In June 2026, the Medicines and Healthcare products Regulatory Agency launched a new AI sandbox to test how artificial intelligence could support medicines development. The programme will look at models that could help predict how medicines behave in the body, including how they are absorbed, processed and whether they could cause harm.
The move highlights how the conversation in life sciences is now closer to evidence rather than early promise. AI tools are now being tested in environments where regulators, researchers and innovators need to understand how they perform, what evidence is needed and where the limits of their use should be drawn.
When considering their talent strategies, life sciences organisations still need people who can build and apply AI tools, but the harder work is increasingly around data quality, validation, model oversight, documentation and regulatory confidence. A broad search for “AI talent” will only go so far when the work is moving into regulated settings.
AI is moving closer to regulated work
Many life sciences organisations have already tested AI in areas such as research, data analysis, automation and drug discovery, providing teams with a way to explore the technology without immediately changing the systems or decisions that carry the greatest regulatory risk.
However, the less stage is next forgiving; once an AI model is expected to support safety assessments, development decisions or work involving clinical and research data, the organisation has to understand far more than whether the tool appears to work. It needs to know how the model was trained, what data was used, where the output is reliable and how the evidence would stand up to review.
The MHRA sandbox reflects this shift, as it gives companies and researchers a controlled environment to work through those questions with regulators, rather than treating regulation as something to address after the technology has already been built.
This is where AI hiring becomes more specific. A technically strong model still needs people around it who understand evidence, risk, patient safety and the expectations of a regulated sector. Without that mix, a project can look promising in an internal pilot and still struggle when it moves towards formal use.
Data quality is no longer a technical side issue
AI depends on the data behind it, and life sciences data is rarely straightforward. Research data, clinical trial data, real-world evidence and laboratory information can all carry issues around completeness, consistency, privacy, bias and suitability for a particular use.
If the data is weak, inconsistent or poorly governed, the AI tool is already compromised. The problem may not sit in the algorithm at all; it may sit in the way information has been collected, labelled, structured or maintained over time.
As such, life sciences organisations need people who can work with complex datasets, understand governance requirements and recognise when data is not strong enough to support the purpose being proposed. That experience may come from data engineering, bioinformatics, clinical data, statistics, software or quality backgrounds, rather than from someone with an obvious AI job title.
The Life Sciences Jobs Plan points in the same direction, with a growing need for people who can combine scientific understanding with digital, AI, regulatory and quality skills. That combination is difficult to find, especially when the same candidates are also attractive to technology companies, consultancies and other regulated sectors.
Job descriptions are struggling to keep up
As AI moves closer to regulated delivery, the roles around it become harder to define. One vacancy can quickly accumulate requirements from data science, software engineering, validation, quality, regulatory affairs and governance, so the brief becomes more impressive on paper and less realistic in the market.
Life sciences employers often need people who can work across more than one discipline, but there is a difference between a hybrid role and an overloaded one. A company may need deep technical AI expertise for model development, while another role may be more focused on documentation, validation or working with regulatory and quality teams. Combining all of those needs into one search can narrow the candidate pool before the process has properly begun.
A clearer brief starts with the work. Is the role building models, preparing data, reviewing outputs, managing validation, improving governance or helping scientific and regulatory teams understand how AI is being used? Some of those responsibilities can sit together. Others need separate expertise or a team built around the project.
Candidates will also judge the opportunity through that level of clarity. Strong AI and data professionals are unlikely to be persuaded by broad claims about innovation if the role itself is vague. Instead, they will want to understand the systems, the data, the project stage and the level of responsibility attached to the work.
Regulation is shaping the skills market
The FDA and European Medicines Agency have set out guiding principles for good AI practice in drug development, covering areas such as context of use, multidisciplinary expertise, data governance, documentation, model design, performance assessment and lifecycle management.
If an AI model is being used in drug development, the organisation needs people who can explain how it will be used, how its performance will be assessed and how it will be monitored after deployment. Documentation cannot be pulled together at the end by someone who was never close to the technical work.
As a result, team structure may be affected: technical specialists need to work with people who understand quality systems, regulatory expectations and scientific context. Regulatory and quality teams may also need stronger technical knowledge, particularly where they are expected to review AI-enabled processes or challenge supplier claims.
The skills needed around AI are becoming wider, but they are tied to specific work: data governance, validation planning, model review, documentation, lifecycle monitoring and the ability to explain technical decisions in language that scientific, regulatory and commercial teams can use.
External expertise still needs internal knowledge
Vendors, consultancies and specialist partners will continue to support AI adoption in life sciences. Many organisations will not have every skill they need in-house, particularly when projects involve new tools, specialist modelling or regulatory uncertainty.
However, internal teams still need enough knowledge to assess proposals, challenge assumptions and review whether the output fits the organisation’s scientific and regulatory responsibilities. A supplier can build or configure a tool, but the organisation still owns the decision to use it.
This becomes more important as AI moves into areas that affect development, safety, operations or patient-facing work, as teams need to know which data has been used, how the model has been tested, what limitations have been identified and who is responsible for monitoring performance once the tool is in use.
Life sciences organisations do not need every AI specialist to sit permanently inside the business, but they do need enough internal capability to govern the work properly and make informed decisions about where external support is being used.
What employers need to think about now
Hiring for AI in life sciences needs more precision than many role briefs currently allow. Employers need to decide whether they are looking for model development, data engineering, validation, governance, regulatory understanding or someone who can connect several of those areas.
The brief should show candidates the stage of the work. A role supporting early experimentation will not look the same as one preparing an AI tool for regulated use. The systems, data, documentation requirements and decision-making responsibilities should be clear enough for candidates to understand what they are actually being asked to do.
Employers should also look carefully at adjacent experience. People from clinical systems, data management, software, quality, cybersecurity or regulatory roles may already have skills that are highly relevant to AI delivery, even if they have not held a dedicated AI role before.
Training will matter too, as some of the skills needed for regulated AI delivery are still emerging, and employers may need to build capability around the people they already have. For example, a data specialist may need more exposure to regulatory expectations, or a quality professional may need a better understanding of model behaviour and lifecycle monitoring.
Moving from pilots to proof
The sector is moving towards a stage where AI tools need to be tested, evidenced, documented and governed with the same seriousness as the systems and decisions they are intended to support.
As a result, life sciences organisations need people who understand technology, data and the regulated environment around them. At nufuture, we work with life sciences organisations to recruit specialist technology talent across AI, data, cybersecurity, cloud, infrastructure and the wider IT function. If you are reviewing the skills your organisation needs to move AI projects forward, get in touch.