HR AI consulting firms: the kinds, what each delivers, how to choose

Search for an HR AI consultant and you get a list. The list mixes a global consultancy, an automation agency, an HR software vendor and a research membership. They all say “AI for HR,” but they sell very different things.

So skip “which firm is best” and ask what you’ll hold on the day the engagement ends. Some firms leave a roadmap. Some leave a trained team. Others leave a running system, or a feature inside their own platform. Each is a different purchase at a different price, and each leaves your team a different job afterwards.

Below, we sort HR AI consulting firms into six kinds by what you’re left with. For each kind we say which company size it suits and which first-call questions tell it apart.

One disclosure first. Culturro is one of these firms. We’re an HR-AI strategy consultancy first and a builder second, and we work with companies of 500 to 5,000 people. We’ve kept ourselves out of the table and put our own offer in a separate section after the six kinds, so you can judge them without our thumb on it.

The six kinds of HR AI consulting firm

The firms in each row are examples, taken from their own websites as read on 19 October 2026. This is not a ranking. Each row fits a different situation.

Kind of firm	Examples	What you hold on the last day	What your team still has to do	Who it tends to suit
Big 4 and global HR or strategy consultancies	Mercer, The Hackett Group	A programme plan, operating model and change plan, often across the whole enterprise	Fund and staff the build, or buy it from the same firm	Large enterprises running a multi-year transformation
HR-transformation boutiques with an AI line	HCM Strategy, Culture Consultancy	Advice, policy and governance, workshops, a trained HR team	Choose the tools and get them built or configured	Small and mid-size teams that need judgement and change support more than software
AI and engineering agencies that build	Folio3, Phos AI Labs	Working agents or automations	Own them, run them, and check they’re fair to employees	Any size, once you know which use case you want
HRIS and HCM vendors’ consulting arms	MHR	Use cases built inside tools you already have, plus a team that can keep using them	Decide how far you want to depend on one vendor’s ecosystem	Any size, if you plan to stay in your current stack
Research and membership advisory	The Josh Bersin Company	Research, benchmarks, peer networks, analyst time	Turn the research into a plan and a build	Large HR functions with a strategy team to do the turning
Fractional experts and independents	Varies by person	A named person’s time and judgement	Everything a team of one can’t cover	Smaller companies, or a gap between hires

What each kind delivers, and what it leaves with you

Big 4 and global consultancies: the enterprise programme

These firms treat AI as one part of a larger change. Mercer’s transformation practice describes AI alongside culture, structure, systems and work design. It also lists “poor implementation of AI and other technologies” as one of the problems transformation runs into. The Hackett Group lists Gen AI in HR as one of its AI services and also sells its own AI products.

Their strength is scope: a whole-enterprise view, a large bench and a brand a board already trusts. What they leave you is usually a programme. If the build is part of the contract, it’s often the biggest line in it. That scale is usually more than a company of a few hundred people needs.

HR-transformation boutiques: judgement and change

These are HR consultancies that added AI to an existing practice. The AI offer is mostly advisory. It covers readiness reviews, governance and policy, plus training and change work. That’s our own read of the firms we’ve mapped, and the firms say much the same about themselves.

HCM Strategy calls itself an “HR and AI integration consulting firm.” Its FAQ says plainly that it doesn’t sell software or platforms. It helps clients evaluate, select and integrate AI tools, and it builds the governance and training around them. Its home page says its frameworks fit teams of “10 employees or 500,” and it works through project, retainer and fractional models.

Culture Consultancy lists AI adoption as one of its services. Its own description is about “building the culture, skills and ways of working” that adoption needs.

This kind is strong on policy, training and change. You’ll come out with clear policy and an HR team that knows how to use the tools. You won’t come out with a working system unless someone else builds it.

AI and engineering agencies: the build

These firms write the software. Some specialise in HR, but most build for any function.

