The Drowning Support Team

Most chatbots are built by people who have never answered a support message in their life.

I architect tools, systems, and processes for an operations platform — but because I touch every layer of the business, from compliance and legal documentation to transport, recruitment, finance, and daily support, I also handle 30+ user requests every single day from Service Providers, On-Site Managers, Area Managers, and department leads.

That cross-functional exposure is exactly why I build better solutions. I don’t guess what needs automating — I know, because I’ve done it hundreds of times.

Part 1 — I improved the Ticketing system.
Redesigned the Support section and issue categorisation:

• Category-first routing: Ask the right questions before anything else
• Clear escalation triggers: When to escalate and to whom
• Ticket lifecycle visibility: Everyone sees status, no more "what happened to my ticket?"

Part 2 — I built an AI Support Chat.
Not a FAQ bot. Not a keyword matcher. A specialised intelligent agent trained on 101 real scenarios — every one extracted from conversations I have had with real users.

When a Service Provider messages "I can’t access the app," it doesn’t show a help page. It works out whether it’s a payment overdue, an expired subscription, a wrong password, or an account that was never activated — and gives the right answer for that specific case.

• Visual screen recognition — a Service Provider sends a screenshot of an error, the agent identifies the screen and walks them through the fix
• PII anonymisation pipeline — personal data is stripped before the AI ever sees it. 62 automated tests make sure it stays that way
• 88 escalation triggers — legal threats, payment disputes, safety concerns are detected and handed off to a human with full context
• Account-aware responses — 11 account states via 3 read-only API endpoints. No database access. Privacy-first by design
• GDPR-compliant conversation logging — every interaction auditable