People Operators are the Future of HR

Everyone building software right now faces the same question: if anyone can prompt an AI to build anything, why would they buy your product?

It deserves a straight answer.

Today you can open any AI tool, describe what you need, and get a working solution in minutes. Onboarding checklist? Done. Leave request form? Built. Compliance workflow? Generated. For single-problem use cases, the economics are punishing traditional software. Why pay for a product that covers 80% of what you need when you can prompt something covering 100%, tailored to you?

This will wipe out a significant number of software companies. Narrow, single-function tools that made data searchable or encoded simple logic into rigid interfaces are in deep trouble. The moats that kept competitors out for decades, specialized UIs people spent years learning, hardcoded business logic maintained by expensive domain engineers, public data wrapped in a proprietary layer, are collapsing.

But general-purpose models will not win everything.

And the reason comes down to something more fundamental than technology. It comes down to how work itself is changing.

The bar for building is dropping. Everywhere.

Over the past two years, AI made it possible for people with no programming background to build websites, apps, and full products by describing what they wanted. The bar for creating software dropped to the floor.

This same force is now moving beyond code.

It is starting to reshape how companies operate. Not the software behind a product, but the processes behind an organization. How you bring someone on board. How you handle a benefits claim. How you verify credentials for a new hire in aviation before they step near an aircraft. How you cycle seasonal workers in and out without losing continuity. How you make sure a departing employee's access gets revoked across every system they touched.

Here is the shift. When you can describe a task and an agent can execute it, you stop spending your day doing tasks. You start spending it mapping, building, running, and governing processes.

The human becomes the architect of how work flows. The system handles execution.

A farmer ten thousand years ago was called a farmer. A farmer today is still called a farmer. The responsibilities have not changed. Grow food. Bring it to harvest. Feed people. But the tools look nothing alike. Tractors replaced oxen bulls. GPS replaced intuition. Sensors replaced walking the field. Every generation, the way tasks get completed transforms while the underlying job remains.

The same shift is now happening across every knowledge profession, faster and more visibly than any previous transition.

This is what we are building for at 50skills. A platform that makes it as easy to set up and run people processes. Just like AI has made it easy to code, build apps and websites. We’re lowering the bar for millions of people to map and build processes which only the tech-savvy 1% was able to do before.

Why context is the moat that holds

Most conversations about AI moats focus on data. Data is important. But there is a more specific version of the argument that matters.

The moats being destroyed right now are the ones built on interface complexity, hardcoded business logic, and data access layers. Three examples:

1. Specialized UIs with years of muscle memory? -> Natural language collapses them into a single conversation.
2. Business logic spread across thousands of lines of code maintained by rare engineers who understand both software and the domain? -> LLMs turn that into something anyone with domain expertise can describe in plain language.
3. Public or licensed data made searchable through proprietary parsers? -> The foundation models themselves are now the parser.

The moats that hold are different. Proprietary data that cannot be replicated. Regulatory and compliance infrastructure. Network effects. Transaction embedding. System-of-record status. These are structural. They do not get cheaper because intelligence gets cheaper. If anything, they become more valuable as everything else commoditizes around them.

The question for any vertical software company is simple: which side of that line are you on?

50skills is on the side that holds.

We are a no-code platform. Our customers encode their own processes without engineers. We do not make public data searchable. We do not depend on interface lock-in. Our data, thousands of real workflows built by customers across all verticals with more than a million actions completed through them in the past year alone, is proprietary and compounding. Our value proposition is rooted in governance and compliance, the exact domain where regulatory requirements create durable switching costs. And we are becoming the operating layer where people processes are defined, executed, and monitored, which is the foundation for system-of-record status in people operations.

The forces destroying traditional vertical software are creating the conditions for platforms like ours.

What the "prompt everything" crowd misses

When a general-purpose model builds a leave request tool, then a compliance form, then an onboarding flow, each works on its own. But you end up with a dozen disconnected solutions scattered across the organization.

