AI agents are moving quickly from prototype to production, but the organizations holding sensitive and restricted research data have no shared way to decide whether they are ready to deploy them. Agent workflows are easy to demonstrate on open data. The difficulty starts when the data cannot leave the institution, when access is governed by consent and ethics approvals, and when institutional policy requires a documented plan before an agent touches anything.

The VANTAGE Working Group addresses that gap. Over 18 months we will develop the Agentic Landscape Assessment Framework (ALAF), an open, tool-agnostic framework that an institution can use to assess its technical environment, governance posture and data-custodian workflows, and reach a defensible conclusion about how, or whether, AI agents should be used in a given workflow. ALAF is designed to return “not appropriate here” as a legitimate result.

Alongside ALAF we will maintain the VANTAGE Technology Radar, a continuously updated knowledge base of agentic developments kept current by agents under human review, and run workshops applying ALAF to real institutional workflows. We will also update the RDA guidance on data visitation language for ethics committees and informed consent to cover agentic access.

We welcome participants from research institutes, data commons, data portals, and any organization weighing agentic AI against real governance constraints. Data stewards, research software engineers, policy and compliance staff, and domain researchers are all needed. No prior work with AI agents is required. The group builds on the RDA Artificial Intelligence and Data Visitation WG and coordinates with the Data Director Agentic AI Blueprint WG and the Dataspaces in the Research Ecosystem IG.

AI agents are moving quickly from prototype to production, but the organizations holding sensitive and restricted research data have no shared way to decide whether they are ready to deploy them. Agent workflows are easy to demonstrate on open data. The difficulty starts when the data cannot leave the institution, when access is governed by consent and ethics approvals, and when institutional policy requires a documented plan before an agent touches anything.

The VANTAGE Working Group addresses that gap. Over 18 months we will develop the Agentic Landscape Assessment Framework (ALAF), an open, tool-agnostic framework that an institution can use to assess its technical environment, governance posture and data-custodian workflows, and reach a defensible conclusion about how, or whether, AI agents should be used in a given workflow. ALAF is designed to return “not appropriate here” as a legitimate result.

Alongside ALAF we will maintain the VANTAGE Technology Radar, a continuously updated knowledge base of agentic developments kept current by agents under human review, and run workshops applying ALAF to real institutional workflows. We will also update the RDA guidance on data visitation language for ethics committees and informed consent to cover agentic access.

We welcome participants from research institutes, data commons, data portals, and any organization weighing agentic AI against real governance constraints. Data stewards, research software engineers, policy and compliance staff, and domain researchers are all needed. No prior work with AI agents is required. The group builds on the RDA Artificial Intelligence and Data Visitation WG and coordinates with the Data Director Agentic AI Blueprint WG and the Dataspaces in the Research Ecosystem IG.