Why Governance Now Needs to Be Built In…
For the last few years, much of the AI conversation has centred on systems that answer questions, generate content and make recommendations. That is changing.
AI is increasingly being given the ability to take action – using tools, making decisions and carrying out multi-step tasks on behalf of an organisation.
That shift is at the heart of a recent move by the Responsible AI Institute (RAI), which announced three new members of its Governing Board in September 2026: AI risk and security specialist Avivah Litan, enterprise AI leader Cameron Davies, and Harvard Business School Professor Suraj Srinivasan.
The appointments reflect a bigger question facing technology leaders: What happens when AI stops simply advising people and starts acting for them?
From chatbot to colleague
A chatbot generating a draft email is relatively easy to understand. An AI agent that can read the email, decide what needs doing, update a database, call an external service and complete the task without waiting for someone to approve every step is a very different proposition. The technology may still use familiar AI models, but the engineering challenge changes significantly.
Developers now need to think about what an AI system is allowed to do, which tools it can access, what information it can see and what happens when something goes wrong. That means responsible AI is moving much closer to the software itself.
Guardrails become part of the architecture
Traditional software has permissions, authentication, logging and rollback mechanisms built around it. Agentic systems need similar protections, but with an additional layer of uncertainty because the system is making decisions dynamically.
For technical teams, that could mean building:
- tightly controlled tool and API permissions
- isolated execution environments
- detailed audit trails
- human approval points for higher-risk actions
- monitoring for unexpected behaviour
- mechanisms to stop or roll back an agent when necessary
In other words, AI governance cannot simply live in a policy document. Increasingly, it needs to be implemented as part of the technology.
A new role for governance teams
This also changes the relationship between developers, security teams and governance specialists. The Responsible AI Institute describes its work in terms of moving from responsible AI principles towards practical controls and evidence that systems are operating within defined boundaries. That is an important distinction.
It is one thing to say an AI agent shouldn’t access sensitive information. It is another to have technical controls that prevent it from doing so, and an audit trail that can demonstrate what happened afterwards.
As organisations deploy more autonomous systems, that difference is likely to become increasingly important.
The next phase of AI engineering
Agentic AI doesn’t make traditional software engineering obsolete. If anything, it makes good engineering discipline more important. The developers building these systems will need to think about permissions, security, testing, observability and failure recovery alongside prompts and models. And governance teams will increasingly need enough technical understanding to work directly with those systems.
The big shift is simple: When AI can act, responsible AI becomes an engineering problem as much as a policy problem.
As AI moves from generating answers to taking action, the importance of professional accountability will only grow. This is where professional registration can play an important role. The IST has developed a dedicated AI Professional Registration framework, recognising practitioners who can demonstrate technical competence alongside responsible and ethical AI practice, governance awareness and a commitment to continuing professional development. With registration levels reflecting increasing autonomy, responsibility and professional impact, it provides a way for individuals and organisations to demonstrate that the people designing, deploying and overseeing AI systems are not only technically capable, but committed to professional standards and accountability.
Find out more about the IST AI Professional Registration Framework.
Source & reference
Primary source: Responsible AI Institute, “Responsible AI Institute Adds Three Leaders to Governing Board for the Agentic AI Era”, September 2026.
Source: Responsible AI Institute
Announcement: PR Newswire — Responsible AI Institute Adds Three Leaders to Governing Board for the Agentic AI Era
IST Professional registration: Institute of Science & Technology — AI Professional Registration