Agentic workflows you can trust

• Trusted authorship and writing. AI co-writers with transparent contribution logs, autonomous citation and source verification agents, or cryptographic authorship attestation for high-stakes academic work
• Enterprise research workflows. Regulatory-grade evidence synthesis agents, governed knowledge graph workflows with provenance at every step, or explainable drug discovery and target identification tools
• Funding and institutional decisions. Researcher identity and disambiguation agents, automated reviewer matching and COI screening, or grant-portfolio intelligence for programme officers
• Publication integrity. Auditable manuscript screening, citation and figure provenance verification, or multi-step editorial workflow agents with full audit trails

If you’re making AI safe to put in front of an institution,  autonomous in service of researchers, with accountability built in, we want to hear from you.

1. Name the research decision your agent affects, and who makes it today (who, when, how often).
2. Say where it sits in the research lifecycle: writing, discovery, evidence synthesis, peer review, funding decisions or research integrity. If it isn’t one of these, it probably isn’t a fit this year.
3. Name the tool or system it has to live inside, and who already uses that tool every day. We are more interested in something embedded in real work than a standalone demo.
4. Check the work genuinely needs more than one step. If a single well-written prompt does the job, it isn’t an agentic workflow.
5. Test where it breaks. Give it a task it should refuse, flag or escalate, and record what actually happened.
6. Name your outcome metric, for example reviewer hours saved, time to funding decision, or duplicate work avoided.

Applying takes one short form and a few minutes.

Who can apply. Anyone with an early-stage idea: founders, research teams, students and individuals. Your tool can be a prototype, a working product or a well-formed concept. We fund early, so you do not need revenue or a finished build.

What to tell us. A clear description of the problem you are solving, your approach, why it fits an agentic, trustworthy workflow, and where you are today. If you have a demo or prototype, show us.

How we choose. Our judges score each application across seven areas: team, problem, solution, competitors, market, progress to date, and fit with Digital Science. Strong applicants are invited to a short interview with a live demo (where applicable) and a few questions.