Data Science Jobs and AI in the UK (2026): Will AutoML and LLMs Automate Data Science?

For a decade, data science was the job that automation was supposed to create, not destroy. That framing no longer quite holds. By 2026, the same machine-learning techniques data scientists pioneered are being turned on the data science workflow itself: AutoML systems that build and tune models without much human input, and large language models that write the SQL, draft the analysis and explain the results. If your job is to find signal in data, the obvious question is whether the tools can now do it without you.

This article is a 2026 data-led cut, focused on what AutoML, LLM-assisted analysis and agentic AI changed in 2025 and 2026 for UK data science jobs, using current hiring figures, named employers and salary benchmarks. It is deliberately not a generic "impact of AI on data science" explainer.

The Short Answer

AutoML and LLMs are automating large chunks of the data science workflow, but the evidence from 2025-2026 points to reshaping rather than wholesale replacement. According to ITJobsWatch data, the median UK data scientist salary sat at £65,000 in the six months to 17 January 2026, up roughly 8.33% year on year, which is not the curve of a collapsing field. Demand softened in places: ITJobsWatch recorded permanent data scientist postings falling from 254 to 184 over the year to January 2026. Yet Adzuna listed around 5,110 data scientist vacancies across the UK in January 2026, with mean data-role pay around £50,412, up 5.8%. The pattern is clear: routine modelling, cleaning and reporting are being absorbed by AutoML and LLM copilots, while demand grows for data scientists who can frame problems, judge results and own projects.

Will AI replace data scientists in 2026?

On current UK evidence, the honest answer is "not wholesale, but the job is being rewritten". The headline macro figure comes from the IPPR, whose task-level analysis of around 22,000 tasks found roughly 11% exposed to current generative AI, rising to as much as 59% with deeper integration. Its worst-case scenario put up to 7.9 million UK roles at risk if firms displace exposed tasks with no offsetting job creation, but its central scenario was far lower, at around 545,000 jobs lost alongside a potential 3.1% GDP uplift. The distance between those numbers is essentially an adoption and policy choice.

Data science specifically sits awkwardly in this picture. Much of the work is exactly what LLMs and AutoML do well, namely well-defined coding, model selection and tuning. But the accountable parts, deciding which question matters, whether a result is trustworthy, and how a model behaves in the real world, remain stubbornly human. The Office for National Statistics reported that 23% of UK businesses used some form of AI by late September 2025, up from 9% in September 2023, yet only 4% of AI-using firms reported a headcount fall as a result. In our reading, that is augmentation outrunning displacement, at least for now. The risk is sharper at the junior end, where routine analysis once provided a training ground that AutoML and copilots are quietly eroding.

Which data science tasks is AutoML automating?

AutoML has moved well beyond hyperparameter tuning. The frontier in 2025 and 2026 is agentic AutoML, where multi-agent LLM frameworks attempt the full pipeline, from data retrieval and preprocessing to model design and deployment-ready output, driven by a plain-English task description. Research systems described in 2025 and 2026 decompose a brief into sub-tasks handled by specialised agents, profile the data, propose hypotheses, write code and iterate toward a result. The practical effect is that the mechanical middle of the workflow is increasingly something a data scientist supervises rather than performs.

Here is how the split typically looks in 2026.

The consistent pattern is that AI is strongest on repeatable, well-specified sub-tasks and weakest on ambiguous, accountable judgement. That is why, despite the hype, several 2026 analyses note that AI-tool usage barely appears in UK job adverts, possibly because employers now treat copilot fluency as too basic to list, much as Excel once became assumed rather than advertised.

How have LLMs changed the data scientist's day in 2025-2026?

The most visible shift is the copilot. Tools such as GitHub Copilot for code, and analysis copilots embedded in platforms like Microsoft Fabric, now sit inside the daily workflow, drafting queries, explaining unfamiliar code and generating first-pass charts. Practitioner accounts through 2025 describe meaningful time savings on literature review, boilerplate and database work, in the order of a fifth of effort on routine tasks, while stressing that the productive pattern is treating the LLM as a copilot, not an autopilot.

