Research knowledge infrastructure: a landscape review for Russell Group universities
The research tool ecosystem spans roughly 65 platforms across five capability segments, but only a fraction are procurement-ready for UK universities today. Three Jisc Dynamic Purchasing Systems now cover CRIS/repositories, electronic lab notebooks, and core identifier infrastructure — meaning compliant procurement routes exist for the most critical layers. However, significant gaps remain in AI-integrated platforms and collaborative knowledge graphs, where no consortium agreements exist and institutions must navigate a rapidly evolving vendor landscape dominated by US-headquartered SaaS providers. This report maps the entire ecosystem, identifies Jisc-procurable options, flags GDPR risks, and recommends practical adoption paths for each segment.
The timing is consequential. UKRI’s new harmonised research data policy (drafted April 2025, expected to take effect early 2026) will mandate FAIR data practices across all research councils. Jisc launched its first Electronic Lab Notebook DPS in February 2025. And the AI agent paradigm is creating urgent demand for grounded, provenance-rich knowledge stores that most institutions have not yet provisioned. Universities that build coherent infrastructure now — combining open-source foundations with strategically procured commercial platforms — will be best positioned for the next decade of AI-augmented research.
Jisc procurement frameworks now cover three critical layers
The most immediately actionable finding is that Jisc has expanded its procurement frameworks significantly, and institutions should leverage these before exploring direct vendor negotiations.
Research Management Systems DPS covers CRIS and repository procurement. Pre-qualified CRIS suppliers include Digital Science (Symplectic Elements), Elsevier (Pure), 4Science (DSpace-CRIS), Worktribe, Vidatum, and MyScienceWork. Repository suppliers include Digital Science (Figshare), 4Science (DSpace, Dataverse), Elsevier (Digital Commons, Pure), CoSector/University of London (EPrints), and Worktribe. This DPS is compliant with Public Contracts Regulations 2015.
Electronic Research and Laboratory Notebook DPS launched in February 2025 — a pivotal development. Approved suppliers are Deltablot (eLabFTW), Labforward (Labfolder), Research Space (RSpace), and SciNote. This framework is open to new suppliers over time. Notably absent are LabArchives, Benchling, and Dotmatics, though LabArchives has a separate standalone Jisc/CHEST agreement running 2023–2026 with a dedicated UK server instance.
ORCID UK Consortium is managed directly by Jisc, providing institutional membership and integration support. UKRI requires ORCID for grant applications, making this effectively mandatory infrastructure.
DMPonline, the de facto UK standard for data management plans, is funded and delivered by Jisc through the Digital Curation Centre. It contains templates for every major UK funder (AHRC, BBSRC, EPSRC, ESRC, MRC, NERC, STFC, UKRI, Wellcome, ERC) and is used by virtually all Russell Group universities.
| Framework | Coverage | Key platforms |
|---|---|---|
| Research Management Systems DPS | CRIS + repositories | Symplectic Elements, Pure, Figshare, DSpace, Dataverse, EPrints |
| ELN DPS (Feb 2025) | Electronic notebooks | RSpace, eLabFTW, Labfolder, SciNote |
| CHEST Agreement | ELN | LabArchives (2023–2026, UK server) |
| ORCID Consortium | Researcher identity | ORCID via Jisc membership |
| DCC/Jisc | Data management planning | DMPonline |
No Jisc frameworks currently cover knowledge graph platforms, AI-integrated platforms, vector databases, literature AI tools, or reference managers. These represent procurement gaps where individual institutional negotiation or open-source self-hosting are the only options.
Segment 1: collaborative knowledge bases and structured knowledge graphs
This segment serves researchers building structured, queryable, linked datasets — from digital humanities catalogues to scientific ontologies. The market divides sharply between open-source community platforms and commercial enterprise tools, with a notable EU-versus-US data sovereignty split.
