Expertise Is Being Repriced. HR Is Next.

I have watched cash rich clients spend millions on HR technology and get almost nothing back. Big name platform. Long implementation. A steering committee with monthly minutes. Two years later the HR Director is still exporting to Excel to answer a board question, and the CEO is asking me why nothing changed.

The answer is nearly always the same, and it has very little to do with the software. They automated on top of a mess. Inconsistent job titles, three versions of the org chart and policies describing a company they stopped being in 2019. The platform did exactly what it was told. It made the mess faster and more expensive.

That has been a costly mistake for years. In 2026 it became a dangerous one.

This year has been unkind to services businesses in this region. I have spoken to owners and CEOs in events management, logistics, supply chain, recruitment and professional services who all describe the same few months. Contracts cancelled or frozen. New clients not signing. Decision cycles stretching from weeks into months. The clients who stayed paying inconsistently.

Underneath that, something else was happening. Frontier AI models started producing outputs in 30 or 40 minutes for work that used to be a 3 to 4 week project.

Most leaders I speak to cannot separate the two. Is demand down because of the uncertainty, or because the work itself is being repriced by AI? I have not fully figured this out either. My best guess is that it is a little bit of both. But the second one does not go away when the news improves, and that is the one worth preparing for.

Expertise is being repriced, and HR is next

Let me take that on directly, because it is the part most consultants will not say to a client’s face.

Consultancies feel it first because we have nowhere to hide. We sell judgement, not products. When a machine produces a passable first draft of that judgement in 40 minutes, the billable hour becomes very hard to defend.

Your HR function is next, and most HR Directors have not seen it coming.

Think about what your team actually produced last month. Policy drafts. Job descriptions. Grade recommendations. Manpower reports. Exit analysis. Onboarding packs. Almost all of it is knowledge work with a repeatable shape. If a consultancy can compress that, so can anyone, including a CFO with a subscription and an opinion about the size of your department.

There are two responses available to consultancies. Sell the same output faster at a lower price, which is a race to the bottom you will lose to someone with lower costs. Or turn the judgement itself into an asset that lives inside the business, so the output gets deeper instead of cheaper.

I will say the unpopular thing. Plenty of firms in this region are already doing the first one quietly. Same deliverable, same fee, a fraction of the effort, described in internal meetings as productivity. That works for a while. Then a client notices the deliverable was always generic, and the whole category gets repriced at once.

We went the other way. Here is what that took, and the order it happened in.

The same mistake, now with AI

The platform story I opened with is being repeated right now, at higher speed and lower cost of entry. Buy the AI license. Skip the basics. Wonder why nothing changed.

An AI tool with no institutional knowledge behind it behaves like a very confident new joiner on day one. It has read the entire internet. It has never seen your company. Ask it to write your HR policy and it will hand you something fluent, generic and completely detached from how your business actually runs, delivered with total confidence.

That is what most HR AI pilots produce. Then everyone concludes AI does not work for HR. The tool was never the problem.

What we actually built, in sequence

The fix is not a better tool. It is giving the tool your company, and there is no shortcut through that.

Consultex AI is built in-house, by our own engineers. The agent was the last thing we built. Everything before it was unglamorous and took far longer.

1. Knowledge harvesting. We sat with our senior consultants and got 16 years of judgement out of their heads. Not the deliverables. The reasoning. Why this job sits a grade above that one. Why a contracting client needs a different allowance structure than a retail client. Which client objection means “explain it again” and which means “you have the scope wrong”. This is the least glamorous work I have ever paid for and it is the entire foundation.
2. Data management. Our pay data was spread across surveys, project files and client submissions in every format a spreadsheet can take. We consolidated it onto Paylense. It now aggregates over 2,500 organizations across sources and industries. Cleaning that data took longer than building anything on top of it.
3. Knowledge management. Harvested judgement is useless if nobody can retrieve it. Everything is structured, versioned and governed, so an agent pulls the current logic rather than something a consultant wrote in 2021.
4. Playbooks. Judgement written down as explicit decision rules. This is the step that turns “our senior consultant knows” into something a system can apply the same way every time.
5. Automating our SOPs. Our own processes, workflow by workflow. Every gate, every review point, every approval.
6. Then the agents, with dashboards and analytics on top.

Five of those six steps have nothing to do with AI. They are knowledge work about our own expertise. That is the asset. The agent only sits on top of it.

More than 50% of our consulting services now run with AI. Of the 18 plus HR service types we deliver, at least 10 use it.

What changed, and what did not

Here is what the asset produces, measured in days.

An HR policy project that took 4 weeks now takes 5 working days.

Evaluating 100 jobs took 3 weeks. It now takes 3 working days, with a senior consultant deep review on every output.

