Apply design thinking to AI

The AI Essentials Framework is a specific grouping of activities to work through so that your team is aligned on your strategy for an AI experience.

There are five phases to the framework:

Before that

  1. Meet the group of your project.
    1. Format: <NAME>, <JOB TITLE>, I focusing on <FIELD> about <TIME> after <START POINT>. I’m good at <YOUR SKILL>
    2. Example: <Beth>,<Data Scientist> , I focusing on <AI product development> about <3 years> after <graduated from university>. I’m good at <translating business needs into models and finding hidden insights in data>
  2. Design Research
    • Design Research the practice of inquiry and discovery that builds knowledge, insight, and empathy for your users
      • Format: <Our users> struggle to <achieve some task> today because <blockers, limitations, etc.>
      • Example: <business travelers> struggle to <check-in experience takes a long time> today because <employees don’t always know guest needs.>
  3. As-is Scenario: the current typical user experience
    1. What specific pain points ?
    2. What do you think users need ?
    3. How do you think AI could address this need ?


Align the business and user intent(s) for your solution.

1. align your team on the business
2. user’s feelings for your solution

How you might use AI to solve your users’ problems.

1. Spend some time with your team coming up with as many ideas as you can.
2. Discuss which idea will have the most value to your business and your users.
3. Move forward with that idea.

6 core AI intents

  1. Which of the 6 core AI intents do you think to apply best to your goal?
    • Accelerate research and discovery
      • Spend less time looking for the information you need and more time acting on it. Easily process millions of data points to focus on the work that matters most.
      • For example, 2.5 million scientific papers are published in English every year. When this unstructured text becomes structured, researchers spend their time looking at the most important insights.
    • Enrich your interactions
      • Train your AI to address common requests, reduce manual response times, increase the number of transactions, and make interactions more productive.
      • For example, AI can enable call center agents to better address caller needs. This helps provide a consistently positive experience and saves businesses time and money.
    • Anticipate and preempt disruptions
      • Continuously monitor the condition of your systems to mitigate problems before they disrupt your work. Use AI to catch potential issues in the systems and processes that are essential to your business.
      • One application of this is public transit maintenance. $398M is lost every year due to subway delays in New York City. Predictive maintenance can help avoid downtime for users and save money for the city.
    • Recommend with confidence
      • Teach systems the nuances of your business to ensure every factor is considered so that you can make better-informed decisions, give tailored advice, and deepen customer relationships.
      • For example, customers enjoy tools or experts that help them purchase the right product. This could be anything from shopping for a dress to choosing the right savings plan.
    • Scale expertise and learning
      • By combining expertise with your industry’s latest learnings, each of you knows as much as all of you. Make deep instructional knowledge available to everyone in your organization.
      • For example, 38% of law partners will retire in the next decade. AI can comb through legal documents to help summarize successful tactics for legal cases.
    • Detect liabilities and mitigate risk
      • Train systems to understand and keep up with constantly-changing regulations and privacy obligations. Identify compliance issues quickly and easily to protect your business and your people.
      • For example, about €20M in fines could be imposed on companies failing to comply with GDPR. With AI, analysts can help uncover compliance breaches and follow the right course of action.
  2. Reflect on the intent you just wrote down.
    • Give 1-2 concrete examples of what the project could make to bring this intention to live.
  3. Bring it back to our users: business travelers.
    • Ask yourself: Would their needs be addressed through your ideas? How? All of the sections?
  4. Vote on the best theme (aka the title of a cluster).
    • Each participant should have 2-3 votes. Have a group discussion about everyone’s reasoning behind their votes. This is more important than the votes themselves.


Hightline the data you could use to make your idea a reality.

** Data is the fuel for AI.**

Next, you’ll look at what it takes to bring your idea to life. The data activity helps document any and all data that can be used to bring your intent and a big idea to life.

How should we prepare the data?

The process of getting usable data takes hard work, and requires you to hit 2 big checkboxes:
quantity and quality.

Following the steps will help you:

  • Set a to-be scenario: to-be scenario is the ideal future user experience
Brainstorm individually, step 2
Imagine your ideas exist in reality Fill in the corresponding rows, using one sticky note per answer.
Identify the highs and lows, step 4
Compare this map to your user’s as-is and identify the various changes this To-be Scenario could bring.
  • Reflect and bring your intentions to life
    1. How much does this to-be scenario meet the user’s needs?
    2. Did you see the big idea (better than others) come through in this to-be scenario?
    3. What data be used in the to-be scenario?
  • What data be used in the to-be scenario? the data belongs to Which of the 3 core data types?
    1. The 3 core data types
      • Public data:
        • This can be found in the world (for free or for purchase). For example, census data is free and highly accessible. Anyone can go download those spreadsheets. Even social media hashtags, likes, and comments are public data. Sometimes, public data isn’t enough, though. For example, businesses regularly purchase additional demographic data for user segmentation.
      • Private data:
        • This type of data is owned and held by your business, which often provides a competitive advantage. This sometimes takes the form of employee records, hardware assets, inventory logs, payroll, and more.
      • User data:
        • User data can be linked to individual users. Think contact information, medical history, or geographic location. All of these variables belong to the user. Businesses must ask for access to store or use the data.
    2. Confirmed data status of analysis Array:
      1. analysis Array is a 3×3 Array, Row fields are HAVE, WANT/NEED, and NICE TO HAVE, the column is Public, Private, and User
Data activity overview, step 1