Wikibase (the platform behind Wikidata) is the standout open-source option. It provides a versioned, SPARQL-queryable triplestore with collaborative editing, multilingual support, and full W3C standards compliance. Deployments include the EU Knowledge Graph (European Commission), FactGrid (University of Erfurt), NFDI4Culture (TIB Hannover), and Enslaved.org (Michigan State). The British Library has used it with Wikimedia UK. Self-hosted via Docker, it carries zero licensing cost and gives complete data sovereignty. Its limitation is the technical expertise required for deployment and the Wikidata-specific data model conventions. AI integration is currently external only (QAnswer for natural language querying, SPARQL endpoints for agent access), but this makes it a viable AI grounding store when paired with an orchestration layer.
Ontotext GraphDB (now merging with Semantic Web Company to form Graphwise, February 2025) offers the most advanced AI-native knowledge graph capabilities. Its “Talk to Your Graph” feature combines LLM-powered natural language querying with SPARQL and vector RAG. It integrates with LangChain and is explicitly designed as a knowledge graph backend for AI systems. The free edition supports research use (though limited to 2 concurrent queries from v11+), and the enterprise edition uses per-CPU-core licensing. Based in Bulgaria/Austria/Switzerland, it carries low GDPR risk. UCL’s Sloane Lab project (with the British Museum and Natural History Museum, UKRI-funded) used GraphDB with metaphactory — the only confirmed Russell Group deployment in this segment.
Metaphacts (metaphactory) provides a low-code knowledge graph management platform with a newly launched “metis” AI platform for LLM-grounded conversational interfaces. Based in Walldorf, Germany, it offers free academic/research licenses and supports self-hosted, cloud, or SaaS deployment. Its FAIR alignment is excellent (built around RDF, SPARQL, SHACL). The Sloane Lab project at UCL demonstrates Russell Group applicability.
OpenAlex functions as critical knowledge graph infrastructure for scholarly metadata. This fully open, CC0-licensed knowledge graph from OurResearch catalogues 250M+ works with authors, institutions, topics, and citation networks. The Leiden University CWTS Ranking switched to OpenAlex in 2024, and Sorbonne deregistered from Scopus in its favour. No procurement is needed — it is free with an open API and full data dumps. For institutions building research intelligence or AI systems grounded in scholarly data, OpenAlex is the most important freely available resource.
Palantir Foundry and Cambridge Semantics Anzo (now Altair Graph Studio) are enterprise platforms with powerful capabilities but poor university fit. Both are US-headquartered with elevated GDPR risk, enterprise pricing inappropriate for academic budgets, and proprietary data models that conflict with open science principles. Palantir carries additional reputational risk from its defence and surveillance contracts and the controversial NHS Federated Data Platform.
Stardog and TopBraid EDG offer strong semantic AI capabilities but are US-based with unclear EU data residency options and commercial licensing that is typically prohibitive for academic use. PoolParty Semantic Suite (Austrian, now part of Graphwise) is EU-based but expensive (70,500 for server licenses).
Low-hanging fruit for this segment:
- OpenAlex — free, immediate, no procurement. Essential scholarly knowledge graph infrastructure
- ORCID (via Jisc) — already universally adopted; deepen integration as identity-layer infrastructure
- Wikibase — zero license cost, strong FAIR alignment, ideal for digital humanities and library special collections pilots
- GraphDB Free + metaphactory academic license — proven combination (Sloane Lab/UCL), both EU-based, free for research
- OpenAlex + Wikibase combination — ingest scholarly metadata into institutional Wikibase for custom research intelligence
Emerging to monitor: Graphwise (Ontotext + Semantic Web Company merger, Feb 2025) — if integration succeeds, this becomes the most comprehensive EU-based knowledge graph + AI platform. Monitor for 12–18 months.
Segment 2: electronic lab notebooks sit at an inflection point
ELN adoption at Russell Group universities remains early-stage and decentralised, with Edinburgh the clear leader (enterprise RSpace licence) and most other institutions lacking centrally mandated platforms. The February 2025 Jisc ELN DPS and a September 2025 Jisc community meetup (127 registrations) signal growing institutional momentum. The landscape divides into discipline-agnostic platforms suitable for institution-wide deployment and specialist tools for specific research domains.
