Summary
This talk is a case presentation about using generative AI and graph languages to come up rapidly with complex enterprise designs. We are using a repository based enterprise architecture tool and EDGY, an open source visual language, to feed GPT4 with context-rich queries. The resulting maps and models are ... wrong. But they have proven to be inspiring or even triggering for conversations across a diverse stakeholder community, and shortcut our way to a set of correct and useful models that inform design decisions. Moreover they can highlight blind spots and interrelationships previously unknown and thereby enrich the design process with minimal effort. Takeaways Recognising blank page moments in complex challenges How to embed context and an ad hoc Training in an LLM prompt How to make generate a web of coherent maps such as Journey, JTBD, Organization, Process Maps that cover a complete design related to a given challenge How to use these maps and how not to use them when co-creating with others When to keep tackling the blank page yourself instead
Key Insights
-
•
Designers often face the 'blank page' problem when tackling complex enterprise domains without sufficient research or domain knowledge.
-
•
Enterprise design requires making the invisible 'dark matter'—complex social, technical, and political systems—visible and understandable.
-
•
EDGY is an open source graph language that models enterprise complexity across three facets: customer experience, enterprise architecture, and organizational identity.
-
•
AI, particularly GPT-4, can augment designers by consuming structured repositories as context, reducing reliance on generic internet knowledge.
-
•
Using AI in enterprise design is more effective as augmentation of human creativity rather than automation or replacement.
-
•
Three AI collaboration modes identified are the 'jester' for provocative incorrect ideas, the 'sidekick' that integrates research context, and the 'matrix' that transforms knowledge across enterprise perspectives.
-
•
AI-generated diagrams are imperfect but valuable as starting points to defeat the blank page and stimulate discussion.
-
•
Inputting domain-specific research and insights into the AI prompt improves the relevance and quality of generated task maps.
-
•
Biases can be intentionally introduced in AI prompts to explore specific enterprise perspectives, such as organization design or brand identity.
-
•
Enterprise design must bridge silos by fostering knowledge sharing and cross-team collaboration, which AI and graph languages can facilitate.
Notable Quotes
"Designers are thrown into environments with experts that have decades of experience, and we have to come up with something relevant really quickly."
"The blank page can be terrifying because you don’t want stakeholders to laugh or say you got it all wrong."
"Behind simple tasks, there’s often big complexity hiding, especially in enterprise environments."
"We think the problem of 'dark matter' in enterprises is about making invisible things visible and bridging silos."
"AI is better seen as augmentation to make us more efficient, not automation to replace us."
"We use our model repository as context to help AI generate answers based on our specific enterprise knowledge rather than general knowledge."
"The jester mode throws something in your face, maybe wrong or uncomfortable, but at least it moves the conversation forward."
"AI-generated diagrams are usually not perfect, but they offer a better starting point than a blank page."
"If you say, ‘let’s look from an organization design perspective,’ AI reproduces the bias of that perspective in its output."
"Enterprise design is about sharing knowledge, analyzing it, improving it, and then consuming it to create better outcomes."
Or choose a question:
More Videos
"Citizen participation is like eating spinach—everyone agrees it’s good, but few actually do it meaningfully."
Aaron Stienstra Lashanda HodgeLeveraging Civic Design to Advance Equity and Rebuild Trust in the US Federal Government
December 8, 2021
"With AI tools, you have to come back and reprompt, it’s not a one-shot operation."
John DonmoyerShipping your code generation experiments to production
June 11, 2025
"Mapping is not just for presentation; it’s a practice to evolve your understanding of a complex problem space."
Sheryl Cababa Alexis OhThinking in systems to address climate with Sheryl Cababa
June 12, 2024
"It would have been really easy for the program manager for street homelessness to say no, but instead she gave us access."
Ariel KennanBuilding a Design Culture
June 9, 2017
"Our evaluation tool is a Google Sheet so we can quickly make adjustments without being bogged down by unnecessary features."
Ignacio MartinezFair and Effective Designer Evaluation
September 25, 2024
"You can’t just look at who wants research and what they want it for; you have to peel apart organizational processes to see if action is possible."
Dave HoraAdvice for Establishing Research
December 8, 2022
"Hearing a participant say thank you for listening is one of the most powerful indicators of meaningful research."
Dr. Jamika D. Burge Mansi GuptaAdvancing the Inclusion of Womxn in Research Practices
September 15, 2022
"If you’re concerned about cybersecurity, buy a used car with fewer electronic features or one that has a strong reliability score."
James RamptonThe Basics of Automotive UX & Why Phones Are a Part of That Future
July 25, 2024
"The state of IT is a daily reflection of what the company thinks and feels about its employees."
Kristin WisnewskiMeasuring What Matters
October 23, 2019
Latest Books All books
Dig deeper with the Rosenbot
How does including metadata like methodology, market, and date improve the quality of research repository search results?
How can play help overcome the ‘messy middle’ between research insights and decision-making in organizations?
What is the evolving role of researchers of one in organizations practicing democratized research?