The skill mix is shifting accordingly. Prompting, retrieval-augmented generation, understanding context windows and knowing when to fine-tune versus retrieve are becoming as ordinary as knowing pandas or scikit-learn once was. At the same time the data engineering bar has risen: SQL, ETL pipelines and tools such as Snowflake and dbt feature more prominently, pushing the role toward a generalist "data" profile expected to own projects end to end. The net effect is not less work but different work, with hours freed from writing baseline models reinvested in framing the right question and checking whether an LLM-generated analysis is actually correct.

Which data science roles are growing in the UK?

Demand has tilted from pure model-building toward deployment, reliability and judgement. As organisations move from analytics pilots into production, the bottleneck has shifted downstream, which favours roles that AutoML cannot easily own.

Roles expanding, in our reading of current UK demand, include:

• Machine learning engineers and MLOps specialists, who put models into production and keep them monitored and retrained.

• Analytics engineers, bridging data engineering and analysis with SQL, dbt and warehouse tooling.

• Senior and lead data scientists, who frame problems, validate AutoML output and own outcomes.

• AI and LLM application specialists, building retrieval and agent workflows on top of foundation models.

• Data scientists with strong domain expertise, in finance, healthcare or retail, where context cannot be automated.

The salary signal supports the seniority shift. According to ITJobsWatch data, the median data scientist salary was £65,000 in the period to 17 January 2026, with London typically at £70,000 to £80,000, and the median Head of Data Science salary around £150,000. Adzuna put mean data-role pay at £50,412 from October 2025 to January 2026, up 5.8% year on year. These are not the salary curves of a field being automated out of existence; they are the curves of a field rewarding the parts that are hardest to automate.

Which data science jobs are most exposed to automation?

Not every role is comfortably placed. The squeeze is sharpest where the work is routine and entry-level, precisely the tasks AutoML and copilots do best.

The concern flagged by the IPPR and echoed by UK skills bodies is structural rather than total: if AutoML absorbs the entry-level rung, newcomers lose the training ground that produced senior data scientists in the first place. Several 2026 commentaries also note that breaking into data science is harder than in 2019, though arguably easier than in the 2023 hiring trough. Hedged honestly, these roles are not disappearing overnight, but headcount per project is typically falling, and the bar for a first job is rising.

Where are the data science jobs, and who is hiring?

The UK map is concentrated but not London-only. Adzuna data for October 2025 to January 2026 put London well ahead with around 1,896 unique data-role postings, followed by the South East at roughly 542 and the North West at about 444, with Manchester and Edinburgh functioning as genuine secondary hubs.

On named employers, the spread is broad. Tesco runs a substantial data science function and recruits scientists tasked with promoting data science across the business and representing the company in the wider community. HSBC, AstraZeneca and BT recruit data scientists across London, Manchester and Edinburgh, spanning finance, life sciences and telecoms. In the public sector, NHS trusts including Imperial College Healthcare NHS Trust hire data scientists for clinical analytics and operational modelling. The BBC has built data science teams around recommendation and audience analytics. This breadth, retail, banking, pharma, telecoms, healthcare and media, is one reason the UK data science labour market has stayed resilient even as individual tasks are automated. It is not dependent on any single sector or on the capital alone.

What do UK bodies and regulators say about data science and AI?

The institutional backdrop matters because it shapes both standards and demand. The Alan Turing Institute, the UK's national institute for data science and AI, has worked through 2025 with Skills England and partners to align apprenticeships and technical education with the skills employers now need, and has examined recruitment data to spot supply-and-demand mismatches. Its work points consistently toward augmentation, with the emphasis on equipping workers to use AI responsibly rather than predicting their replacement.

Professional standards are firming up too. The Royal Statistical Society, through its Data Science and AI Section and as a founding member of the Alliance for Data Science Professionals, offers the Advanced Data Science Professional certificate and has pursued chartered status for the profession. That push toward accreditation is itself a signal: as AutoML commoditises the mechanical work, demonstrable judgement, statistical rigour and accountability become the differentiators, and the profession is moving to certify exactly those. For an individual, the takeaway is that recognised statistical and governance skills are becoming more valuable, not less, precisely because they are the parts a model cannot be held responsible for.

