From blueprint to bot: Designing resilient AI-powered services
Summary
Our goal was to introduce an AI-powered voice agent to help customers of a B2B building materials supplier manage their orders. But making it work meant looking beyond the tech and working through layers of complexity like siloed systems, long-established processes, and both staff and customers used to traditional, human-led interactions. This case study shares how service design helped us make that shift by: 1. Helping front-line staff see AI as a collaborator 2. Preparing customers for a new kind of interaction at different stages of their customer journey by involving sales and customer service agents 3. A data driven approach to aligning internal processes, systems and workflows
Key Insights
-
•
82% of companies worldwide are using or exploring AI due to competitive and investor pressure.
-
•
AI failures are unpredictable because AI models hallucinate, so designing for resilience is essential.
-
•
C-suite leaders often expect flawless, scalable AI impact but must see AI rollout as a learning pilot.
-
•
Service design can untangle organizational complexity by aligning strategy, orchestration, and implementation layers.
-
•
Orchestration teams translate high-level AI strategy into actionable workflows to anticipate downstream effects.
-
•
Product teams must design fallback mechanisms for predictable failures and iterate based on real-world AI errors.
-
•
Involving frontline staff as co-creators reduces AI anxiety and improves AI adoption and fallback design.
-
•
Continuous feedback loops across organizational layers enable timely learning and more resilient AI services.
-
•
Framing AI deployment as an experiment or pilot aligns stakeholders to expect failures and focus on learning.
-
•
Performance metrics should go beyond containment rates to reflect customer experience and resilience over time.
Notable Quotes
"With AI, we don’t want to save a few minutes here and there. We want something that has a large impact and can scale really fast."
"AI models hallucinate when they fail, they fail unpredictably."
"Every feature release with AI is really an experiment, not a finished feature."
"If you only design for the ideal experience, you might panic and hire back human help when things don’t go as planned."
"Service designers can help the organization untangle complexity across layers so they collaborate effectively."
"Frontline staff should feel ownership, not fear, around AI implementation."
"Involving frontline staff as co-creators helps ensure AI is a collaborator rather than a threat."
"We must build our service vision not just for the ideal, but for how to handle things when they go wrong."
"It’s important to establish feedback loops so each organizational layer learns from the others continuously."
"Framing AI introduction as a pilot helps stakeholders understand it’s a learning experience to build resilience."
Or choose a question:
More Videos
"AI won’t transform all things at once; it’s about piloting and delivering discrete business value first."
Jon Fukuda Amy Evans Ignacio Martinez Joe MeersmanThe Big Question about Innovation: A Panel Discussion
September 25, 2024
"We move beyond checklist thinking to focus on real needs of real people."
Sam ProulxAccessibility: An Opportunity to Innovate
March 9, 2022
"Nothing about us without us — cultural content should be created and tested with people who represent that culture."
Anna Avrekh Amy Jiménez Márquez Morgan C. Ramsey Catarina TsangDiversity In and For Design: Building Conscious Diversity in Design and Research
June 9, 2021
"Don't name your bot Bob, because it's not alive."
Greg NudelmanDesigning Conversational Interfaces
November 14, 2019
"Our tool is not partial to Sketch — we support Adobe XD and will soon support Figma, enabling flexible integration in design workflows."
George Abraham Stefan IvanovDesign Systems To-Go: Reimagining Developer Handoff, and Introducing App Builder (Part 2)
October 1, 2021
"Synthetic data created by AI—like fake personas and journeys—is super derivative and often not insightful."
Shipra KayanMake your research synthesis speedy and more collaborative using a canvas
January 24, 2025
"If something is only visible on hover, it either needs to be visible to screen readers at all times or have an alternative way to access it."
Sam ProulxDesigning For Screen Readers: Understanding the Mental Models and Techniques of Real Users
December 10, 2021
"We moved from Evernote to spreadsheets, to a giant affinity board in Mirror, then to UserVoice, and finally to idiomatic software with AI."
Shipra KayanHow we Built a VoC (Voice of the Customer) Practice at Upwork from the Ground Up
September 30, 2021
"The environment can disrupt critical interactions, like when Stacey gets caught up in cords and takes her eyes off the patient."
Dane DeSutterKeeping the Body in Mind: What Gestures and Embodied Actions Tell You That Users May Not
March 26, 2024
Latest Books All books
Dig deeper with the Rosenbot
What are effective behaviors that design operations teams should focus on to deliver value and build trust?
How does including metadata like methodology, market, and date improve the quality of research repository search results?
How can a product vision avoid becoming just a pretty prototype with no strategic impact?