Who controls access? When someone leaves, how do you revoke permissions across all these tools? When someone joins, how do they discover what exists? When something breaks, where do you look?

Governance disappears. Visibility disappears. The consistency holding an organization together disappears.

This matters everywhere. It matters most for people data. Compensation. Health information. Performance records. Compliance status. Sensitive, regulated, consequential. You need access controls, audit trails, and a coherent interface designed for exactly this. The stakes are too high for stitched-together prompts.

There is a reason every industry with high-stakes data ends up with purpose-built vertical platforms. Healthcare has them. Financial services has them. Legal has them. And make no mistake HR will have them.

A new category is emerging, people operations, which will consist of people that will map, build and manage flows of people and AI agents running organizations.

The deeper advantage comes from what sits beneath the governance layer: accumulated domain context.

When a new customer describes their company and asks the system to help build an onboarding process, it should not start from a blank page. It should draw on what has worked for similar companies, similar sizes, similar industries, similar regulatory environments. It should know that aviation companies need credential verification workflows that are fundamentally different from retail companies needing seasonal onboarding at scale. It should understand that a 200-person company in Denmark has different compliance requirements than a 2,000-person company in the UK.

A general-purpose model can generate a plausible workflow. It cannot tell you which one performs. That distinction is the entire game.

What customers are building today

The range of what companies build on 50skills goes far beyond what most people picture when they hear "HR tool."

Onboarding is the most common workflow, running across hundreds of companies. But right behind it: temp worker cycling for companies with high seasonal turnover. Event registration for internal programs. Benefits claims. Offboarding. Role changes. AI-powered candidate screening. Contract management. Incident reporting. Expense reimbursement.

Then it gets more specific. Travel requests. System access provisioning. Background checks involving police clearance for airport workers. Salary change approvals. Stock option administration. Criminal record submissions. Equipment orders. Parental leave applications. Payroll requests. Remote work agreements.

And then the edges: Christmas gift coordination. Employee kudos programs. Birthday automations. Company medical appointment scheduling. Suggestion boxes. Language education requests. Even internal games.

Aviation companies use the platform to manage seasonal crew end-to-end: credentials, licenses, training, offboarding, and nurture sequences to bring workers back next year. Hotels use it to cut benefits processing time by 80%. Fast-growing companies use it to reduce onboarding time by 90%.

Each of these workflows started simple and evolved. Customers iterate as they grow, enter new markets, or face new regulations. Each person moving through a workflow generates signal: completion rates, bottleneck patterns, time to resolution, escalation triggers. All generating context that makes the next workflow smarter.

This is not a dataset you can assemble from scratch. It is the accumulated output of all organizations solving real problems over years. And it compounds.

Describe your thoughts and review the output

At the core sits an AI agent you can talk to. Describe what you need, and it builds the workflow. Need an onboarding process for a new office in Germany with specific compliance steps? Describe it. The agent creates the stages, adds the right integrations, and gets it running. Something broken in your offboarding flow? Tell the agent what is not working. It updates the process.

The agent is live today. We use it internally to build most workflows, and customers in our beta program are building and updating workflows through voice and chat. It handles the majority of requests correctly, and it is only getting better day by day.

But the more important architectural decision is this: the platform is designed for both people and agents to access. Today, you can connect any workflow to an external AI through an open protocol, describe how it should be triggered, and let people interact with it through conversation. Someone can ask their AI assistant "I need to go on paternity leave, what do I do?" and it will find the right workflow, explain the steps based on company policy, and trigger the process if the person agrees. This works now. It will only get smoother, faster, and more capable. But the foundation is already there. And it works for any process in any language in any modern AI chat tool via MCP.

It works for agents that people are starting to deploy to do tasks for them. And that’s where this is heading: agents that can enter the platform autonomously, review every active workflow, analyze how each process performs, identify bottlenecks, draft improvements, and present summaries. External agents, internal agents, and people interacting with the same system. The platform becomes the shared operating layer where processes get built, run, monitored, and improved, regardless of who or what is doing it.