Folio3 sells custom HR AI agent development. Its page covers discovery and integration with HRIS, ATS and payroll systems. It also lists approval gates, audit logs and monitoring after launch. The page says sensitive employment decisions still need review and approval by an authorised person. One detail in its own comparison table is worth reading closely: under ownership it says “contract-dependent custom IP.” That’s honest, but it means you need to read the contract.

Phos AI Labs describes itself as “the AI department for $5M+ companies.” It runs strategy, training and build in three phases and offers to “stay embedded as your long-term AI team.” Phos also publishes a ranked list of AI consulting firms for HR that places Phos first.

The strength here is speed to something that works. The risk is building the wrong thing well. An agency will build what you ask for. Whether that use case is worth building is your call. So the strategy needs to exist first, whoever writes it.

Culturro builds too, so we have a stake in this kind. Our build comes last, after a strategy month and a proof on test data. At handover, the code and prompts are yours to run. Whoever builds for you, get that ownership written into the contract. Our full offer sits further down, under “Where Culturro fits.”

HRIS and HCM vendors’ consulting arms: inside your current stack

Some HR software vendors now sell AI consulting. MHR’s AI consultancy page describes three steps called Understand, Build and Empower. It builds use cases “directly within tools like Copilot or ChatGPT,” works with the HR team hands-on, and says you don’t need to be an MHR customer.

This kind is practical and usually quick. It suits dashboards, policy chatbots and small automations in tools you already pay for. The trade-off is independence. A vendor’s consultants know their own ecosystem best. That’s useful if you plan to stay in it, and worth naming out loud if you might not.

Research advisory and fractional experts: knowledge and a person

The Josh Bersin Company sells research and memberships for CHROs and for vendors. It also runs the Josh Bersin Institute and sells Galileo, which it calls “AI intelligence for HR.” Analyst firms in this kind give you a view of what peers are doing and what the market looks like. They don’t run your project.

Fractional experts and independents sell one person’s time. A good one can write a roadmap and sit in vendor meetings with you. One person can’t build, test and run a production system alone. Some boutiques blur this line too, and HCM Strategy also offers fractional leadership.

Where Culturro fits

We’re a mix of two of the kinds above. Our strategy work overlaps with the HR boutiques. Our build work overlaps with the agencies. We do strategy first and build only what the strategy calls for.

The engagement runs in three steps, each on a fixed fee and each ending with a written go or no-go:

1. Strategy, about a month. You get a sequenced roadmap, a business case for each use case, and written guardrails for your security team and works council. You also get an inventory of AI already in use in HR. The roadmap is yours whether or not you go further.
2. Proof, a few weeks per use case. We build a working system on synthetic or de-identified test data, in our sandbox or in your own cloud tenant. Your team uses it, and we measure accuracy, failure cases and cost per query.
3. Build, scoped after the proof. A production system in your environment, under your security controls. The code and prompts transfer to you at handover. So do the evaluation sets, documentation and runbooks. You run it, with no dependency on a Culturro platform.

Two limits we hold to. We don’t build systems that make autonomous decisions about hiring, pay or promotion, and the same goes for discipline and termination. AI can inform, rank or draft, but a human decides. We also hold no certifications and claim none. The system runs under your controls, so what your team assesses is our process.

If you’d like to see what agents can do before a call, the HR Agent Map lays them out by HR process. Our explainer on AI agents in HR covers the basics. Our post on what to tell the board about HR AI maturity helps you place your function today.

What you hold at the end of each kind of engagement

The table above gives the short version. Here’s what to check before you sign, deliverable by deliverable.

A roadmap. Almost every kind can write one. Ask what’s in it. A list of use cases is not a roadmap. A useful one puts the use cases in order, attaches a business case to each in numbers your CFO already tracks, and names an owner. We wrote a post on how to prioritize HR AI use cases if you want to order them yourself.

A working system. Only agencies and vendor arms, some Big 4 programmes and a few hybrid firms deliver one. Ask where it runs. A system in your cloud tenant and under your security controls is very different from a hosted platform you rent.