RSpace is the strongest strategic choice for institutions prioritising open science and data sovereignty. Edinburgh-based (Research Innovations Ltd), it went open-source under AGPLv3 in June 2024 and is deployed at the University of Edinburgh and UCL — two Russell Group universities. It integrates with Figshare, Dataverse, DSpace, and DMPTool for FAIR data deposit, supports IGSN persistent identifiers, and is a member of the ELN Consortium supporting the .eln interoperability format. It can be self-hosted on institutional servers (zero GDPR risk) or used as managed SaaS. It is available through the Jisc ELN DPS. Its main limitation is that it comes from a relatively small company, though the open-source transition mitigates vendor dependency.
LabArchives offers the easiest procurement path — it is the only ELN with both a Jisc/CHEST agreement (2023–2026) and a dedicated UK server instance (uk-mynotebook.labarchives.com). Oxford has held an institutional contract since 2018. It is discipline-agnostic, suitable for arts through health sciences, and assessed by Oxford’s Information Security team as suitable for confidential data. The risk factor is its ownership chain: LabArchives is now part of Dotmatics, which Siemens acquired for $5.1B in July 2025 — introducing product direction uncertainty. The platform lacks native AI features and has limited FAIR alignment compared to RSpace.
eLabFTW is the most widely deployed fully open-source ELN globally, with massive adoption in German universities (TU Munich, RWTH Aachen, Heidelberg, Freiburg, Hamburg) and CNRS in France. Developed by Deltablot (France), it is available through the Jisc ELN DPS and can be self-hosted or use EU-hosted managed SaaS (Paris, Scaleway). It costs nothing to license and supports trusted timestamping and the .eln export format. Its limitation is that it requires Linux sysadmin capacity for self-hosting and lacks the polished institutional integration (LDAP/Shibboleth) that RSpace offers.
Benchling dominates in life sciences with its molecular biology toolkit, inventory, and registry system, and has launched Benchling AI (using Anthropic Claude, Google Vertex AI, OpenAI). However, it is US-hosted SaaS only with elevated GDPR risk for the academic tier (no confirmed EU data residency). The free academic plan has limited features, and data export is primarily PDF — poor for FAIR compliance. Suitable as a supplementary tool for individual molecular biology labs, not as an institutional solution.
OMERO (Open Microscopy Environment, University of Dundee) is essential infrastructure for any life sciences imaging core facility. UK-developed and open-source, it is used by the Euro-BioImaging UK Node (Edinburgh, Francis Crick Institute, King’s College London, Liverpool, York, Oxford Brookes). It supports 150+ microscopy formats via Bio-Formats, and its data model is the community standard for FAIR bioimaging data.
Chemotion ELN (KIT, Germany) serves chemistry departments with deep cheminformatics integration and direct deposit to the Chemotion Repository. It has excellent FAIR alignment as the reference ELN for NFDI4Chem but requires self-hosting expertise and has limited UK deployment.
Jupyter Notebooks are ubiquitous for computational research but are not a proper ELN — they lack audit trails, electronic signatures, access controls, and timestamping. Best used as a complement to a formal ELN for analysis workflows.
| ELN | Jisc availability | Russell Group evidence | GDPR risk | FAIR alignment |
|---|---|---|---|---|
| RSpace | ✅ ELN DPS | Edinburgh, UCL | Low (self-host/UK) | Excellent |
| LabArchives | ✅ CHEST 2023–26 | Oxford | Low-Moderate (UK server) | Moderate |
| eLabFTW | ✅ ELN DPS | None confirmed UK | Low (EU/self-host) | Strong |
| Benchling | ❌ | None institutional | High (US SaaS) | Limited |
| OMERO | ❌ (open-source) | Euro-BioImaging UK Node | Low (self-host) | Excellent |
Recommended dual strategy: LabArchives for broad institutional coverage (easiest procurement via CHEST) + RSpace for FAIR-focused research groups and departments prioritising open science and data sovereignty.
Segment 3: the institutional RDM stack is largely settled
Research Data Management is the most mature segment, with clear market leaders and established Russell Group adoption patterns. The typical stack comprises a DMP tool + CRIS + repository + long-term preservation, and Jisc frameworks cover most components.