How should you future-proof a data science career in 2026?

The defensible move, on current UK evidence, is to climb from tasks to judgement and from analysis to ownership. First, treat AutoML and LLM copilots as collaborators to be supervised, not rivals to be out-typed; the value is in directing and validating them. Second, deepen the parts machines struggle with: causal reasoning, statistical rigour, problem framing and clear communication with decision-makers. Third, build the deployment side, MLOps, monitoring and analytics engineering, where production demand is concentrated. Fourth, develop genuine domain expertise, because context in finance, healthcare or retail is hard to automate. The pattern across 2025 and 2026 UK sources is consistent: the mechanical core of data science is being absorbed, and the premium is shifting to the judgement wrapped around it.

Frequently Asked Questions: Data Science Jobs and AI

Will AutoML replace data scientists in the UK?

Not wholesale, on current evidence. AutoML automates model selection and tuning well, and agentic frameworks now attempt full pipelines, but the accountable work of framing problems and validating results stays human. ITJobsWatch put the median UK data scientist salary at £65,000 to January 2026, up around 8.33% year on year, which suggests reshaping rather than replacement.

Are data science jobs still in demand in 2026?

Yes, though the picture is mixed. ITJobsWatch recorded permanent data scientist postings falling from 254 to 184 over the year to January 2026, but Adzuna listed roughly 5,110 UK data scientist vacancies in January 2026. Demand has tilted toward senior, engineering and domain-heavy roles rather than entry-level analysis.

How much do data scientists earn in the UK?

According to ITJobsWatch data to 17 January 2026, the median UK data scientist salary was £65,000, with London typically £70,000 to £80,000 and Head of Data Science around £150,000. Adzuna put mean data-role pay at £50,412 from October 2025 to January 2026, up 5.8% year on year, comfortably above the national average.

Which data science tasks are being automated first?

Routine, well-defined work goes first: model selection and tuning via AutoML, first-draft SQL and exploratory analysis via LLM copilots, and standard reporting and dashboards. Agentic tools increasingly profile data and propose models automatically. Judgement-heavy tasks, problem framing, validation and stakeholder translation, remain far harder to automate.

Which data science roles are growing in the UK?

Demand is tilting toward machine learning and MLOps engineers, analytics engineers, senior and lead data scientists, LLM application specialists, and domain experts in finance, healthcare and retail. These roles centre on deployment, reliability and judgement, which AutoML cannot easily own, and they command the strongest pay in current UK salary data.

Which UK companies hire data scientists?

Major employers include Tesco, HSBC, AstraZeneca and BT, recruiting across London, Manchester and Edinburgh, alongside NHS trusts such as Imperial College Healthcare and media organisations including the BBC. Roles span retail, banking, pharma, telecoms, healthcare and media, and are typically advertised via boards such as Adzuna and specialist job sites.

Do I still need to learn statistics if LLMs can do the analysis?

Yes, arguably more than before. The Royal Statistical Society and Alan Turing Institute both emphasise rigour and accountability, and the RSS has pursued chartered status for the profession. As LLMs generate plausible-looking analysis quickly, the ability to judge whether a result is actually correct, and to stand behind it, becomes the core differentiator.

Summary: AI Is Rewriting Data Science Jobs, Not Ending Them

AutoML and LLMs have automated a real share of the data science workflow, from model tuning to first-draft analysis, and agentic systems are pushing further into full pipelines. Yet UK evidence from 2025 and 2026 points to reshaping rather than replacement: ITJobsWatch shows a median data scientist salary of £65,000, Adzuna lists thousands of vacancies, and the ONS finds AI adoption rising without much headcount loss. Demand is shifting toward MLOps, analytics engineering, senior judgement and domain expertise, supported by employers from Tesco to the NHS and by bodies such as the Royal Statistical Society and the Alan Turing Institute. The strategic move for 2026 is to direct and validate the tools, not compete with them on raw output.

Ready to find your next role in a fast-changing market? Browse current vacancies and career advice at datascience-jobs.co.uk.