Digital Science has launched its 2026 Catalyst Grant, inviting individuals, startups and research teams worldwide to apply for equity-free awards of up to £25,000 to develop what it calls trustworthy agentic workflows for research. The funding round, now in its 16th year, closes on 5 October 2026, the company said on 1 September 2026.
What the grant seeks
The programme targets early-stage ideas — from well-formed concepts to prototypes — that go beyond single-prompt generative tools to create autonomous, multi-step systems that can plan, execute and review research tasks. Digital Science emphasises that these systems should include built-in provenance, governance and auditability, enabling institutions to rely on agents when they act on behalf of researchers.
"AI tools are increasingly able to act on research, not just describe it. The next breakthroughs will come from agentic workflows, autonomous and multi-step, that researchers, institutions and funders can actually rely on," said Steve Scott, VP Portfolio Development at Digital Science.
Why "agentic workflows" matter
An agentic workflow is an AI-driven sequence that carries out several linked tasks — for example designing an experiment, running data analysis, and producing a written summary — with limited human intervention. Digital Science draws a distinction between such systems and the generative models commonly funded in recent years: trust, the company argues, comes from systems that can "show their working" with verifiable records of decisions and data lineage rather than post-hoc detection of errors or misuse.
The grant brief asks applicants to explain the problem they are solving, why their approach fits the agentic, trustworthy workflow theme, and the current state of development. Applicants need not have revenue, a finished product or a complete business plan to be eligible.
What the award supports
- Equity-free awards of up to £25,000 for selected projects.
- Open to individuals, startups and research teams globally.
- Suitable for prototypes, working products or robust concepts; emphasis on audit, governance and provenance.
The Catalyst Grant has a history of supporting novel software and technology aimed at advancing research. Digital Science describes the programme as intended to accelerate tools that increase research impact on society by helping early-stage innovators move from concept to demonstrator.
Implications for South African research
For South African institutions and researchers, the call highlights two practical priorities. First, the need to design AI systems that preserve data provenance and clear chains of decision-making so that outputs can be audited and reproduced. Second, the governance question: institutions must decide what level of autonomy is acceptable for AI agents operating on sensitive data or in regulated areas such as clinical or environmental research.
These priorities echo longstanding national conversations about research integrity, data stewardship and institutional accountability. Practical tools that record the steps an AI agent takes — who authorised actions, what datasets were accessed, which models produced which outputs — could help universities, funders and ethics committees assess risk and compliance without stalling useful automation.
How to apply and timeline
Digital Science opened applications on 1 September 2026 and set the deadline for Monday, 5 October 2026. Applicants should outline the problem, their approach to agentic workflows and the current development stage. The award is intended for early-stage builders who prioritise safety, provenance and institutional fit.
| Item | Detail |
|---|---|
| Award | Up to £25,000 (equity-free) |
| Eligible applicants | Individuals, startups, research teams (global) |
| Focus | Agentic workflows with audit, governance and provenance |
| Application deadline | 5 October 2026 |
Researchers and developers in South Africa who are exploring autonomous research assistants, reproducible pipelines or institutional governance of AI may find this call relevant. The emphasis on demonstrable provenance and accountability is likely to shape which proposals are competitive, reflecting broader sector concerns about safe deployment of systems that act, not just respond.
Further information and application details were provided by Digital Science in its announcement on 1 September 2026.
