Coaching · Hong Kong General Chamber of Commerce
I will be an AI coach at the AI-Powered Workflow Automation Lab
9 Sep 2026 · 5 min read
This October I will be one of the HKU student coaches at the AI-Powered Workflow Automation Lab. Four evenings, 7, 13, 22 and 28 October from 6pm to 9pm, at the Hong Kong General Chamber of Commerce. A 12-hour programme where each organisation brings one real workflow and turns it into something that runs on its own.
Esperanza runs the Lab with HKU. The instructor is Keith Rumjahn. Every participant or participating team gets a named HKU student coach, one on one, to set up tools, translate business requirements into technical instructions, troubleshoot prototypes and push for improvements. I am one of those coaches.
Why most organisations stay stuck
A recent McKinsey study found Hong Kong is good at early-stage AI adoption, but turning that into sustained advantage needs executives who practice, governance that answers security concerns, and talent plans that attract the AI-native generation. The Lab was built around the four traps that keep teams stuck.
1.The Chatbot Ceiling. AI gives quick answers, but the team still does the heavy lifting. Scaling AI means delegating whole tasks, not collecting better prompts.
2.The Tool-Pile Problem. One more subscription does not fix the stack. Teams need a proven method for turning operational bottlenecks into automated systems.
3.The Demo-to-Desk Gap. A workflow that only works while the instructor watches is a demonstration, not an operational solution. Real workflows need testing, explicit boundaries and human approval checkpoints.
4.The Prototype Graveyard. Deployment brings hallucinations, missing data, edge cases and governance risk. Sustainable adoption needs clear ownership, quality controls and measurable outcomes.
What happens across the four sessions
1.Session 1, 7 Oct. Select your use case and build an AI specialist. Map the process, define success before choosing technology, and create a reusable specialist with Gemini Gems, Microsoft 365 Copilot agents or an approved equivalent.
2.Session 2, 13 Oct. Cross the delegation gap. Move from asking questions to assigning multi-step work to an agent, with permissions, boundaries, citations and approval checkpoints.
3.Session 3, 22 Oct. Build the tool the workflow needs. Use Cursor to wrap the process in a small internal app that colleagues can use without reading a prompt.
4.Session 4, 28 Oct. Make it survive the real world. Test against missing information, hallucinations, user mistakes, permission failures and adoption resistance, then prepare the internal pilot.
What you leave with
A tested AI workflow prototype for your own organisation, a governance checklist and a 30-day rollout plan.
A reusable AI specialist, one working agent task, and toolkits covering delegation, workflow architecture, prompt templates and a human-in-the-loop checklist.
A direct line to HKU AI talent through the coaching model and the HKU Co-operative Education Programme.
No coding background is required. The Lab is for managers, SME owners, school and NGO leaders who know their organisation's daily workflows and have basic AI familiarity. Teams of three or more get a group rate.
On privacy: coaches sign NDAs, and the Lab runs on a 100% synthetic data protocol. You can build and test on mock data, so no confidential material ever has to touch an external model.
Build internal AI capability, not vendor dependency. Bring one real workflow and leave with a tested prototype and a 30-day rollout plan, built on systems your IT already approves, with a dedicated coach assigned to your team.