DMPonline is the universal baseline — used by virtually all Russell Group universities, funded by Jisc, containing templates for all UKRI funders. Manchester alone has created over 3,000 DMPs through it since 2018. Any institution not already subscribed should do so immediately. No viable alternative exists for UK-specific funder compliance.
Figshare for Institutions is the dominant and growing data repository choice. Manchester (institutional deployment since ~2021), UCL (Research Data Repository since June 2019), and Oxford (supplementary to ORA Data) all use it. It visualises 1,200+ file formats in-browser, mints DOIs, integrates with Symplectic Elements, and is listed on the EOSC Marketplace. Available through Jisc’s Research Management Systems DPS. Its risk is commercial vendor dependency (Digital Science/Holtzbrinck) and AWS infrastructure.
Symplectic Elements is the dominant CRIS among top Russell Group institutions — literally founded at Imperial College in 2003. Oxford, Cambridge, Imperial, and many others use it. Its deep integration with Figshare creates a seamless CRIS + repository stack. Available through Jisc’s Research Management Systems DPS.
Pure (Elsevier) is the main CRIS competitor, used at Edinburgh, Manchester (alongside Figshare), Bath, York, and approximately 24 UK/Ireland institutions. The strategic risk is Elsevier vendor lock-in at a time when several Russell Group universities have tensions with Elsevier over journal pricing. Institutions should assess whether the Pure ecosystem creates unacceptable dependency.
DSpace remains widely deployed for institutional repositories (Cambridge’s Apollo, Imperial’s Spiral, Edinburgh’s DataShare), though some institutions have migrated to Figshare. Available through Jisc’s DPS via 4Science. EPrints (University of Southampton) continues at Birmingham, Glasgow, Nottingham, and the White Rose consortium (Sheffield, Leeds, York).
Zenodo (CERN-hosted, free, EU Open Research Repository) should be recommended as a default deposit target for researchers without a discipline-specific repository. It supports all file types up to 50GB per record, mints DOIs, and is a core EOSC component. Data stored at CERN (Switzerland) under adequate GDPR protections.
Arkivum (UK-headquartered, spun out of University of Southampton) addresses the long-term preservation gap. OAIS-compliant, deployed at Loughborough (integrated with Figshare) and Oxford Brookes. Important for EPSRC’s 10-year data preservation requirement. Available through direct procurement.
DataLad (MIT-licensed, Git-based data version control) is worth monitoring for computational research groups. It provides provenance tracking and distributed data management but requires command-line literacy. Growing adoption in the Helmholtz and NFDI ecosystems but limited UK Russell Group deployment.
AI integration across RDM tools is nascent. Zenodo’s MCP (Model Context Protocol) server for AI agent access and CKAN’s AI-ready data roadmap (CKAN 3.0) are the most forward-looking, but none currently offer production-ready AI agent integration.
Segment 4: literature tools split between proven managers and emerging AI assistants
This segment spans from established reference managers through to purpose-built AI research assistants. The key distinction for institutional decision-making is between tools that are proven, low-risk, and universally applicable versus AI-powered tools that offer transformative potential but carry accuracy and data sovereignty concerns.
Zotero remains the universal baseline reference manager — free, open-source (AGPL), recommended by virtually every academic library globally. Its plugin ecosystem now includes AI tools: ZotSeek provides 100% local semantic search using Transformers.js with no data leaving the institution, while PapersGPT connects to commercial LLMs for chat-with-PDF capabilities. The recommended institutional approach is to endorse Zotero with a curated AI plugin list, emphasising local-only plugins like ZotSeek for GDPR-safe AI search.
Semantic Scholar (Allen Institute for AI, non-profit) indexes 225M+ papers with AI-powered TLDR summaries, influential citation detection, and an experimental ScholarQA feature. It is free with an open API and open datasets, making it fundamental infrastructure for scholarly AI applications. Research Rabbit provides free citation-network discovery with Zotero integration. Both require no procurement.
