Hong Kong Trails Global Peers in Workplace AI Readiness, Cisco Survey Finds
Hong Kong ranks last for workplace AI readiness among 30 global markets, according to a Cisco survey conducted in August and released this week. Only 2% of organisations in the city qualify as "pacesetters," compared with 13% globally.
Pacesetters outperformed peers across road mapping, investment, infrastructure, measurement, and security. Hong Kong's gap is clear: just 32% of organisations in the city have a defined AI road map.
What the best performers do differently
- They document an AI road map (nearly all pacesetters did).
- They prioritise AI as a top investment and plan to expand use across functions.
- They add data-centre capacity and modernise infrastructure for AI workloads.
- They track outcomes and ROI, not just pilots and proofs of concept.
- They take AI-specific security threats seriously and invest ahead of incidents.
Signals of "AI infrastructure debt"
The survey flags technical debt holding companies back, describing "AI infrastructure debt" built up through compromises, deferred upgrades, and underfunded architecture. That debt slows deployment, inflates costs, and limits what teams can ship.
Hong Kong by the numbers
- 38% expect workloads to grow by 30% within three years.
- 68% are struggling to centralise data.
- Only 14% report adequate chip/GPU capacity for AI.
- Fewer than 1 in 5 can detect or prevent AI-specific threats.
- 71% plan to deploy AI agents to work with employees within a year.
Regional context
Simon Miceli, Cisco's managing director for cloud and AI infrastructure in Asia-Pacific, said he was "somewhat surprised" by Hong Kong's low readiness. Ben Dawson, senior vice-president and president of Asia-Pacific, Japan and Greater China, added: "Comparatively, Indonesia showed surprisingly high adoption, as did Thailand, so there may be stronger opportunities in emerging economies to leapfrog than their more developed peers."
AI is shifting from chatbots to agents
"We're moving past the era of question-answering chatbots and stepping into the next major phase of AI: agents that independently execute tasks," said Jeetu Patel, Cisco's president and chief product officer. Companies further along are seeing stronger returns.
Agents raise the bar for infrastructure, security, and governance. They trigger more transactions, touch sensitive data, and require observability across prompts, tools, and outputs.
What IT, engineering, and business leaders should do now
- Define a 12-18 month AI road map tied to 3-5 clear business outcomes. Prioritise use cases with measurable value and available data.
- Centralise data. Build or buy pipelines for quality, lineage, consent, and retention. Reduce copies and shadow datasets.
- Plan compute strategy. Right-size GPU/accelerator capacity; balance on-prem with cloud burst. Track utilisation and cost per token/inference.
- Modernise the stack. Ensure high-throughput networking, fast storage, and containerised deployments with autoscaling.
- Stand up MLOps and LLMOps. Version models and prompts, manage feature stores, and automate evaluation and rollback.
- Strengthen AI security. Add model gateways, secret and key management, prompt and tool-use policies, and detection for data exfiltration, poisoning, and jailbreaks.
- Prove value fast. Run time-boxed pilots, publish dashboards, and retire low-yield experiments.
- Prepare for agents. Start with narrow tasks, human-in-the-loop review, audit logs, and clear escalation paths.
- Upskill teams. Train product, data, security, and operations to ship AI safely and efficiently.
KPI checklist to track
- % of priority workflows augmented or automated by AI
- Time from idea to production (days)
- GPU/accelerator utilisation (%) and queue times
- Cost per 1,000 tokens/inference and per successful task
- Data lineage coverage and policy compliance (% of datasets)
- AI incident detection coverage and mean time to respond
- Agent task success rate and human override rate
- Model and prompt version adoption across environments
Helpful references
Level up skills, fast
If your team needs a focused path to ship real use cases and deploy AI agents safely, explore role-based programs and certifications here:
The gap is fixable. Start with a clear plan, fund the foundations, and measure outcomes. Delay increases debt; action compounds returns.
Your membership also unlocks:
