BAT Just Cut 9,000 for "AI Transformation." The Omni Engineers Are Live.
British American Tobacco's June 29 cut hides a real Azure and MS Fabric engineering org. Here's how to source it before Accenture absorbs the second wave.
On June 29, 2026, British American Tobacco announced it would restructure 9,000 roles under a program called Fit2Win, framed publicly as an "AI transformation." Every tech recruiter I know scrolled past the headline because BAT sits under "Tobacco" on LinkedIn. That is the mistake. The cut is squeezing an Azure, MS Fabric, and GenAI org that most boolean strings will never surface.
Here is what actually got announced, who actually got cut, and how to source them before Accenture's 12-month clock starts running.
What the June 29 announcement actually said
The number is 9,000 role changes: roughly 5,500 direct layoffs and 3,500 roles transferred to strategic partners, excluding the US. CEO Tadeu Marroco described the goal as building a "more agile, cost disciplined and technology enabled" organization. Fit2Win, the umbrella program launched in 2025, targets around £600M in annual savings by 2028 and £500M by 2027.
The transfers go to named partners. Accenture picks up Global Service Hubs across Costa Rica, Mexico, Poland, Romania, and Malaysia, plus Supply Network Ops in the UK and Singapore. Systems Ltd. absorbs work in Pakistan. ITC Infotech takes Information, Digital, and Technology roles in Poland and Romania. A new BAT Future Capabilities Centre stands up in India.
Read that partner list twice. "Information, Digital, and Technology roles in Poland and Romania" is not admin. That is the digital engineering org. The 2025 Accenture partnership was explicitly announced to "simplify digital processes" and "improve speed to market." The layoff and the digital reorganization are the same event.
The tech org tech recruiters didn't know existed
BAT is coded as a CPG company. Under the hood it runs like a mid-sized enterprise SaaS shop that happens to sell nicotine.
From the BAT HY 2025 investor transcript and public job specs, the real footprint:
- 85% cloud hosting, with roughly 2,000 workloads migrated to Azure.
- Microsoft Fabric as the enterprise data platform, under a direct Microsoft partnership.
- A GenAI lab launched in Dubai in 2025, shipping "Ask Omni," AI sustainability bots, and consumer insight tools.
- The Omni digital stack activated in 13 markets in H1 2025.
- Digital hubs in Warsaw (BAT DBS Poland), Bucharest, Prague, Budapest, London Chiswick, Hamburg, and Monterrey.
- A Warsaw Supply Network Operations hub running AI/ML/automation for over 100 countries.
- Skill stack from current job specs: Python, SQL, Azure, MS Fabric, T-SQL, Power BI, RAG/LLM prompt engineering, dbt-adjacent modeling.
Fujitsu handled the cloud migration ops. BAT was an early Udacity enterprise customer, running data engineering and product management upskilling since 2020. Jean-Pierre Cussac, BAT's global head of IT enterprise services, went on record in I-CIO saying BAT was "shifting to become more like a technology company." Martin Whitehouse ran the Future Capability Development function that built the internal engineering bench. Janusz Dziurzyński is on public record as Global Delivery Digital Business Solution Lead and Warsaw Hub Lead.
None of this shows up when you filter LinkedIn Recruiter by "Tobacco" and exclude it, which most tech recruiters do by default.
The industry-tag bug
Here is the sharp version of the problem. Filter for Data Engineer, Software Engineer, MLE, or Platform Engineer inside Tobacco, Consumer Goods, and F&B across the UK, Poland, and Romania. You get roughly 26 profiles. The results are dominated by KFC UK, Domino's, Kitopi, and Huel.
That number is obviously wrong. BAT's DBS Poland office in Warsaw employs hundreds. Bucharest is substantial. Chiswick has a real platform team. The tools are not broken, the taxonomy is. Every affected engineer self-tags "British American Tobacco," LinkedIn buckets that under "Tobacco," and every tech recruiter's saved search excludes "Tobacco" as noise.
This is the exact failure mode we designed Refolk around. You describe the person in plain English ("data engineers in Warsaw with Azure and MS Fabric, currently at BAT or recently exited") and get a ranked shortlist that ignores the LinkedIn industry field entirely. Industry tags are a sourcing artifact, not a talent signal.
Why "AI transformation" is the wrong mental model
The public narrative is that AI replaced humans. The mechanics say something else. A Microsoft, Accenture, and ITC Infotech-led simplification means the affected staff include the digital and data people who built the AI stack in the first place. Ask Omni did not fire the engineer who wrote the RAG pipeline. Accenture's contract did.
That is the arbitrage. The layoff description repels the exact recruiters who should be running toward it.
The layoff description repels the exact recruiters who should be running toward it. That is the arbitrage.
The affected roles map cleanly to hot 2026 stacks. Azure plus MS Fabric plus dbt plus Python plus RAG and LLM prompt engineering is the same profile a Series B AI startup is paying a 30 to 50% premium for. A BAT data engineer in Warsaw is a Snowflake and dbt data engineer in Warsaw with an industry label problem. Fix the label, close the offer.
The two waves: June 29 and Accenture plus 12 months
Sourcing engineers from non-tech layoffs is a timing game. Two waves matter.
Wave one: the direct 5,500
Consultations for affected roles are ongoing under local labor laws, with all changes to finalize by year-end 2026. That is a rolling six-month exit window, not a single WARN date. Direct layoffs from digital hubs will start surfacing on LinkedIn with vague "open to work" banners through Q3 and Q4 2026. Most will not say "British American Tobacco layoffs 2026" in their headline. They will say "exploring next chapter."