Obsidian (local-first, markdown-based) is the strongest personal knowledge management recommendation, particularly for arts and humanities researchers. Its Zettelkasten-compatible workflow, zero cloud dependency by default, and AI plugins (Smart Connections for semantic search) make it ideal where GDPR sensitivity is paramount. Free for personal academic use with a commercial licence at $50/user/year.
Elicit is the most capable AI research assistant, searching 138M+ papers with sentence-level citations using Claude models. It supports systematic review workflows, interactive extraction tables (up to 20,000 cells), and structured evidence synthesis. Institutional/enterprise pricing is available with SSO and single-tenancy options. Its limitation is STEM-focus and US-hosted SaaS, though accuracy mitigation (all claims cite specific passages) is strong. Suitable for a departmental pilot in health sciences or psychology.
Consensus offers a distinctive “Consensus Meter” showing evidence distribution across 200M+ papers, built on a multi-agent architecture using GPT-5. It claims 5,000+ university subscribers globally (Princeton, Stanford, Harvard, UCLA). The key limitation is no arts or humanities coverage — it explicitly serves sciences and social sciences only. It also primarily analyses abstracts rather than full text.
Mendeley (Elsevier) is in decline for institutional purposes. Elsevier cross-shares usage data across ScienceDirect, Scopus, and PlumX. Feature removals (public groups, mobile apps, profiles, funding features all discontinued) signal strategic divestment. Its new AI features carry explicit accuracy disclaimers. Many institutions are actively migrating researchers to Zotero.
Lateral.io shut down in June 2025. Its successor, coordination.network, is experimental and open-source. Remove from procurement consideration.
Roam Research shows declining community activity and carries sustainability risk. Not recommended.
| Category | Recommended tools | Cost | Risk level |
|---|---|---|---|
| Reference management | Zotero + ZotSeek plugin | Free | Very low |
| Academic discovery | Semantic Scholar + Research Rabbit | Free | Low |
| Personal knowledge management | Obsidian + Smart Connections | Free–$50/yr | Very low |
| AI research assistant (pilot) | Elicit Enterprise | Institutional pricing TBC | Medium |
| Document summarisation | Scholarcy (acquired by Texthelp) | From $8K/yr institutional | Low-medium |
No tools in this segment have Jisc/CHEST agreements. Scholarcy’s acquisition by Texthelp (which has established UK education sector relationships) may open a procurement pathway worth monitoring.
Segment 5: AI-integrated platforms require a build-not-buy approach
This is the most rapidly evolving segment and the one where institutional technical capability matters most. No turnkey “AI knowledge platform” is procurement-ready for Russell Group universities at scale. The practical path forward is assembling an open-source stack on existing infrastructure (Azure or on-premise), supplemented by commercial platforms already in institutional use.
Azure AI Search is the strongest institutional recommendation. Most Russell Group universities already have Microsoft/Azure agreements (many via Jisc). UK South (London) and UK West (Cardiff) data residency is available. A pilot RAG chatbot can be built for approximately £120/month on the Basic tier (~£60/month) plus Azure OpenAI costs. Integration with Azure AI Foundry provides a complete managed RAG pipeline. The limitation is that Azure OpenAI model availability in UK South changes frequently — some models are only available via “Global” deployment, which may route data outside the UK. Institutions must verify model-by-model.
Qdrant (Apache 2.0, headquartered in Berlin, Germany) is the recommended open-source vector database for sensitive research data. It provides high-performance vector similarity search written in Rust, supports self-hosted Docker/Kubernetes deployment with full feature parity, and carries structural GDPR advantages as an EU-native company. SOC 2 Type II certified. Free when self-hosted. Managed cloud available in EU regions.
Weaviate (BSD-3-Clause, Netherlands-founded) is the strongest alternative, with hybrid vector + keyword search and multi-modal capabilities (text, images, audio). Self-hosted via Docker is free; managed cloud starts at $45/month. Both Qdrant and Weaviate integrate natively with LlamaIndex and LangChain.