I want to be precise about those old timelines. They were not slow work. They were the honest cost of doing the job properly by hand, and the clients who bought them got rigorous work that is still standing today. What changed is the cadence, not the standard.

I should also be straight about where the reclaimed time went, because the obvious assumption is that it came off the invoice. It did not. It went back into the work. The days a consultant used to spend grinding through evaluation now go into deeper senior review, more scenarios modelled and data refreshed continuously instead of once a year. Deeper, not cheaper.

The review gate is not decoration. Nothing reaches a client without a senior consultant signing it off. On our early attempts the agents produced confident output that was wrong in ways only an experienced consultant would catch. The confident new joiner is still there. He is just supervised now.

Some services resisted completely. Anything that depends on reading a room. Board level negotiation. Culture work. Those take exactly as long as they always did, and they should.

Where this lands on your P&L

Everything so far is what the asset did to our own delivery. Here is what it does for a client, and it lands on the largest controllable number you have.

Go back to step 2 of that sequence. Consolidating the pay data onto Paylense was the most tedious part of the entire build, and it is the one that changed what a client can actually decide. Payroll is usually the biggest controllable cost in the business, and it is set with the weakest data of any decision that size. Most companies fix it once a year, against a survey that closed months before the decision was taken. You are pricing today's offer against last year's data.

With benchmarking data running live across 2,500 plus organizations, two things become possible at the same time, which sounds contradictory until you see the distribution. You can pay properly where the market is genuinely tight, and hold the line everywhere else.

This is not done blindly. It is done with confidence from data. That is the difference between a cost decision and a guess.

Most companies do the opposite. They apply a flat percentage across the whole population. They overpay in roles nobody is competing for, underpay in the few roles that actually walk out, and then wonder why attrition and payroll cost both went up in the same year.

What separates two proposals that look identical on paper

All of this creates a new problem for you as a buyer.

Speed is now cheap. Two firms can build the same deliverable in completely different ways, and the document will not tell you which is which. A template and 16 years of harvested judgement look the same in a PDF.

So: two firms quote you a job evaluation for 100 roles. One is 40% cheaper. The methodology names sound the same. Ask these five questions and they will stop looking similar.

1. Whose judgement is inside this? Ask to see the factor logic and who wrote it. If the answer is a methodology name and nothing else, you are buying a template.
2. How old is the pay data, and when does it refresh? Annual snapshot or continuous? How many organizations sit behind the number, and across which industries?
3. Who signs each deliverable, by name? If nobody is named, nobody reviewed it.
4. What happens 6 months later, when the structure meets a real payroll cycle? A firm that has never run payroll for a client does not know what breaks.
5. Show me a case where your own recommendation failed and what you changed. Everybody has one. Only some will tell you.

Our answers are 16 years of harvested judgement, over 2,500 organizations live on Paylense, a named senior consultant on every deliverable, an outsourcing function that has to live with our advice through real payroll cycles and our own Technology Center in India with engineers across AI, ML, cybersecurity, product development, QA and infrastructure.

Most HR consultancies have no engineers. Most HRMS vendors have no consultants. Very few have the implementation muscle to find out whether the advice was any good in the first place.

Make every firm on your shortlist answer those five. Including us.

If you want to build this inside your own function

You may decide the asset should sit inside your own HR function rather than be bought. That is a legitimate answer, and for some companies it is the right one. The first steps are yours either way, and you do not need a consultant for them.

1. Pick one process you run repeatedly. Not your most important one. Your most repetitive one.
2. Write the SOP as it is actually done, not as the manual says it is done.
3. Get your best person to narrate the judgement calls out loud and capture the reasoning, not the output.
4. Fix the underlying data before you automate anything above it. If your job titles are inconsistent, nothing you build on top will work.
5. Keep a human sign-off gate, and name the person.
6. Measure elapsed days, not software licences.

Do steps 1 to 4 and never buy a single AI tool, and you will still be ahead of the company that bought the tool and skipped them. That is the whole difference between the clients who lost millions on a platform and the ones who did not.

What I still do not know

I opened by saying I cannot cleanly separate the war from the AI, and that is still true.

My best guess is that some of our older service offerings will largely vanish or become very occasional, and the rest will need upgrading to the new age.

I expect the same inside your HR function. A good part of the work that fills your team's month today will largely vanish or become very occasional, and the roles that remain will need upgrading to the new age. Which parts go and which stay will depend a lot on what you captured before the tools arrived.

I do not think anyone in this market honestly knows the split yet, and I would be careful of anyone who says they do.

What I am certain about is the sequence. The knowledge comes first. The agent comes last. Anyone selling you the reverse is selling you a licence.

If you are somewhere in the middle of this, I will be happy to compare notes. Tell me which process you picked.