AI-Driven Agents, Human-Centered Service
July 21 - 22, 2026
| Free CX Network Webinar Series
Andrew has built three generations of conversational AI at Super Dispatch (Intercom Fin, then Zendesk AI, now Botpress LLM agents) on a logistics platform that handles about 70,000 tickets a year. Each migration was mine end to end: platform selection, knowledge grounding, orchestration, guardrails, evaluation, and cost. The latest one tripled AI resolution from 21% to 66% and brought cost per conversation down to $0.67, roughly a fifth of what was0 paid before.
Every quarter he reports the numbers to leadership: 99.5% one-touch rate, 1% reopens, 77.9% AI CSAT, and 8,000+ LLM calls a month with zero errors.
Most of that comes from unglamorous plumbing rather than the model itself. Andrew and his team gate every agent change behind evals, keep a human approving promotions, use deterministic guardrails so the bot can't invent answers, and rebuilt the knowledge base for retrieval instead of human readers.
Andrew also runs support operations for Support, Onboarding, Compliance, Fraud, and Legal. The workflows he has automated there cut fraud resolution from 1,239 minutes to 159.
If you're working out where AI fits in your support org, or you launched a bot and the numbers let you down, that's a conversation he's always open to.
Day One - July 21
10:00 AM Full coverage, not full automation: How the best support teams are rethinking AI
The teams winning at AI support in 2026 aren't chasing the highest automation rate. They're building for full coverage: a model where AI and humans together resolve every conversation as efficiently as possible. This session unpacks what that looks like in practice.
We'll explore why "automate everything" is the wrong goal, and what to measure instead (resolution and coverage, not deflection rate). We will outline the three ingredients of a full-coverage operation: AI that resolves by taking real action (not just answering), handoffs so seamless the customer never notices, and a reinforcement loop where every human intervention makes the system smarter over time. To conclude, we'll explain why most tools and pricing models - built for deflection, not collaboration - actively work against this.
This session will be grounded in real numbers from a support team that moved from a fragmented "AI-plus-humans" setup to true full coverage, and shows a side-by-side look at why the underlying engine is what makes or breaks this model.
You'll leave with a practical blueprint for a support operation that gets better the more it runs - and full coverage neither AI nor humans could reach alone.
Attendees will learn:
- Why full coverage beats full automation, and the metrics that actually prove ROI (end-to-end resolution + coverage, not deflection-rate vanity)
- The three ingredients of a full-coverage model: AI that resolves by taking action, context-preserving handoffs customers never notice, and a human-feedback loop that compounds over time (and how to design each)
- Why most platforms and pricing models are built for deflection, not collaboration, and what to demand of your AI platform (with a live side-by-side look at why the engine underneath matters)
In Partnership With:
©2026 IQPC. All rights reserved.