LlamaIndex (MIT licence) and LangChain (MIT licence) are the essential orchestration frameworks. LlamaIndex specialises in data ingestion and RAG indexing with 300+ data connectors and built-in source node tracking for provenance. LangChain excels at complex agent workflows via LangGraph with human-in-the-loop capabilities. Both are free, self-hosted, and carry zero GDPR risk when run on institutional infrastructure. The recommended pattern is LlamaIndex for retrieval and LangChain for orchestration.
Neo4j offers the strongest provenance tracking of any platform reviewed. Knowledge graphs inherently capture entity→source document→chunk provenance chains, enabling explainable, multi-hop reasoning traces. The Community Edition is GPL-3.0 (free, self-hosted), while AuraDB managed cloud is available in EU/UK regions from $65/GB/month. Neo4j’s GraphRAG pattern — combining knowledge graph structure with vector search — is emerging as the gold standard for provenance-rich AI grounding. The LLM Knowledge Graph Builder (open-source) enables automated entity extraction from unstructured documents.
Confluence with Atlassian Intelligence (Rovo) represents genuine low-hanging fruit. Already deployed at most Russell Group universities, activating AI features requires only upgrading the plan tier (Standard from 10.44/user/month with 500 credits). Data Center self-hosting is available for full sovereignty, though AI features are cloud-only. This provides immediate institutional knowledge search with AI at minimal incremental cost.
Pinecone (proprietary, US-only SaaS, no self-hosting) carries high GDPR risk — the free tier is locked to US-East-1, and the company is subject to the US CLOUD Act with no EU data processing guaranteed on all tiers. Not recommended as a primary institutional store for sensitive research data. Glean is prohibitively expensive for most universities (~$60K annual minimum). Kùzu was abandoned by its maintainers in October 2025 — do not adopt; monitor community forks (Bighorn, Ladybug).
Recommended institutional stack for AI grounding:
- Azure AI Search (leveraging existing agreements) for rapid deployment with UK data residency
- Qdrant or Weaviate (self-hosted) for sensitive research data requiring full sovereignty
- LlamaIndex + LangChain (self-hosted) for orchestration with source provenance tracking
- Neo4j Community Edition (self-hosted) for provenance-critical applications requiring explainable retrieval
- Confluence + Atlassian Intelligence for immediate institutional knowledge search
Russell Group adoption patterns reveal a clear hierarchy
Research across the Russell Group reveals a maturity hierarchy. CRIS and repository infrastructure is well-established, with Symplectic Elements dominant at the top (Oxford, Cambridge, Imperial) and Pure prominent elsewhere (Edinburgh, Manchester). Figshare is the fastest-growing repository platform.
ELN adoption is the critical gap. Edinburgh leads with an enterprise RSpace licence and documented use of multiple platforms (Benchling, LabArchives across research groups). Oxford has LabArchives institutionally since 2018. Most other Russell Group institutions have no centrally mandated ELN — individual labs choose their own tools, creating fragmentation and compliance risk.
AI-integrated platforms are almost entirely absent at the institutional level. Individual research groups use vector databases and orchestration frameworks, but no Russell Group university has yet deployed institution-wide AI knowledge infrastructure. This represents both a risk (unmanaged researcher use of US-hosted AI tools) and an opportunity (first-mover advantage in FAIR-compliant AI grounding).
| Tool category | Russell Group adoption | Maturity |
|---|---|---|
| CRIS (Elements/Pure) | Near-universal | Mature |
| Repository (Figshare/DSpace/EPrints) | Universal | Mature |
| DMP (DMPonline) | Universal | Mature |
| ELN | Edinburgh, Oxford only at institutional level | Early |
| Knowledge graphs | UCL (Sloane Lab project only) | Nascent |
| AI platforms | None institutional | Pre-adoption |
GDPR and data sovereignty risk requires tool-by-tool assessment
The UK legal framework under UK GDPR and the Data Protection Act 2018 does not specifically prohibit public sector cloud data storage in any country, but institutions must conduct Transfer Impact Assessments for US-hosted services due to the US CLOUD Act. A 2025 survey found 82% of UK IT leaders would consider switching from Big Tech for greater data control, and 84% are concerned that geopolitical developments could threaten data access.
