The Agentic Maturity Ladder: A practical guide for HR leaders trying to make sense of “AI Agent” claims

There is no shortage of ambition in today’s HR technology market. In the past year, we have seen an explosion of platforms claiming to use “AI agents” to transform recruitment, onboarding, service delivery and talent management. It is an exciting moment, filled with possibility, yet also thick with confusion. Every demo sounds compelling, every slide features the same cluster of words, and every vendor insists their product is already operating at a frontier of autonomy and intelligence.

But when you look beneath the surface, the reality is often different. Many of these supposed agents cannot act without a human clicking a button, selecting a workflow or approving every step. Some rely entirely on rules engines. Others are powered by an LLM that generates helpful text yet has no control over tools, systems or decisions. A few are genuinely pushing towards autonomous capability, but even then, their language rarely matches their maturity.

As a result, HR and talent leaders are being asked to evaluate technologies without a shared reference point. They need to understand what they are being sold, what is feasible, what is theatre and what is genuinely new. The industry needs clearer language and better frameworks to navigate the next decade of AI in the workforce.

This is where The Agentic Maturity Ladder comes in.

It is a simple, evidence-informed model that describes the five stages of development from basic automation to true AI agents. It helps buyers understand what they are seeing, helps vendors describe their capabilities more accurately and creates a shared map for organisations planning their AI journey.

It is also the framework behind the Agent Impostor Test, a light-hearted but research-grounded assessment designed to classify HR tech claims and cut through the noise. You can try it here:👉 https://agent-analyser.eunomia-hr.com(The assessment is gated, and you will receive your results and category badge by email.)

But first, let us walk through the ladder itself.

Why agents matter now

To understand the ladder, it helps to revisit what distinguishes an agent from a chatbot or a workflow. In computer science, an agent is defined as a system that can:

• Perceive its environment

• Reason about what it perceives

• Plan a course of action

• Act upon those plans

• Learn from the outcome

This is not a new definition. It predates generative AI by decades. What has changed is that large language models, tool-calling architectures and retrieval systems have made these behaviours more accessible and more powerful in practical business contexts.

True agents are not simply conversational. They are not reactive. They do not wait for a prompt. They operate with some level of autonomy, within guardrails, and are capable of meaningful delegation.

Very few HR platforms are there today. Many are on the way. Some are still quite far from the destination but speak as though they have already arrived.

The Agentic Maturity Ladder was built to make these distinctions clear.

Introducing the Agentic Maturity Ladder

The ladder has five stages. Each reflects a distinct shift in capability, design and organisational impact.

Stage 1: Basic Automation

“If X happens, do Y.”

This is the foundation of most HR technology. Rules and triggers drive predictable actions. If a candidate applies, send a confirmation email. If a case is raised, assign it to a queue. If a new starter is added, generate a checklist.

There is nothing wrong with this level. It is reliable and efficient. But it is not intelligent, nor does it adapt. The actions are entirely predefined.

Many HR teams still rely on this stage for core operations. Many vendors still sell this stage with a modern interface. It is familiar territory.

Stage 2: Automation with Lipstick

“A workflow wearing an AI mask.”

At this stage, vendors take deterministic automation and add a layer of generative capability. A chatbot answers questions. An LLM drafts messages. A user can converse with the system rather than navigating menus.

The underlying workflows, however, do not change. The system cannot create new plans, adjust its actions or make decisions beyond the rules built for it. It appears more intelligent than Stage 1 but behaves exactly the same once the cosmetic layer is removed.

This is the most common type of “AI agent” currently marketed in HR technology. It is helpful, and sometimes delightful, but fundamentally predictable.

Stage 3: Assisted Agentic Systems

“The system can reason, but still asks permission.”

This is where generative AI begins to influence process design. The system can use an LLM to interpret requests, propose steps, surface insights or suggest alternative actions. It can call tools or APIs, but usually only after explicit confirmation.

This stage introduces emerging agentic behaviours: planning, reasoning and tool use. The system begins to look more like a partner than a workflow, yet it remains human-directed. It is capable, though not autonomous.

Many organisations will spend the next few years moving between Stage 2 and Stage 3, experimenting with this new flexibility.

Stage 4: Agentic AI

“The system can act within boundaries.”

This is the first true shift into agentic capability. The system can:

• Create multi-step plans

• Act across systems

• Maintain short- and long-term memory

• Initiate work within defined triggers

• Adapt its behaviour based on context

It still needs guardrails. It still needs organisational governance. But it can execute meaningful tasks without requiring human approval for each step. This is where vendors start to offer genuine delegation.

For HR, this means an agent might coordinate interview scheduling, resolve standard employee queries or progress candidates autonomously, while escalating exceptions or risks.

