This is the story of how a small Hong Kong sports agency went from manually managing everything in spreadsheets to running a system that handles the boring parts automatically — so the team could focus on the work that actually matters.
The problem
The agency helps young athletes navigate opportunities with schools and programs abroad. The work involves a lot of matching: athlete profiles against school requirements, eligibility criteria, timelines, and family communication. All of it was done manually.
The team was spending hours each week on repetitive tasks: pulling data from emails, cross-referencing spreadsheets, writing similar messages to different families, tracking deadlines across multiple calendars. The work wasn't hard — it was just endless.
The team was good at their jobs. The tools were making them slow.
They came to us with a simple question: "Can AI help with any of this?"
What we built
We didn't start with AI. We started by watching how the team actually worked for two weeks. Where did they spend time? What was repetitive? What required judgment? What was just copying data from one place to another?
The answer was clear: about 60% of their process was data movement and pattern matching. The other 40% was human judgment — talking to families, understanding nuances, making calls that required context. AI is great at the first part. It's terrible at the second. So we built for the first part.
Profile processing
Athlete profiles used to be manually extracted from emails and forms, then entered into spreadsheets. We built a pipeline that ingests profiles from multiple sources, extracts key data points (academic records, athletic stats, preferences), and structures them into a searchable format.
Eligibility matching
Matching athletes to programs used to mean cross-referencing multiple spreadsheets with different criteria. Now the system automatically checks eligibility against program requirements and flags matches — or flags when something doesn't fit and needs human review.
Communication drafts
The team was writing similar emails to different families, manually customizing each one. The system now drafts personalized communications based on the athlete's profile and the specific program, ready for the team to review and send. Not auto-send — draft. The human always decides.
Deadline tracking
Application deadlines, document submissions, follow-ups — all tracked automatically with proactive alerts. No more "oh no, that was due yesterday" moments.
The results
The team went from spending 30+ hours a week on data management to about 8. They used the freed-up time to actually talk to families, build relationships, and grow the business. The system didn't replace anyone — it gave them their time back.
What we learned
A few things stood out from this project:
- Start with observation, not assumptions. We almost built a chatbot for family communication. Watching the team work showed us the real bottleneck was data entry, not conversation.
- Keep humans in the loop. Every output gets reviewed before it goes out. The AI drafts, the human decides. This builds trust and catches edge cases.
- Measure before and after. We tracked hours spent on manual tasks for two weeks before building anything. That baseline made the impact undeniable.
The bottom line
This wasn't a flashy AI project. No one-demo-to-impress-the-board moment. It was a quiet, practical system that made a small team significantly more effective. The technology isn't the interesting part — the outcome is.
The agency now serves three times more families with the same team. The quality went up because the humans spend time on the parts that need humans. And the system keeps getting better as we learn from real usage patterns.
That's what "AI that actually works" looks like in practice. Not magic. Not a demo. Just software that does the boring parts so people can do the important parts.