Tools with the highest GDPR risk in this landscape:
- Pinecone — US SaaS-only, free tier locked to US, no self-hosting, CLOUD Act exposure
- Benchling (academic tier) — US-hosted, no confirmed EU data residency for free accounts
- Palantir Foundry — US-based, defence/intelligence origins, NHS data controversy
- Diffbot — US SaaS-only, no confirmed EU data processing
- Mendeley — Elsevier cross-shares user behaviour data across its commercial ecosystem
Tools with structural GDPR advantages:
- Qdrant — EU-headquartered (Berlin), Apache 2.0, self-hostable
- Ontotext GraphDB/Graphwise — EU-based (Bulgaria/Austria/Switzerland)
- metaphactory — EU-based (Germany), free academic licenses
- RSpace — UK-based (Edinburgh), self-hostable, open-source
- eLabFTW — EU-based (France), EU-hosted SaaS option (Paris)
- All self-hosted open-source tools (Wikibase, Weaviate, LlamaIndex, LangChain, Neo4j CE, Chroma) carry zero inherent GDPR risk when deployed on institutional infrastructure
For sensitive personal data (student records, patient data, qualitative interview transcripts), the recommendation is unambiguous: use self-hosted solutions or Azure UK South with an institutional DPA. For non-sensitive research data, risk-assessed US SaaS tools with appropriate DPAs and Standard Contractual Clauses may be acceptable.
FAIR, UKRI, and EOSC alignment concentrates in specific tools
The new draft UKRI research data policy (April 2025, expected early 2026) will harmonise expectations across all research councils with explicit FAIR data requirements. Tools that directly support compliance include DMPonline (all UKRI funder templates), Zenodo (widely accepted deposit target), Figshare (funder compliance workflows), and the DSpace/EPrints repository ecosystem.
EOSC Marketplace listings confirmed for: Figshare for Institutions, Zenodo (core EOSC component), FAIRsharing (Oxford-based, ELIXIR Recommended), and European Galaxy Server. Zenodo is deeply embedded in EOSC infrastructure through the CERN EOSC Node and FAIRCORE4EOSC project.
FAIRsharing (fairsharing.org), based at Oxford, deserves particular attention as a curated registry of 1,641+ standards, 2,041+ databases, and data policies. It is an RDA-endorsed output, ELIXIR Recommended Interoperability Resource, and listed on the EOSC Marketplace. Institutions should treat it as a reference resource for FAIR compliance decisions.
Tools with the strongest FAIR alignment across all segments: OpenAlex (CC0, exemplary), Wikibase (W3C standards, linked open data), RSpace (designed around FAIR lifecycle), Zenodo (DOIs, OpenAIRE, EOSC), OMERO (community bioimaging standard), Chemotion ELN (NFDI4Chem reference), Zotero (open formats, open source), Semantic Scholar (open datasets, non-profit).
Five strategic recommendations per segment
Segment 1 — Knowledge graphs
- Deploy OpenAlex as scholarly knowledge graph infrastructure (free, immediate)
- Deepen ORCID integration via Jisc consortium as foundational identity layer
- Pilot Wikibase for digital humanities and library special collections
- Evaluate GraphDB Free + metaphactory academic licence for semantic web projects
- Monitor Graphwise merger integration over 12–18 months
Segment 2 — Electronic lab notebooks
- Procure LabArchives via Jisc CHEST for broad institutional coverage
- Deploy RSpace via Jisc ELN DPS for FAIR-focused research groups
- Offer eLabFTW (Jisc ELN DPS) as zero-cost option for self-hosting departments
- Mandate OMERO for bioimaging core facilities
- Evaluate Benchling Academic as supplementary free tool for molecular biology labs (with GDPR guidance)
Segment 3 — Research data management
- Ensure institutional DMPonline subscription (Jisc) if not already in place
- Deploy Figshare for Institutions (Jisc DPS) as primary data repository
- Recommend Zenodo as default deposit target for researchers without discipline-specific repositories
- Evaluate Arkivum for EPSRC 10-year preservation compliance
- Assess Pure versus Symplectic Elements with explicit vendor-dependency risk analysis
Segment 4 — Literature and document knowledge bases
- Endorse Zotero + ZotSeek as the institution-recommended reference management stack
- Recommend Semantic Scholar + Research Rabbit (both free) in library discovery guides
- Recommend Obsidian for personal knowledge management, particularly arts and humanities
- Pilot Elicit Enterprise with a research-intensive department (health sciences, psychology)
- Evaluate Scholarcy institutional licence for accessibility-aligned summarisation
Segment 5 — AI-integrated platforms
- Build initial RAG pilots on Azure AI Search (leveraging existing Azure agreements, UK data residency)
- Activate Confluence + Atlassian Intelligence for immediate institutional knowledge AI
- Deploy Qdrant + LlamaIndex (both self-hosted, open-source) for sovereign research AI infrastructure
- Pilot Neo4j Community Edition for provenance-critical knowledge graph applications
- Establish institutional guidance on acceptable use of LangChain/LlamaIndex orchestration frameworks
Emerging platforms warrant monitoring but not yet adoption
Several platforms are too early or too volatile for institutional commitment but could become significant within 12–24 months.
