Poland's Capability Centres in 2026: Hiring After the GBS Growth Slowdown

The numbers behind the slowdown

ABSL's 2026 report on Poland's business services sector puts employment at 500,500 people at the end of Q1 2026, spread across 2,179 service centres run by 1,303 companies. Growth was 1.8% year on year, about 8,860 jobs. The sector now accounts for 6.1% of Polish GDP and 7.8% of enterprise-sector employment.

That top line looks healthy. The composition underneath is the story.

Of those 8,860 new jobs, only 3,240 came from newly opened centres. Forty-six new centres opened in 2025 and four in the first quarter of 2026, the slowest pace of new openings in a decade. Poland is no longer winning the site-selection race it dominated from 2012 to 2021. It is growing by deepening what is already there, which is a completely different commercial and hiring problem.

Related reading: Bulgaria IT Outsourcing in 2026: A Buyer's Guide to Sofia's Software Sector · EU AI Act AI Literacy Training in 2026: What CEE Employers Must Deliver · Nearshoring Software Development to Poland in 2026.

What "qualitative growth" means when you are the one hiring

The industry phrase for this is a move from cost centre to capability hub, and it gets repeated at every conference in Kraków until it means nothing. Translated into hiring terms, it means three concrete shifts.

First, the work coming into Polish centres is higher up the value chain than it was five years ago: financial planning rather than accounts payable, product ownership rather than ticket triage, model governance rather than data entry. Second, the entry-level intake that used to absorb thousands of graduates a year has thinned, because that tier of work is the tier automation ate first. Third, the skills profile a centre needs has become far more specific, which makes each hire slower and each mis-hire more expensive.

You can see the second point in the ABSL data if you read it sideways. Employment grew 1.8% while the value of services delivered grew considerably faster. That gap is productivity, and productivity in a services centre means fewer people doing more, which nobody puts on a recruitment poster.

Here is the uncomfortable version for anyone running graduate intake in Poland: the model where you hired 200 economics graduates a year, put them on invoice processing, and promoted the good ones into analyst roles after eighteen months has stopped working. The bottom rung of that ladder has been automated away, and most centres have not built a replacement. They just recruit fewer graduates and hope the mid-level market supplies them, which it will not, because everyone is doing the same thing.

Kraków, Warsaw, and the tier-two cities

Kraków remains the largest single location, with close to 108,000 people across 312 centres. Warsaw, Wrocław, Tri-City and Katowice make up most of the rest, and the tier-two cities (Łódź, Poznań, Lublin, Rzeszów, Bydgoszcz) have been the quiet growth story.

The trade-off is not what most site-selection decks say it is. The usual pitch is that a tier-two city gives you 15% to 25% lower salary costs and lower attrition. The salary part is real and shrinking. The attrition part is real and durable, and it is the more valuable half.

What the decks understate is depth. In Kraków you can replace a senior treasury analyst in six weeks because eight other centres employ them. In Rzeszów you might be looking at four months and a relocation package. For a centre with 400 people and a broad skills mix, that risk is manageable. For a specialist unit of 40 where three people hold the domain knowledge, it is not.

Treat the salary index as directional. It moves with role family, and for AI and data engineering roles the spread between Kraków and a tier-two city narrows to almost nothing, because those people price against remote work in Western Europe rather than against the local market.

Ukraine, and the thing nobody puts in the deck

Poland's labour supply since 2022 has been shaped by Ukrainian migration in a way the official sector reports handle carefully and private conversations handle bluntly. A substantial share of the growth in Polish delivery capacity over the past four years came from Ukrainian professionals, particularly in engineering and Russian and Ukrainian language support.

Any workforce plan for a Polish centre that runs to 2030 has to hold two scenarios: one where a large part of that population stays and integrates permanently, and one where a meaningful share returns. Most plans I have seen assume the first and have no answer for the second. It is not a comfortable planning conversation and it is a necessary one, particularly for centres whose language coverage depends on it.

What this changes for vendor selection

If you are buying delivery capacity in Poland rather than building your own centre, the slowdown works in your favour on price and against you on availability of the good teams.

The questions that separate a serious Polish partner from a body shop have changed since 2021. Rate cards are no longer the interesting part. Ask these instead:

• What proportion of your delivery team on this account is dedicated versus shared, and what is your contractual notice on rotating a named person off?
• What is attrition on accounts of our size and profile, over the last eight quarters, not the last one?
• Which parts of this scope does your tooling now automate, and how is that reflected in the price rather than in your margin?
• Where is your bench, and what does the ramp look like if we double scope in six months?
• Who is your Article 4 owner, and can we see your AI literacy tiering? This is now a live procurement question for anyone delivering into EU enterprises.

The third question is the one vendors dislike. A partner who has automated first-pass reconciliation and is still billing the same FTE count for it is counting on you not asking. Some will give you a straight answer and a revised price. Those are the ones to shortlist.

Building a centre in Poland in 2026: still worth it?

For most companies, yes, with a narrower thesis than the one that drove the 2015 wave.

The old thesis was labour arbitrage at scale: move 500 transactional roles, save 60% on cost, accept some quality loss in year one. That thesis is mostly dead. Polish salaries in business services have risen for a decade, the transactional roles are being automated, and if pure cost is your driver there are cheaper markets with rising capability.

The thesis that still holds is capability concentration. Poland gives you a deep, EU-based, English-and-German-capable talent pool inside the same regulatory perimeter as your headquarters, at 40% to 55% of Western European cost, with the legal and data-residency simplicity of an EU member state. For anything touching GDPR-heavy data, financial regulation or the AI Act, that perimeter argument is worth more in 2026 than it was in 2019, and it is worth more than the remaining salary gap.

What I would not do is open a new centre for fewer than 150 seats in a tier-two city. The fixed cost of employer branding in a market where nobody has heard of you is brutal, and the ABSL numbers on new-centre openings suggest a lot of people have quietly reached the same conclusion.

The graduate pipeline problem, and who fixes it

Someone has to rebuild the bottom rung. The centres that are doing it well have stopped hiring graduates into process roles and started hiring them into what you might call assisted-analyst roles: from week one they work alongside automation, checking outputs, handling exceptions, and learning the domain through the exceptions rather than through the volume.

It is harder to design and it produces a better analyst in eighteen months than the old model did in three years. It also requires managers who can teach, which is a scarcer resource in Polish centres than any technical skill, and one that almost nobody is measuring for at promotion time.

If you run a centre in Poland and you are choosing one thing to fix this year, fix that. The salary benchmarking will sort itself out. The market will keep supplying senior engineers to whoever pays. What nobody can buy off the shelf is a management layer that can turn a smart 23-year-old into a domain specialist while the automation handles the volume, and in three years that will be the thing separating the centres that grew from the ones that flatlined at 1.8%.