What a week! We’re coming off two milestone experiences—connecting with over 1,000 global service design leaders in Germany at the Service Design Global Conference (SDGC) and unpacking the future of enterprise AI governance in San Francisco.

Whether we’re designing human-centered service systems or architecting agentic workflows, one truth remains clear: technology only creates durable value when it is anchored in clear human intent, organizational context, and systems thinking.

Service Design Global Conference (#SDGC) Recap: Global Community in Motion

​
Heading home from Germany after an incredible experience supporting and participating in the Service Design Global Conference!

While I originally planned to share a run-of-the-mill event recap, what resonated most deeply was the genuine warmth, kindness, and hospitality of this community. The other major thread of all the talks was: Access to information does not mean everyone understands information. Whether designing a service blueprint or a multi-agent system, human context is everything.

A sincere thank you to the Service Design Network (SDN) leadership, the SDN community, and the organizing team for designing such an intentional, seamless gathering.

I’m leaving energized, inspired, and grateful. Until next year in Japan!

​New Blog Post: The AI Bottleneck Has Shifted​

​
Over the last 6 weeks I attended the AI Conference, Tech Week, and Product Con in San Francisco and the SDGC conference in Germany, I learned a lot! Though, I continue to believe that we’re repeating history, not disrupting as much as we think. Here are the four key beliefs and shifts I feel are required to move past fragile AI pilots and build durable, enterprise-grade AI capability:

1. Treat Agents Like Workforce Roles, Not Software Feature Sets: AI agents inherently see only what is above the waterline—they lack past intent, organizational memory, and subtle human nuances. Giving an airline chatbot full authority led Air Canada to be held legally liable in court for its bot's mistakes (Moffatt v. Air Canada, 2024). Make sure your existing human knowledge and systems are well documented (RAG) and give every agent a defined scope, explicit context, and a named human manager.
   ​
2. Don’t Let AI Grade Itself (Start with Human-Calibrated Prototypes To Build Trust): As enterprise workflows shift from deterministic rules to probabilistic intent, our operating model must be: "Police the rules, coach the intent" (Quote from AI conference talk). Deploying uncalibrated algorithmic judges creates hidden variance. For example, UnitedHealth's coverage-denial model getting overturned 90% of the time on appeal (Estate of Lokken v. UnitedHealth Group, 2023). Build real-time evaluation traces paired with a reliable "Control-Z" rewind mechanism.
   ​
3. Cost Per Outcome > Pure Token Cost: Over 70% of enterprise AI pilots stagnate due to unclear success criteria rather than technical limitations. Klarna ignored raw token expense and measured a single downstream outcome: cost per resolved interaction. By redesigning the service workflow from scratch, their system absorbed two-thirds of inquiries, cut resolution times from 11 to under 2 minutes, and drove an estimated $40M annual profit improvement.
   ​
4. Data Sovereignty in Motion: Traditional perimeter security addresses data strictly at rest. Agentic workflows put data in continuous, autonomous motion across global enterprise ecosystems—introducing active safety vectors across Human-to-Agent, Agent-to-Human, and Agent-to-Agent boundaries. Protecting enterprise data requires real-time architectural guardrails around every active interaction layer.
   ​

Preparing for 2027 Annual Planning?

At Lini, we help enterprise leaders bridge human-centered service transformation with agentic AI architecture. Book a 2027 Strategy & Architecture Session with Lini​

Make sure to follow Lini LLC on LinkedIn for more thoughts, project updates, and learnings on AI and transformation.

Best,
Sylvia