Design Thinking with AI

Nov 27, 2017

3 min read

Design Thinking workshop in progress: four team members stand at a wall covered in wireframes and sticky notes, writing on a whiteboard under blue lighting. Handwritten labels include 'Interactions' and notes on natural language processing and analyst behavior.

Design Thinking with AI

I want to share some lessons around how my team has adapted IBM Design Thinking to fit our needs and fulfill our mission developing AI prototypes and tackling SWAT opportunities. I’ve led & facilitated IBM Design Thinking workshops for executive-level clients, senior Watson leadership, IBM teams and external agencies. These workshops have led to co-creating user-focused experiences with clients that deliver impactful Watson integrations that solve a need for our users and have added capabilities for IBM Watson. During these workshops, I’ve identified a workflow that my team has used to design Watson Proof of Concept experiences and MVP prototypes that we’ve built leveraging the outputs of the exercises.

Looking beyond needs alone

The key to any successful Watson experience is making sure the objective based on user need — or data insight. This focus allows us to align core capabilities around the application that meet a necessity, not functionality. In the AI age, we need to move beyond needs alone — we have to develop trust and empathy within our experience to create a cooperative relationship between Watson and our users.

Hand-drawn experience-based roadmap on paper, annotated with white circles highlighting Proof of Concept, Phase 2, and Phase 3 milestones. Each phase labels Watson services used: Conversational Agent, Watson Discovery Service, Watson Knowledge Studio, Command and Control, and NLU.

Identifying a Proof of Concept using Experience Based Roadmaps to identify key milestones and Watson integrations

Integrating Watson into Design Thinking

Watson can be incorporated into every Design Thinking exercise — but I’ve found it used best, with least influence over the concept, when integrated into activities after the initial brainstorming. That focuses our central thinking on our users — afterward we focus on determining how we can leverage Watson to meet those needs. We include client stakeholders & leads, Watson architects, data scientists, and developers, as well as our stakeholders and design leads. These folks work together to fulfill this part — I find it best to let our clients prioritize the direction on their own, and then we work on integrating Watson collectively.

Prioritization matrix plotting Watson services by market readiness against feasibility. Grouped clusters show progression from APIs to Data and Insights to User-Centric Experience, with services like Watson Multimedia, KnowledgeGraph APIs, People Insights, and Watson Conversation mapped across the quadrants.

Identifying our core experience using prioritization and designing data and insight we can find with Watson.

Forming a hypothesis

Our team’s goal walking out of any Design Thinking session with Watson is a precise definition of the experience — a Hill or commander’s intent matched with data and insights that form the building blocks of our application. We can take the output of the Design Thinking sessions and focus on what we discovered. The ‘what’ of our Hill statement is the seed basis for our data hypothesis — which will leverage the data we identified or deliver the insights we outlined.

Typographic slide on black background reading 'Hill statements seed your AI hypothesis' with 'AI hypothesis' set in blue.

Hill statement and AI hypothesis example for a coffee shop app: 'Jacob, a coffee enthusiast, sees his local store come to life based on the combined interests of the people inside, forging a personal community connection and promoting civic engagement, by merely opening his app.' Paired hypothesis reads: 'If we can ingest our coffee shops' customer data, purchasing habits and social interests than we can predict trends and shared human values based on a physical stores' location creating digital personas of physical spaces.'

Hill statement and AI hypothesis example for car insurance: 'Deanna, a first-time car buyer, can buy the perfect car insurance for themselves quickly and as painlessly possible, just by having a conversation.' Paired hypothesis reads: 'If we integrate microservices into a Watson conversation with improved relevance and usefulness than we'll see a measurable reduction in time and frustration due to an increase in ease of use validated by user testing.'

Hill statement and AI hypothesis example for a financial analyst: 'Hilary, a financial analyst, is empowered to predict market changes and volatile swings in company sentiment, as they happen.' Paired hypothesis reads: 'If we can ingest social data, press releases, quarterly calls, and traditional media we can predict a change in stock price based on a company's social sentiment trend.'

Prototyping with Watson

obile app wireframe prototype for a coffee store experience powered by Watson. Screen shows store details at top, a 'Coffee Store Stir' section displaying 'The Life Changing Magic of Tidying Up' by Marie Kondo with user avatars and a recommendation: 'You've read other philosophical novels. Check it out.' Blue annotation icons call out Watson integration points for user profiling, content recommendations, and data connections.

Using the Hill statement we defined in our workshop sessions, we start to prototype an experience built around the Hill’s core tenets. Functionality, technology, and AI decision are based entirely on achieving our hypothesis in context to our user.

This approach is appropriate for prototypes of all fidelities and levels of interaction. It gives our integrated teams common language and syntax to work with and makes the experience tangible for high-level client executives.

For every interaction, touchpoint, and pivot in our prototype we have to:

define what our user is doing
what data sources they’re accessing
what APIs Watson is using
how is it analyzing that data
how we make that data accessible
how that info helps our user make a decision and/or move forward

Watson sentiment analysis dashboard displayed on a laptop showing a 6-month trend line of positive and negative media mentions. Below, individual mention cards from Reuters and Globe and Mail show 67% positive sentiment scores. Left sidebar displays 31% positive trend across 1,935 total mentions with breakdown: 600 positive, 316 negative, 1,019 neutral. Annotated with labels: 'Hill Statement What — Visualized trend' and 'Hypothesis Proof — Visualized evidence.

Visualizing your data insights

Combining all we learned during the Design Thinking process and the prototype phase allows us to develop the right visuals that add clarity for our users but also demonstrate what Watson is ‘doing,’ which adds trust to the relationship.

So, how do you prove your hypothesis visually?

Your experience may not require a data visualization but using the same thought process will help you ‘prove’ the decisions Watson made (i.e. ‘why is this recommended to me?’).

IBM Design Thinking infinity loop graphic on black background, rendered as a dark gray figure-eight path with a blue dot at the center crossing point and a white dot on the upper right curve, representing the continuous cycle of observation, reflection, and making.

Validating and iterating

User testing will validate your experience against their needs which will map directly back to the Hill statement from the initial exploration. Validating Watson performance should be carried out with the user in mind — is this data valid, did it prove or disprove our hypothesis? Be iterative toward all phases of your design — a second aspect of data enrichment may prove your hypothesis and meet your user’s needs.

ypographic slide reading 'IDT elevates AI and creates real insights for users' with 'real insights for users' in blue. Right side shows a Venn diagram of three overlapping circles labeled User, IDT, and Watson, with arrows indicating the cyclical relationship between them.

IBM Design Thinking elevates AI & creates real insights for users

—

This private publication is not affiliated with my employers or professional associations. Personal blog, personal opinions. Not speaking for anyone but myself. ✌️

Back
←

A group of individuals collaborates on design concepts by sketching and placing notes on a large, illuminated board, demonstrating teamwork and creativity in a workspace setting. Rob Harrigan Design. Rob Harrigan.

Design Thinking with AI