Designing for Truth in an Era of AI Hallucinations / Inside Service Design / Episode #08
Service Design Show
Designing an agent: step-by-step framework
Jessica walks through framing the problem, scoping users, personality, knowledge, testing, and iteration steps.
We need to talk about the "intern" sitting on your desktop...
Come on, you know the one.
Sure, they are fast, very eager to please, and can process data at lightning speeds.
But they also have a bad habit of hallucinating facts and making things up just to make you happy.
Of course, I’m talking about AI.
It is fair to say that we are past the initial "wow" phase of generative AI.
Now, for us service design professionals, the real question is: How do we actually hire, train, and trust this new digital colleague?
That is the focus of this episode of our Inside Service Design series.
We sit down for a chat with two brilliant professionals: Jessica Dugan and Judith Buhmann.
They share a grounded, hype-free look at how they are integrating AI into their own existing workflows. Not as a replacement for our work, but as a "Junior Associate" who needs some (sometimes a lot) management.
To make this real, Jess walks us through the framework she uses for building her own custom AI agents. She explains how to define their "persona," scope their tasks, and curate their knowledge base so they can actually be useful (and safe).
And Judith shares a critical perspective on why we can’t fully trust AI yet. We explore why we need to treat AI as an "unreliable narrator" especially when working with vulnerable groups.
So if you are feeling a bit somewhat by the pressure to "use AI" but aren't sure how to do it responsibly, this conversation has some key insights you don't want to miss.
Here's a question: If you had to give your current AI tools a "performance" review, what rating would you give them? A) Employee of the month B) Promising intern (needs supervision) C) Chaos agent (fires random info at me).
Let me know, I’m really curious where we are all at!
Be well,
~ Marc
--- [ 1. GUIDE ] ---
00:00 Welcome to the November Round Up
04:00 Jess's journey into service desig
09:45 Judith's challenge
12:30 Designing for the employee experience and internal systems
14:00 The "Pros" of in-house service design
15:30 The necessity of patience and deep knowledge for in-house success
18:30 Judith topic
19:00 Jess topic: Building (and trusting) your own AI agent
23:00 Why we cannot fully trust any AI
27:00 Scoping the AI agent's role and understanding user need
29:00 Designing the "Human" side: Setting personality and tone for your agent
33:45 Accessibility: Is it actually hard to build your own agent?
35:30 Human-in-the-loop: Regulation and ensuring data accuracy
40:00 Why transparency matters more than just "trust"
47:00 Getting organizational buy-in for AI tools
54:45 Markers of success: How service blueprints live on after the workshop
56:30 Closing thoughts and Question to Ponder
--- [ 2. LINKS ] ---
--- [ 3. CIRCLE ] ---
Join our private community for in-house service design professionals.
https://servicedesignshow.com/circle
--- [4. FIND THE SHOW ON] ---
The AI-powered Podcast Player
