JSM’s Joe Choy on why AI should support professional judgment, not replace it

Taking us through a case study on deploying internal AI agents, the legal expert stresses successful AI adoption should empower people to focus on higher-value work.

Professional judgement, responsiveness, accuracy, and confidentiality are non-negotiable in client-facing industries, particularly in legal services.

As artificial intelligence (AI) becomes increasingly embedded in everyday work, organisations across highly regulated sectors face a common question: how can they unlock productivity gains while maintaining accountability, trust, and professional standards?

For Hong Kong law firm Johnson Stokes & Master (JSM), the answer lies in a ‘governance-first, human-led’ approach to AI adoption.

The firm has recently developed a growing suite of internal AI agents designed to help its lawyers spend less time on routine tasks and more time on legal analysis, client engagement, and strategic advisory work. Crucially, all AI-generated output continues to be reviewed, refined, and approved by qualified lawyers before it reaches clients.

“Our focus has been to use the technology to enhance the quality, consistency, and efficiency of client service, while keeping legal judgement, professional oversight, and accountability central to every outcome,” said Joe Choy, Employment & Benefits partner at JSM and the leader behind the initiative.

In an exclusive interview with HRO’s Tracy Chan, Choy shares why responsible AI adoption is not about embracing technology for its own sake, but about creating a scalable model that supports people, strengthens governance, and enables professionals to focus on higher-value work.

Q: What were the biggest challenges JSM encountered when developing and deploying its internal AI agents, and how did you overcome them?

As organisations across regulated industries explore AI adoption, we believe many face the same question: how can AI be scaled responsibly without undermining accountability, trust, or professional judgment? Our answer was guided by a clear principle: AI should support professional judgment, not replace it.

To put this into practice, we adopted a governance-first approach and a clear human-in-the-loop model. Our AI agents operate within predefined parameters, a limited scope and controlled access to internal knowledge. They do not provide autonomous legal advice, do not self-learn from client data, and do not bypass professional review. Lawyers remain fully accountable for all client-facing work, ensuring that professional judgment, accountability, and high standards are maintained throughout the process.

Q: How do you ensure AI-generated outputs don't cloud human judgement, and that internal talent continues to think and act for themselves while becoming more productive?

The value of AI in regulated legal services is not simply about speed. It is about delivering consistent and quality work, strengthening governance, and providing faster access to relevant legal insights, while preserving clear human accountability for every client-facing outcome. This approach comes to life through a growing suite of purpose-built AI agents

One example is the ‘Employment Legal Advice Copilot Agent’, designed to support high-volume employment advisory work. The agent helps organise matter context, identify key issues, and generate structured first pass drafts covering risks, options, and recommended next steps. Lawyers then review, refine, and finalise the advice before anything reaches a client.

Deploying this agent has changed how we handle high-volume employment instructions. What stood out most was seeing the system perform in live matters.
The output provides a genuinely strong starting point, allowing our lawyers to spend more time on the strategic judgement and client counsel that matter most especially for nuanced employment issues.

The firm also has a dedicated Learning & Development (L&D) team responsible for driving and overseeing AI training and adoption across the firm. We provide a comprehensive range of legal skills training programmes that focus on developing skills such as legal analysis, risk assessment, and critical thinking.

The objective is to equip young lawyers with strong legal fundamentals rather than encouraging reliance on ready-made answers. This is a key example of our ongoing investment in cultivating critical thinking among the next generation of lawyers.

Q: What steps has JSM taken to drive adoption and participation in using these AI agents across the firm, and what seems to work best?

Rather than introducing standalone AI tools, JSM chose to embed AI directly into the platforms lawyers already use every day, including Outlook, Teams, and Word. By integrating AI into existing workflows, adoption has been gradual, practical, and closely aligned with the way our lawyers already work.

We formalised our AI policy before rolling out any tools firm-wide, ran structured pilots and consulted widely, so legal teams could see where AI was useful and where the boundaries were.

A key factor in driving adoption has been demonstrating the practical value of AI in lawyers’ day-to-day work. Our experience is that trust was earned, not mandated.

Rather than requiring lawyers to use AI, we focused on ensuring that the tools could be applied responsibly and in a way that supports, rather than compromises professional standards. As lawyers saw firsthand how AI could enhance efficiency and assist with routine tasks while maintaining quality and oversight, this gradual, experience-based approach to build trust has proven to be one of the most effective drivers of adoption across the firm.

Q: JSM has emphasised a ‘governance-first’ approach to AI adoption. What does responsible AI implementation look like in a highly regulated profession such as law, and what lessons can HR and business leaders in other industries learn from your experience?

Responsible AI means putting governance, accountability, and safeguards in place before technology is deployed at scale. The broader lesson for business leaders is that governance should never be treated as an afterthought. We were deliberate about defining what the agent could and could not do, while ensuring lawyers remained firmly in control throughout the process.

We also recognise that technology adoption is not solely a technology challenge. Successful implementation requires investment in people.

Through targeted training, practical guidance, and ongoing support, we help our lawyers understand how AI can be integrated into real legal workflows while maintaining the highest professional standards.

Photo / Provided