AI implementation for Hong Kong SMEs

Fix one painful workflow. In days, not quarters.

Small, fixed-scope AI operations projects for teams buried in messages, spreadsheets and manual follow-up. You work directly with a former Series B COO—not a junior agency team.

  • HK$2,800–8,800 fixed scope
  • Working system in 5–14 business days
  • English · 廣東話 · 普通話
live workflow
01
INPUTCustomer message + photo
received
02
AI PROCESSExtract · classify · check
03
OUTPUTOwned task + deadline
done
manual steps7 → 1
responsehours → min
Built for operating teamsFixed scopeDirect founder deliveryPractical handover

Productized services

Start small. Ship something useful.

Every offer has a clear boundary, price and handover. If your problem does not fit, I will say so before you spend money.

01Start here if the problem is fuzzy

AI Workflow Audit

HK$2,8005 business days

Turn one painful workflow into a practical automation blueprint—with ROI, risks and a build recommendation.

  • 60-minute workflow interview
  • Current-state process map
  • ROI-ranked automation opportunities
  • Implementation blueprint and fixed build quote

Best for: teams that know where time is being lost but are not sure what to automate first.

Services →
03For decisions stuck in spreadsheets

Decision Dashboard Sprint

HK$8,80010–14 business days

Replace recurring spreadsheet assembly with one decision-ready dashboard and an automated management digest.

  • Up to three existing data sources
  • Cleaned daily data flow
  • One focused KPI dashboard
  • Daily or weekly email digest plus handover

Best for: sales, revenue, inventory, service levels and multi-location operations.

Services →
Scope boundary

Platform subscriptions, major data migration, custom mobile apps and work outside the listed scope are excluded. Any extra work is quoted and approved before it starts.

Services →

Selected anonymized implementations

From scattered tools to one accountable operating workflow.

These are real systems I have implemented for operating businesses. Broad Link can connect the intake, data, decisions and follow-up—not just add an AI chatbot on top.

01Multi-location retail operator

Sales, stock and customer incidents became one operating system.

Before

Payment records, machine sales, stock levels and customer reports lived across separate portals, spreadsheets and inboxes.

What I built
  • Reconciled payment and machine data into a controlled sales record and live dashboard
  • Sent an 08:30 sales and replenishment digest with route-ready field tasks
  • Turned customer messages and photos into matched incidents with human-approved corrective actions
Operating result

Management can see trusted revenue, the field team knows what to refill, and customer issues are traceable to the right transaction, machine and product slot.

Payments + machines + messages → verified operations workflow
02Regional unattended-retail operator

Technician requests, sales reporting and live errors stopped depending on manual checking.

Before

Operational requests arrived in chat, managers checked sales portals by hand, and customer-facing system errors were easy to discover late.

What I built
  • Parsed chat requests into de-duplicated, owned technician tasks with urgency, due date and checklist
  • Delivered a daily multi-market sales digest with location and period comparisons
  • Monitored customer-visible errors every minute with duplicate protection and escalation controls
Operating result

Technicians receive structured work, management starts the day with the same numbers, and important service failures surface quickly.

Chat + sales portal + error logs → tasks, digest and alerts
AI intake & routingData reconciliationDashboards & digestsMonitoring & alertsHuman approval controls

How it works

One bottleneck. One owner. One short sprint.

01

Qualify

Tell me the workflow, current process, budget and timing. Strong-fit enquiries can book immediately.

02

Map

We define the happy path, exceptions, source data and a measurable before/after outcome.

03

Build

You review a working version early. I add safeguards, documentation and handover.

04

Verify

We test with real examples and confirm the workflow works before the sprint closes.

A good first project

Small enough to ship. Important enough to matter.

✓ Good fit

  • A workflow repeats every week
  • One owner can provide examples and access
  • The result can be measured in time, errors or visibility
  • You want to start within 60 days

× Not yet a fit

  • A general “teach us about AI” session
  • A large enterprise transformation disguised as a pilot
  • No process owner or sample data
  • A request for unpaid speculative development

Your implementation partner

Operator judgment, AI delivery speed.

I’m Justin Chiu. I co-founded and ran operations for a Series B AI fintech for seven years, scaling to 120+ people across four countries. I owned finance, HR and IT, secured HK$40M+ in public funding, and led ISO 9001/27001 certification. Broad Link applies that operating discipline to small, useful AI systems.

LinkedIn profile ↗
01Former Series B COO
02120+ team · 4 countries
03US$13M raised
04HK$40M+ grants
05ISO 9001 / 27001

Start with the bottleneck

Get a fit check—and a call slot if it makes sense.

This takes about three minutes. You will get a clear next step; qualified projects can choose a Tuesday or Thursday afternoon call immediately.

Current booking windows

Tuesday & Thursday · 2:00–5:00 pm HKT

No mailing list. No sales handoff. Your details go directly to Justin.

Before you enquire

Is the listed price final?+

Yes, for the listed deliverables and assumptions. If discovery reveals extra scope, you receive a separate quote before work begins.

Do I need clean data?+

No. You do need representative samples and someone who understands the current process. Data cleanup beyond the sprint boundary is quoted separately.

Who owns the system?+

You do. The handover includes the working files, configuration and a plain-language operating guide.

Can this become a larger project?+

Yes—but only after the first sprint proves the workflow and economics.