A trained team. Boutiques and vendor arms tend to be strongest here. A complex agent still needs someone technical to run it.

Who runs it after handover. When the firm leaves and nobody owns the system, usage drops. That handover gap is a big part of why HR AI pilots stall. Name the owner on your side before you start.

Who owns the code. Read the contract, not the website. You should receive the code and prompts if you’re paying for a custom build. Test sets, documentation and runbooks should come too. If the firm keeps the IP, you’re only renting the system.

Whether IT security can approve it. Ask for the architecture document and the data-flow diagram before the first security review, not after the build. A firm that can’t produce them early will slow the review down later.

Which kind fits which company size

Company size changes the answer more than any other factor. These lines are our view, based on who each kind says it serves.

Under about 500 people. A fractional expert or an HR boutique is often enough. HCM Strategy says its frameworks fit teams of 10 to 500. A vendor’s consulting arm can also work well if the plan is fast results in tools you already own.

About 500 to 5,000 people. This is where the gap is widest. The company is big enough to need real systems and real governance. It usually has no AI team of its own. The Big 4 are often too large a purchase. An agency on its own may start building before anyone has agreed what’s worth building. A firm that does both strategy and build tends to fit. So does a boutique paired with an agency. This is also the band Culturro works in, so weigh our view here with that in mind.

Above about 5,000 people. Programmes at this size usually span many countries, several works councils and a large IT estate. This is where the Big 4 and global consultancies are built to operate, and where research memberships pay off because there’s an in-house team to use them. We mostly leave this band to the large players.

Questions that tell the kinds apart in a first call

Take these into the first meeting. Each group shows you something different about the firm. Before the call, our guide to HR data readiness covers what to pull together on your side.

What you’ll actually get

What will we hold on the last day: a document, a trained team or a running system? Is the build part of this contract or a separate one? Or is it someone else’s job? If we stop after the first phase, what do we keep?

Who owns and runs it

Who owns the code, prompts and documentation when you hand over? Where does the system run, and whose security controls does it sit under? After handover, what do we need on our side to keep it running?

Whether they’ll tell you no

Which use case on our list would you advise us not to build, and why? What’s a project you turned down?

Who answers for people decisions

Which decisions will this system make about employees, and which will a human make? Is that written into the design, or only into a policy? Who on our side is accountable when the system gets an employee case wrong?

This group matters most. A system that ranks candidates or flags performance risk touches someone’s job, and the firm should have an answer ready.

Independence

Do you sell or resell any software, or earn fees from a vendor you might recommend?

EU and UK obligations for HR AI

If you employ people in the EU, the EU AI Act treats many HR uses of AI as high-risk. Annex III lists systems used to recruit and select candidates. It also lists systems used to decide on promotion or termination and to allocate tasks. A last group covers systems that monitor and evaluate performance.

Article 26 requires employers that use such systems to provide human oversight by people with the competence, training and authority to do it. It also requires employers to inform workers’ representatives and affected workers before a high-risk system goes live at work. These Annex III obligations apply from 2 December 2027 after the Digital Omnibus amendment. That’s close enough that a system you scope this year should be designed for it now.

The UK’s Data (Use and Access) Act 2025 widened when organisations can make solely automated decisions with significant effects. It still requires safeguards. People must be told about significant decisions, be able to challenge them, and be able to get human intervention.

When you choose a firm, ask each one how its design meets these rules, and whether it will put that in writing for your works council or employee representatives. A firm that has done this before will answer quickly. This is not legal advice, and your own counsel should review any system that touches hiring, pay or performance.

The next step

If you’re comparing firms, bring your shortlist and the numbers you already report on to a 30-minute call. We’ll tell you what we’d do first, which kind of firm we think fits, and whether that’s us. We build on your own systems, and the result is yours to run. If you want to look around first, start with the HR Agent Map.

Frequently asked questions