Wave two: the 3,500 transferred
Engineers TUPE'd or novated into Accenture, Systems Ltd., or ITC Infotech usually leave within 12 to 18 months. Comp bands compress. The career track collapses from "product engineer at BAT" to "delivery consultant on a client engagement." Recruiters who ignore the June 29 headline and instead source from Accenture Poland, Accenture Romania, and ITC Infotech's UK bench starting in Q3 2027 will win the second wave. But only if they know which internal team names to filter for: Omni, Ask Omni, BAT DBS, Fit2Win, MS Fabric at BAT, RGM.
The un-crowded search terms
The best signal is not "laid off." Every recruiter searches that. The un-crowded searches are internal program and product names that only insiders and their immediate network know:
- "Omni" plus BAT (the digital stack activated in 13 markets)
- "Ask Omni" (the GenAI assistant from the Dubai lab)
- "Fit2Win" (the restructuring program itself)
- "BAT DBS" or "DBS Poland" (the Warsaw digital hub)
- "SNO" plus Warsaw (Supply Network Operations, running AI/ML for 100+ countries)
- "MS Fabric" plus tobacco or CPG (the enterprise data platform)
- "RGM" plus BAT (revenue growth management, a heavy analytics function)
- "Future Capabilities Centre" (the new India build)
Paste any of those into a normal boolean tool and you get thin, noisy results because the tool wants exact string matches on job titles. Paste them into a plain-English sourcing query in Refolk and the semantic layer treats them as team and product signals, then ranks profiles by proximity to the actual stack. This is where hidden engineering talent in CPG stops being hidden.
The five offices to source from, ranked
If you only have time to work five geographies, work these in order:
- Warsaw (BAT DBS Poland and SNO Hub). The largest concentration of Azure, Fabric, and Python engineering. Also the epicenter of the Accenture Poland and ITC Infotech Poland transfers, meaning wave two lands here hardest.
- Bucharest. ITC Infotech's Romania transfer target. Strong data engineering pool, thin local competition from tech recruiters.
- London Chiswick. Senior platform, architecture, and product engineering. Highest comp bands, most likely to be hired by UK fintech and AI startups within 90 days.
- Monterrey. Accenture Mexico hub target. Underrated data analytics and platform bench.
- Dubai GenAI Lab. Small headcount but the RAG and LLM prompt engineering talent is disproportionately valuable.
The wider pattern: this is not a one-off
AI transformation layoffs from non-tech companies are becoming the most underpriced talent pool of 2026. In the same quarter, Gambling.com cut 25% for an "AI-first world." Takeda cut around 4,500. Volkswagen has flagged up to 100,000. In every case, digital and data staff sit inside larger workforce reductions and disappear from tech-recruiter searches because the parent company is coded pharma, auto, or CPG.
Sourcing outside the tech industry means treating the industry field as noise. It means learning which non-tech companies actually run modern stacks: Unilever's Foundry, Nissan Digital Hub, Reckitt data science, Diageo's Baileys AI team, Maersk's platform org. Every one of those is an engineering shop dressed as something else.
Communities where displaced BAT engineers will surface first: the Warsaw Azure User Group, the Microsoft Fabric Community, dbt Slack (#job-postings and #local-warsaw), the London Data Engineering meetup, and the Accenture Poland alumni network for the 12-month second-wave play. Set a calendar reminder for June 2027 on that last one.
What to do this week
Three concrete moves.
First, build a saved query for BAT DBS Poland, Bucharest, and Chiswick engineers by team name, not industry. Ignore the "Tobacco" field entirely. If your tool forces you to pick an industry filter, your tool is the problem.
Second, write a first-touch message that names the stack, not the company. "Saw you shipped on Ask Omni and MS Fabric" beats "noticed you're at British American Tobacco" every time. The former reads as respect for the work. The latter reads as a layoff scraper.
Third, put a Q3 2027 reminder on your calendar to source Accenture Poland, Accenture Romania, and ITC Infotech UK for ex-BAT engineers. That is when comp compression bites and the second wave opens up. Recruiters who work the calendar, not the news cycle, win here.
The 400-plus engineers in the rubble of the June 29 announcement are not gone. They are just under the wrong label.
FAQ
How do I identify BAT engineers if they don't announce the layoff publicly?
Most affected engineers will not post "laid off from British American Tobacco." They will quietly update their headline, drop the company from their current role, or add "open to work." The reliable signals are internal program names (Omni, Ask Omni, Fit2Win, BAT DBS, SNO) that appear in their experience descriptions, plus stack keywords like Azure, MS Fabric, and dbt. Semantic search across LinkedIn, GitHub, and the open web catches these where boolean strings on job titles miss them.
Isn't sourcing from a tobacco company a hiring manager problem?
Sometimes, rarely. In practice, hiring managers care about the stack and the shipped product. "Built the RAG pipeline for Ask Omni across 13 markets on Azure and MS Fabric" is a stronger resume line than most fintech projects. If your hiring manager reflexively rejects CPG or tobacco backgrounds, that is a calibration conversation, not a sourcing constraint. The underlying engineering is real.
Why focus on Accenture 12 months out instead of hiring direct layoffs now?
Do both. Direct layoffs from the 5,500 are available now through year-end 2026 and will move fast, especially in Chiswick and Warsaw. The Accenture, Systems Ltd., and ITC Infotech transfers are a second wave that opens in Q3 2027 when comp compression and career-track collapse hit. Recruiters who only work the June 29 headline will miss the larger, less contested pool 12 months later.
What makes non-tech "AI transformation" layoffs different from regular tech layoffs?
Two things. First, the affected engineers self-tag under a non-tech industry, so tech-recruiter searches actively exclude them. Second, the "AI transformation" framing makes the layoff sound like AI replacing humans, which repels sourcers who assume the cut targets low-skill roles. Both effects compound: the pool is real, the stack is modern, and almost nobody is looking. That is the definition of an underpriced market.