UKRI Enabling AI Adoption Across Science and Engineering is a £5 million fund offering up to £250,000 (80% funded) for 12-month exploratory AI projects in 5 priority areas, with a closing date of 1 December 2026 at 4:00 pm UK time.
For research teams, the challenge is making sure their data is ready to support the research question within that 12-month window. Datavid, a C5i company, helps organisations connect, structure and prepare data for AI applications, so teams can focus on demonstrating what AI can achieve.
An AI expert and a domain expert may have a research question worth exploring, but can the data behind it support meaningful results within a year? Finding datasets, connecting information across systems and establishing shared meaning can consume valuable project time.
With the UKRI project start date fixed at 26 April 2027, assessing data readiness early can help project leads and heads of bioinformatics identify potential gaps before implementation begins. Datavid can help teams build the connected, AI-ready data foundations needed to support exploratory projects through its AI services .
The call funds exploratory projects that pair an AI expert with a non-AI domain expert around a research challenge in 1 of 5 priority areas. Among AI grants UK institutes can apply for, it is unusual to count data requirements as a valid output, so the data work behind the bid is fundable rather than overhead.
The UKRI Enabling AI Adoption Across Science and Engineering call, led by EPSRC (official call page), supports exploratory AI projects that address research challenges across five priority areas. For research teams, the challenge is not just identifying a promising AI use case, but ensuring the data needed to explore it is accessible, connected and meaningful within the project's 12-month timeframe.
With projects scheduled to start on 26 April 2027, teams can use the preparation period to identify data gaps, clarify access requirements and establish what needs to be in place before experimentation begins. Datavid, a C5i company, supports this work through data integration, semantic modelling and knowledge graph solutions that help research teams build connected, AI-ready data foundations.
Projects primarily focused on AI methods are out of scope. Expected outputs range from feasibility evidence and prototypes to data requirements and research roadmaps, so funding under UKRI Enabling AI Adoption Across Science and Engineering sits closer to a UKRI proof of concept than to a methods program.
BridgeAI deployment catalyst: an Innovate UK competition for SME-led AI assurance and compliance, closing 8 December 2026.The UK research organisation applies for and leads the project, while other organisations may participate as project partners where they have an integral role in the work. Each named project partner must provide its own letter or email of support explaining its contribution to the project, limited to 2 A4 pages. Applicants should confirm partner eligibility and any proposed financial arrangements with their research office.
Applications receive up to 8 reviews through distributed peer review, with no panel and no applicant response. Reviews cover 5 assessment areas plus the 1,500-word data management and sharing section, and reviewers weigh whether the AI fits the domain rather than AI novelty alone.
|
Assessment area |
What the call asks for |
Where data decides it |
|
Vision and approach |
The challenge, the AI method, the workplan |
Can the AI reach the data needed to test the question |
|
Applicant and team capability (1,650 words) |
Who delivers, track record |
The team's delivery record, including any partners |
|
Resources and cost justification (1,000 words) |
Staff, partners, compute, data costs |
Data management and sharing requirements should be considered when planning project resources |
|
Ethics and responsible research and innovation (RRI) |
Data collection, release, storage; information standards |
Which standards and access rules apply |
|
Risk and reward (1,000 words) |
What could stall it, why it is worth doing |
Data access and quality tend to stall 12-month projects |
A reviewer with 8 applications and no panel discussion is likely to reward a proposal whose data is already in reach and whose delivery plan is clear. Showing that control is achieved through data governance services, and a documented data management plan addresses 3 of the 5 areas, with Resources being where partner costs are justified.
In a 1-year exploratory project, the AI method is rarely the bottleneck. Finding the data, agreeing what its fields mean, connecting it across systems and proving where each value came from is where the months tend to go. In AI for science, AI readiness is largely a property of the data rather than the model.
Most 12-month projects that run late lose time at 1 of 3 points, and each is visible to a reviewer before the award.
The UK Government's AI Opportunities Action Plan (January 2025) and the Open Data Institute's 2025 white paper on an AI-ready National Data Library both prioritize documented, interoperable research data over the model.
A project that begins with FAIR data and a connected foundation can spend its year on the science. That foundation can come from the institute's own team or from a data partner such as Datavid through research data management work, so the domain and AI experts are not also the data engineers.
The 4 foundations below are in the order a project needs them. Each is something the application can describe, and a named partner can commit to in a letter of support.
These foundations are not just technical considerations; they help research teams identify data gaps, define realistic project requirements, and prepare for AI experimentation.
A documented inventory is a catalog of the datasets the project depends on, where they reside, who owns them, and the access conditions. These are FAIR's first 2 letters and a quick way to show a reviewer that the data is within reach, which is why they belong in the data management section and the Resources justification.
The Syngenta research platform shows what that looks like at scale, where decades of siloed R&D documents can be searched in minutes rather than weeks.
Shared meaning comes from an agreed model of the entities and terms the project uses, linked to controlled vocabularies. AI prompted on data that means different things across systems tends to produce results that are hard to defend, and reviewers look for this in the information standards question in Ethics and RRI, and in the Vision section.
With ontology management in place, a follow-on project reuses the same definitions, so the work keeps paying off.
Connected data means a knowledge graph linking the inventoried datasets through the ontology, so the AI retrieves entities and relationships rather than files, which is the core of knowledge graph solutions. Many domain questions span more than 1 source, and data silos can cost months, so this is what the AI approach and its feasibility evidence rest on.
Unifying Biobank is the evidence, with a governed metadata knowledge graph and a RAG workflow delivered across 2 biobanks in 8 weeks, aligned to OMOP 5.4.
Provenance means origin, version, and transformation recorded for each data point the AI uses, which a sound data architecture captures by default, so results trace back to source. It helps make feasibility evidence credible to a reviewer and reusable by a follow-on bid, and it speaks directly to the data management and sharing section and to Risk and reward.
Roche Helios shows the payoff, where study forms mapped into a common semantic model delivered 5x faster clinical trial data processing, 80% fewer manual errors, 40% lower operational costs and on-demand audit readiness for FDA and EMA.
The section must demonstrate how the project will comply with the UKRI data-sharing policy, which expects data to be as open as possible and as restricted as necessary. It also has to be specific to the project's datasets, since a generic data management plan tends to lose marks, and the access rules are where data compliance experience saves time.
Reviewers tend to read the section as 5 questions, and the strongest answers name the project's own datasets rather than restating policy.
The 4 foundations above answer these 5 elements almost line for line. Research teams that have these foundations in place, or have identified a partner to help build them through enterprise data management, can develop this section using concrete details about their data, rather than general intentions. This also supports the broader goal of keeping research data reusable.
The gap between the 1 December deadline and the 26 April 2027 start is about 4.5 months. The sequence below is what a partner does alongside the institute in that window.
Each step builds on the previous one, and the first step can start before the bid is submitted.
Datavid Rover can support the development of a connected semantic data foundation. The scope and delivery timeline will depend on the project's data sources, access requirements, and technical complexity.
The foundation and the partner add the most when the research question spans sources, formats, or institutions, as many life sciences data questions do.
Even when data already sits in a well-governed system, ontology and provenance layers can help turn a prototype into reusable evidence of feasibility. A named partner's contribution can also be described in the team capability section. Project teams can start with a data inventory and use the findings to determine what additional support or integration work may be needed.
The call asks what AI could do for a domain, and the honest answer depends on the data it can access. Show the reviewers that the data is ready and who will deliver the AI, and the case tends to write itself.
Preparing a UKRI AI adoption proposal? Assess whether your research data is ready to support your project goals.