How the AI Canvas helps companies develop their artificial intelligence roadmap

If you are new to AI, I bet you are struggling (or will) with where to start.

If you are just getting started, odds are it’s not going so hot. It could be that your models are not providing measurable value. Or maybe they never worked in the first place. But wait! The benefits of AI are well-documented, right? What’s going on here?

Most companies task their data scientists and/or engineers to explore models. I mean, why not? It’s an obvious place to start. But if there’s a disconnects between business needs, data science projects, and the people who use intelligence, you risk solving the wrong problem with models that will never be used. So how do you know what problems should be solved with AI? How do know where to start?

 

The AI Canvas

Canvases are not new. There are Lean  Canvases, Business Model Canvases and many more. Their purpose is simple: stop wasting time on ideas that get stuck. Canvases help you:

 

  1. Solve the right problem
  2. Create solutions to complex problems
  3. Identify where to start
  4. Foster alignment between different stakeholders
  5. Move quickly

 

The AI Canvas is a tool to help business align on strategy, identify data requirements, sketch ideas, and define how to use artificial intelligence in the business immediately.

Example: Voltage Control AI Canvas

 

AI Readiness Canvas

 

The AI canvas is filled out through a series of exercises throughout a 2-4 day workshop.

Workshops should include various individuals that have stake in the success of any AI program. The must haves are; executive sponsor, data leader, technology leader, and those closest to your customer. The goal is to create alignment between different groups of people, accounting for the perspectives of those different people.

 

KUNGFU.AI Practical AI Workshop

 

Practical AI Workshop

Objectives of the workshop:

 

  • Understand business goals, challenges, and strategy
  • Create an AI strategy that supports the business strategy and guides all AI development
  • Define measures of success
  • Identify data wealth, dependencies, and constraints
  • Create a roadmap of AI use cases most relevant to the business needs and data available
  • Find data augmentation opportunities
  • Identify technology and cultural risks
  • Define the first project that drives most immediate value

The AI Canvas is an extremely useful tool get started on AI projects quickly while reducing risk of wasted time and effort. You also get a long-term strategy for future use cases and identify data gaps. This part is key! As you build your first AI capability, you also start collecting the right data now for AI capabilities down the road.

The AI Canvas for Baxter Planning.

Baxter Planning is an Austin-based software company who provides service inventory planning and optimization solutions that support service supply chain requirements across diverse industries. They’ve been recognized as innovators in the planning space for over 25 years.

The next frontier for planning innovation is using artificial intelligence to do everything from predict product demand to automate planning tasks. And Baxter Planning thrives by remaining on the forefront.

We at KUNGFU.AI teamed up with Voltage Control to conduct a workshop using the AI canvas. The goal was simple. We needed to identify where Baxter Planning should adopt AI that would drive the most customer value.

We had an opportunity bring together business leaders, data leaders, and technology leaders who don’t typically get a chance to collaborate. And through their perspectives something very interesting happened. While we did find a starting point for AI product enhancement, we also identified some critical data gaps.

This insight led to an important conclusion. We must create a plan to acquire the data necessary to create the capability. And based on the data collection strategy, we can build an iterative AI product roadmap, selecting techniques that can work on limited data, while building more advanced techniques as more robust data become available.

This insight was big. The net result would save wasted time and effort experimenting with AI tools that would never achieve levels of accuracy and provide value.

You never know what you will unearth in these workshops. But a no go conclusion is still a positive result. Instead of starting work on ideas that will eventually get stuck, we can identify constraints up front, come up with a game plan to overcome the constraints, and build the best product roadmap accounting for the constraints.