflow, often with an AI feature on top that doesn't change the era.
- Does an AI chatbot make my HR software agentic?
- No. A chatbot or copilot helps you with work you still do yourself (augmentation). Era 4 means the software runs the full workflow and only asks you to approve.
- Do I need to replace my HRIS to reach era 4?
- Not necessarily in one step, but an AI feature on an existing HRIS won't lift you into era 4 - that requires software built around execution. An agentic platform takes on the system-of-record role as well.
- Is agentic HR safe?
- When properly designed, yes: approval gates for sensitive actions, scoped permissions, a full audit trail of every action, and EU-hosted, GDPR-compliant, ISO 27001-certified infrastructure.
- What is agentic HR?
- HR software at level 4: AI agents that autonomously execute HR work instead of organising it, with human approval where policy requires.
More Peepel resources · Book a demo