Speaking · The University of Hong Kong
First talk at HKU: what two students did with AI agents
8 Sep 2026 · 4 min read
Today I gave my first ever talk at The University of Hong Kong, in IIMT3689 Cloud Computing: Essential and Application in Entrepreneurship, for Prof Chan and Raymond's class. Not about a finished product. About what happens when two students decide to build something with AI agents.
The setup
Two students who met on a tennis court, then started coding together every single night.
The project started on 17 April. Two months of nights on a Lap Chee TV-room sofa, full-time students the whole time.
Two people (Hong Kong + Edinburgh) and N agents, working like a team of ten.
What we actually did with AI agents
1.Distribution. We could not afford marketing, so an agent became our marketer, drafting, sending and following up. One of those emails landed with a professor who said yes. This talk exists because an agent wrote a better email, sent at a better time.
2.Volume work. Agents parse documents, chase clauses and draft the reports. Humans review, sign and decide. Machines do the volume; people do the judgement.
3.Trust by design. Every answer cites its source. Every output routes to a named human reviewer. SHA-256 audit trail. For SME finance, compliance is the product, built from the first commit, not bolted on.
4.One niche. One platform, working title Automation, custom for each SME FI today, SaaS later.
What I told the students (and believe even harder now)
Fall in love with the field first. Agents amplify obsession; they do not create it.
Stay curious. This field changes every week.
Specialise. Boring plus deep beats broad plus shallow.
It was my first talk, so I am still learning. But if you are a student with an idea you keep not building, start tonight. Two people and a laptop are enough.
The best part came after the class. I got to sit down with Prof Chan and Raymond, two experts in this field, and talk about our approach face to face. Their questions and feedback were worth more than any slide I showed.
Thank you Prof Joseph Chan and Raymond Tsang for the opportunity and the feedback, and Lewis Simpson for being the other half of the team.
P.S. If you work in or with SME financial institutions in Hong Kong, we co-build custom AI workflows tailored directly to your operations and engineered for 100% compliance with SFC, HKMA, and PCPD standards. We published a live interactive demo on our website at automation.jonathansimpson.co by Jonathan Simpson & Co so you can see governance in practice: zero data retention and no login required, verifiable outputs with direct source citations, full SHA-256 audit trails and human-in-the-loop review routing. Test it on the site, see how it cites its sources, and send me a message if you want to map out a custom workflow for your firm.
Try the live demo