Few products sit squarely at Stage 4 today, but several are approaching it quickly.

Stage 5: True AI Agents

“Perceive → reason → plan → act → learn.”

Stage 5 is the point where the full definition of an agent is realised. A system at this level has:

• Continuous awareness of its environment

• The ability to set sub-goals and update plans dynamically

• Proactive initiation, not just reactive responses

• Coordination with humans and other agents

• Feedback loops that genuinely alter future behaviour

This is the frontier of enterprise AI. It is powerful and transformative, but also complex. It demands strong governance, thoughtful policy and a clear understanding of where autonomy helps and where it introduces risk.

Very few HR technologies are at Stage 5 today, and that is entirely expected. The industry is only beginning to explore what responsible autonomy looks like.

Why the ladder matters for HR leaders

The Agentic Maturity Ladder is not an academic exercise. It is a practical tool for decision-making.

1. It makes vendor claims intelligible.

When a salesperson says, “We have autonomous recruiting agents,” you can now ask:

• Which stage of the ladder does this actually match?

• Can you show me examples of planning, tool use and autonomy?

• Does your system act without human initiation?

• Does it learn from outcomes?

• What guardrails are in place?

If their answer places them at Stage 2 while their slide suggests Stage 5, you have clarity.

2. It brings structure to buying decisions.

Most organisations do not need Stage 5 capability immediately. Few are ready for that level of autonomy. Many will see significant value in Stage 3 or Stage 4 systems that improve efficiency without altering job roles or governance.

The ladder helps teams understand what they need rather than what they are being sold.

3. It exposes the real organisational challenge.

The shift to agentic capability is not only technological. It affects:

• Operating models

• Process ownership

• Policy and risk management

• Workforce design

• Data governance

• Employee experience

Technology readiness is only half the picture. Organisational readiness is the other half.

4. It aligns HR teams, IT and leaders.

The ladder provides shared language across departments. It removes ambiguity and supports strategic planning.

Introducing the Agent Impostor Test

To make the ladder practical, I created the Agent Impostor Test. It is a short, cheeky diagnostic that classifies technologies into four categories:

• AI Agent

• Agentic AI

• Automation with lipstick

• None of the above

You can try it here:👉 https://agent-analyser.eunomia-hr.com

It is gated, so you will need to add your name and email, but you will receive:

• Your full results

• Your maturity classification

• An official badge that you can share

• A quick interpretation guide

It is designed to be fun, but it is grounded in well-established agent frameworks and current research in AI systems.

Vendors can use it to benchmark their claims. Buyers can use it to evaluate demos more critically. HR leaders can use it to educate teams and spark discussion.

What HR teams should do next

Understanding the ladder is a first step. Acting on it requires a structured approach.

1. Map your current HR tech stack against the ladder.

Most organisations will discover:

• Some systems at Stage 1

• A few appearing to be at Stage 2

• Rare examples at Stage 3 or beyond

This is a helpful baseline.

2. Decide where you want to be.

Not every part of HR needs agentic capability. Ask:

• Where would autonomy add value?

• Where is it unnecessary?

• Where is it risky?

• What capabilities do employees expect?

• What does the business expect?

3. Reassess vendor relationships.

Ask vendors to articulate:

• Their current stage

• Their roadmap

• Their governance model

• Their integration strategy

• Their data and memory architecture

If they struggle to answer, you have insight into their maturity.

4. Build internal readiness.

Agentic systems require:

• New governance routines

• New process design principles

• Clear accountability

• Transparency and auditability

• Skills development

Many organisations will need support navigating this shift.

Where I can help

Much of the confusion in today’s HR tech market stems from a gap between ambition and reality. The Agentic Maturity Ladder and the Agent Impostor Test were created to help close that gap. They are free to use, easy to share and designed to support critical thinking.

For organisations that want more structured support, I offer:

AI due diligence for HR tech buyers

A clear evaluation of vendor capability, claims and risks.

AI maturity assessments for HR teams

Where you are today, what is feasible next and how to bridge the gap.

Operating model design for agentic HR

Governance, accountability, process redesign and workforce implications.

Executive education and workshops

Helping leaders understand what agentic capability means for their strategy.

If you would like to explore these, please get in touch.

Final thought

We are entering an era in which AI will take on meaningful work alongside HR teams. It will not replace the human elements of our profession, but it will reshape the operational core of it. To navigate this shift responsibly, we need clarity, shared language and practical tools.

The Agentic Maturity Ladder offers a way to think about the journey.The Agent Impostor Test offers a way to diagnose where we stand today.

Together, they help us move from marketing noise to informed action.

👉 Try the assessment: https://agent-analyser.eunomia-hr.com I would love to hear where your systems land.