The entire premise is lowering the bar for building processes to the same level AI has lowered the bar for building products. If you can describe what you need, you can build it. You should not need to be an engineer, a systems architect, or even particularly technical to set up how your organization operates.

Where humans belong

Here is the part most people get wrong about AI and people operations.

It will not replace the function. It will not automate away departments. What it will do is handle everything manual and redundant so people can work where their strengths matter.

The core responsibilities have not changed and will not change. Empower employees. Help people become the best version of themselves. When something goes wrong, handle it with care. Ensure people feel heard. Keep values visible. Make objectives clear.

What is changing is how these responsibilities get fulfilled. Today, the knowledge of how things work lives in people's heads. When the person who built the onboarding process leaves, the knowledge walks out the door. Tomorrow, that knowledge lives in the platform. Structured. Governed. Continuously improving.

New employees will interact through voice, through text, through their own personal agents sharing information with company systems on their behalf. The experience on every side moves closer to seamless.

And because this touches people, there will be laws. Different countries, continents, and unions will put rules in place. Human-in-the-loop requirements will expand, not shrink. Approval flows, review steps, sign-offs. The platforms built for this future need governance baked in from the start. Not bolted on.

Some decisions should remain stubbornly, necessarily human and in some regions enforced by law. Termination conversations. Sensitive accommodations. Restructuring decisions requiring empathy, judgment, and presence. The platform should make these boundaries explicit. Surface where humans must stay in the loop. Ensure AI accelerates everything around those moments rather than replacing them.

The person running people operations in the future is an architect. Designing how the organization functions. Deciding where automation belongs and where it does not. Using a platform to execute at a speed and quality impossible before.

This function may not be called HR much longer. What we are describing is people processes. People operations. The work of designing, running, and governing every process touching every person in an organization. And soon AI Agents working with people in that organisation. The name changes. The need does not.

The pattern

Every technology wave follows the same arc. When the internet arrived, it was slow. Your desktop handled everything faster. But the internet could do something the desktop could not: accessible from anywhere, shareable with anyone, at any time. Speed caught up later. The transformative quality was always there.

AI follows this arc. Early on, we worry about hallucinations and reliability. A calculator will always know five times five is twenty-five. You can count how many r's the word strawberry has, but an AI couldn't. These concerns are valid today. They will smooth out. The capabilities AI enables, building and running processes through conversation, adapting in real time, learning from outcomes across thousands of companies, are so fundamentally different from what existed before that the rough edges become irrelevant once the foundation matures.

There are things nobody knows. Whether the dominant models will be open source or proprietary. Where they will be hosted. Which regulations arrive and which do not. How fast agentic capabilities mature. How quickly enterprises adopt. Known unknowns everywhere.

What is not uncertain is the direction. The shift from tasks to automated workflows is happening. AI agents will function alongside people, carrying out work like employees, at a fraction of the cost. The people function, which holds the map of every role and responsibility in a company, becomes the orchestration layer for a hybrid workforce of humans and machines.

What we carry into this

I have been in the startup ecosystem for close to 20 years. Built my first website as a kid. Ran accelerators and startup competitions in Iceland. Watched hundreds of founders build, fail, and build again. Scaled inside a fast-growing company where we hired so fast we had no time to build proper processes. The experience of drowning in broken workflows as the person responsible for fixing them is what led me to start my own company. With AI I have never experienced anything like the change we are all going through.

The lesson from all of it: the companies that win are not always the ones with the most advanced technology. They are the ones who understood the problem early enough and built the right context around a solution before anyone else understood why it mattered.

We do not know how the next years unfold. Nobody does. But lowering the bar so anyone can build and govern people processes is one of the most important shifts in software. The vertical of people operations is one of the most consequential places to prove it. It will be the vertical that manages people and agents moving forward. The core of every organization.

HR is evolving into people operations. Operators designing, executing, and governing processes for a workforce that is part human, part AI.

The sooner a company recognizes that shift, the sooner it builds the infrastructure to win in it.