What Should You Do Before You Deploy an Agent?
Plan how you will watch it, steer it and undo what it does. The panel's shorthand was that agents should be observable, controllable and reversible. Start them in parts of the business where the blast radius is manageable, so that a mistake stays small.
How Should You Oversee Agents While They Run?
Monitor them. Where you can, keep a human in or over the loop to check their work. Where that is not practical, at least use other AI agents to check the primary ones.
Who Is Accountable When an Agent Gets It Wrong?
The organisation is. The panel was clear that a company has to own an incident as a whole, and cannot blame the model or the tools. This is where reversibility pays off, because the first question after an incident is whether the damage can be undone.
What Is the Cheapest Way to Prepare?
Planning, which costs nothing. The closing advice from the stage was that the best thing an organisation can do to get ready for agentic AI is to plan properly before it starts.
Where Can I Read More?
Tech in Asia has not published an article on the panel. Our summary and Singdata's recap are on LinkedIn.
- How do companies think about managing the risks of agentic AI?, StrongKeep on LinkedIn, 17 September 2026
- How can enterprises safely deploy agentic AI in practice?, Singdata on LinkedIn
Related Reading
- Who is watching when an AI agent acts on its own?, Gaurav on CyberSG Voices on autonomy and accountability.
- Is it safe to let an AI agent act for you?, Gaurav on CNA938 on personal agents.
- Is AI becoming a new kind of cyber attacker?, Gaurav at Caixin's Asia New Vision Forum.
Frequently Asked Questions
Who was on the Tech in Asia agentic AI panel?
Gaurav Keerthi, CEO of StrongKeep, moderated and spoke. The panellists were Too Ping Tan of Changi Airport Group, Ethan Yu of Singdata and Smile Huang of Whale. The session, Trust at Machine Speed: Governing and Securing the Agentic Enterprise, took place at the Tech in Asia Conference in Singapore on 16 September 2026.
What does it mean for an AI agent to be observable, controllable and reversible?
Observable means you can see what the agent is doing. Controllable means you can steer or stop it. Reversible means you can undo what it has done. The panel recommended planning for all three before an agent is deployed.
Who is responsible when an AI agent causes an incident?
The organisation that deployed it. The panel's view was that a company has to take accountability for an incident as a whole, and cannot pass the blame to the model or the tools.
What Does StrongKeep Do?
StrongKeep is a self-serve cybersecurity platform for small and medium businesses. It bundles endpoint protection, phishing and malicious-link filtering, credential management, staff awareness training and phishing simulations, all run by the business itself from one dashboard. Compliance tooling is available on higher plans. StrongKeep holds the CSA Cyber Essentials Mark. Built for teams without an IT department. Plans are on the pricing page.