Understanding your Data

Determine what you will need to teach your AI.

you’ll break down each data source relevant to what your AI needs to learn.
This will help you determine what you need to teach your AI so it can understand the ins and outs of your domain.It can be

Daunting to make an AI system understand all of the things.

You have to think of your AI as a toddler and train it over time. It could take months of training for a machine to start making sense of a single data source, like a loyalty profile.

Because of the large number of data points and concepts, it’s impossible to tackle everything at once.
They must start small and iterate as they figure out the best way for their system to learn the necessary vocabulary and information.

  1. Could you break out the data set into its individual components?
  2. What’s the component in your data that an AI would need to understand?

How to broke out the data set?

  1. Find The Data Sources
    • Data Sources could be a Database from public, or private data, or some action of your system. find all you need and set those as the center of the circle.
  2. Deconstruct your data sources
    • Think of all of the individual components that make up the data source. Place them around your original data source. These components will directly inform your AI models and their training.
  3. Discuss with your team
    • Determine if the individual components have jargon you need to define.
    • Check your work as a team to make sure the data components match correctly to each data source.
    • Talk through the implications if a machine makes associations between the stickies you posted (e.g. different behavior based on gender, race, and nationality).
    • Thinking about Who will be curating the data and training these models.
Discuss with your team, step 4

Reasoning your Idea

Bring your ideas down to earth.
Descript the system How to understands the data you feeds and how our system is going to reach conclusions?

  • A reasoning statement to answer How will your AI use what it knows to meet your intent?
    • Format: <Your Business> can <intent> by <big idea> based on <the AI’s data and understanding>.
    • Example: <School> can <offer a personalized experience> by <student’s homework with custom question> based on <last score, study preferences, and the personality for they have.>.
Review and identify phases, step 4


Brainstorm the direct and indirect effects of your AI.

Our goal here is really to design a relationship with our users, a healthy relationship.

This is a process and mindset that never ends. You need to have a plan for growing and maintaining the relationship between your users and your AI over time.

  • Layers of Effect
    1. Primary
      • These are the effects that you think of first when you think of a product or experience. Primary effects are pitched as the heart of a product.
        • For example, Yelp’s primary effects is that it is a platform for people to review and learn about establishments.
      • Primary effects may evolve over the course of time, but they typically do not change too drastically, especially if they have already set off and made an impression on the market.
      • They are always intended and known. these effects are confined in the innermost, opaque box to highlight that they are known and intentional.
    1. Secondary
      • These are the effects that might not pop up immediately as the defining characteristic of a product but are still just as relevant to the company’s shareholders.
        • For Twitter, while its primary effect might be serving as a social media platform, a secondary effect would be how it acts as an ad revenue generator.
      • These effects are confined in the outermost, opaque, coral box to highlight that they are also known and intentional.
    2. Tertiary
      • These are the effects that are either unintended or unforeseen. These can be good or bad; in any case, tertiary effects are always surprises that start cropping up after users have had their hands on the product.
        • For example, a tertiary effect of Facebook would be the role it played in perpetuating the spread of fake news due to its algorithm.
        • A tertiary effect of Twitter is the massive trolling and cyberbullying (as well as the ensuing emotional trauma) that happens on its platform.
      • In the above graphic, the tertiary circle is feathered out, representing the fact that there can be an unlimited amount of tertiary effects. They aren’t quite the same as edge cases, although there is some overlap;
      • tertiary effects are not always the result of extreme parameters. Those with the privilege of creating products have the responsibility of defining ethical primary and secondary effects, as well as forecasting tertiary effects to ensure that they pose no significant harm.
  • 5 AI ethics focus areas
    • Accountability
      • Every person involved in the creation of AI at any step is accountable for considering the system’s impact on the world, as are the companies invested in its development.
    • Value alignment
      • AI should be designed to align with the norms and values of your user group in mind.
    • Explainability
      • AI should be designed for humans to easily perceive, detect, and understand its decision process. Imperceptible AI is not ethical AI.
    • Fairness
      • AI must be designed to minimize bias and promote inclusive representation.
    • User data rights
      • AI must be designed to protect user data and preserve the user’s power over access and uses.
One Comment

Add a Comment