Graphwise (Ontotext + Semantic Web Company, merged February 2025) could become the default EU-based knowledge graph + AI platform if integration succeeds. Its combined GraphDB + PoolParty + AI RAG capabilities are unmatched in the EU vendor landscape, but merger execution risk is real. Monitor through mid-2027.
Microsoft Foundry IQ (launched November 2025) provides fully managed RAG pipelines with minimal configuration on Azure. If it matures and proves stable, it could dramatically lower the barrier to institutional RAG deployment. Evaluate stability by mid-2026.
Neo4j’s GraphRAG ecosystem is maturing rapidly and represents the emerging standard for provenance-rich AI grounding. The pattern of knowledge graph + vector search + LLM provides explainable retrieval that regulatory and research integrity contexts increasingly demand. Watch for university-specific packaging.
InvenioRDM (the CERN-developed software powering Zenodo) is being developed as a turnkey institutional repository. If it reaches institutional deployment maturity, it could offer a free, EOSC-integrated, FAIR-native alternative to Figshare.
Consensus is growing rapidly (8M users, 8× revenue growth) with institutional licensing expanding. Watch for UK university deals and potential Jisc engagement — though its exclusion of arts and humanities limits cross-disciplinary utility.
Coordination.network (open-source successor to the shuttered Lateral.io) is very early-stage but philosophically aligned with open science. Worth tracking for FAIR-aligned document knowledge base applications.
Conclusion: build the layers, not the monolith
The landscape analysis reveals that no single platform spans the full research knowledge lifecycle. Institutions need a layered stack integrating data management planning, active data capture, repository deposit, long-term preservation, and AI-ready knowledge infrastructure. The good news is that Jisc frameworks now cover the first four layers through compliant procurement routes. The fifth layer — AI-integrated knowledge platforms — remains the frontier where universities must assemble open-source components on institutional infrastructure or leverage existing Azure agreements.
Three structural insights should guide commissioning decisions. First, the open-source ELN market has matured significantly — RSpace and eLabFTW are now available through Jisc’s DPS, eliminating the primary objection to open-source adoption in risk-averse procurement environments. Second, GDPR risk concentrates in specific tools (Pinecone, Benchling academic tier, Mendeley, Palantir) rather than being a blanket concern — most segments have EU-based or self-hostable alternatives that eliminate data sovereignty risk entirely. Third, the AI agent paradigm favours tools with strong provenance tracking — knowledge graphs (Neo4j, Wikibase, GraphDB) and RAG frameworks with source attribution (LlamaIndex) will become more valuable than simple vector databases as research AI moves from experimentation to production.
The most impactful near-term action is to establish an institutional AI knowledge infrastructure baseline: Azure AI Search for managed RAG with UK residency, Qdrant/Weaviate self-hosted for sensitive data, and LlamaIndex/LangChain as standard orchestration. Combined with the established RDM stack (DMPonline + Symplectic/Pure + Figshare + RSpace/LabArchives), this provides a coherent, procurement-compliant, FAIR-aligned foundation for the next generation of AI-augmented research.