Getting Started with AI

At a recent meeting focused on AI and Automation for Customer Engagement, we had the opportunity to explore how the Predictive Customer Engagement model has evolved over the years, and how we might apply it to the emerging artificial intelligence landscape.

One topic we covered was a way to understand the structure of an AI implementation, including

  • a data repository, where data gets collected and accessed,
  • machine learning techniques, where we apply these technologies,
  • capabilities, where we analyze what we learn from applying AI,
  • and visualizations, where we communicate what we’ve learned.

This brief video explains a little more.

We then reviewed and updated a framework started at an Open Space session convened by Bonnie Chase in 2018 that focused on getting started with AI. Much of the structure was still relevant, and Members offered additional current advice on how to get started.

Start small, be clear about the use cases,
be prepared to learn, take incremental measures. 

Two sentence summary from “How to Bring AI Into the Company” Open Space session, August 2018

Some Members are hearing, “We should do AI!” in their organizations, which can lend itself to a solution chasing a problem. What follows is the advice of Members who are approaching the AI landscape from a use-case-first perspective.

  • Leverage formal change management practices and the structure of a KCS adoption (plan and design, then adopt in waves, then build proficiency, then optimize and innovate) to adopt AI/ML. Incorporate trust.
  • Technology and capabilities are changing rapidly! In the spirit of “just because we can, doesn’t mean we should,” make every effort to adopt at the speed of absorption into the organization, not at the speed of what’s becoming available.
  • Start with concrete, objective use cases. This allows for building understanding and consensus within the organization, which you need to secure business buy-in.

Success requires defining clear objectives, securing business buy in, and obtaining proper resources.

Define Clear Objective(s)

What’s your use case? Consortium Members have access to a whole list of example use cases, but suggested categories include:

  • How does this help customers be successful?
    • Improve/remove/streamline existing processes
    • Pattern recognition/healing defects before they’re customer-reported
    • AI offerings as a product
  • How does this help employees be successful?
    • Improve/remove/streamline existing processes
    • Act as co-pilot for problem solving
    • Pattern recognition/bug detection/script-writing
  • For an example of how Members used a large language model to analyze data collected in an empathy mapping exercise, see Sara’s post: ChatGPT is Our Latest Collaborator, Not a Job-Stealer

Secure Business Buy-In

Defining clear objectives enables you to communicate how AI will help solve the problem, and allows you to tie the value of the objective to existing business goals.

  • Agree on measures and baselines
    • Data points from examples of success are always helpful
  • Set expectations and timelines
    • What are the requirements the project must meet?
    • Involve stakeholders in building a communications plan.
      • What kinds of visualizations are we expecting from this project? Do they need to be different for different stakeholders?

Obtain Proper Resources


  • Minimum to start: someone with a clear understanding of the objective and someone who can speak data science.
  • Who is responsible for which layers: data repository, machine learning techniques, capabilities, visualizations?
  • Security: who is accessing what data and how?
  • How will you collect feedback, and who will be monitoring and acting on it?
  • Oversight and risk management: who watches for, reports on, and/or fixes unintended consequences?


  • Starting with clear objectives helps avoid “we have a really cool tool; what could we throw it at?” What AI/ML techniques specifically address your objective(s)?
  • Proof of concept experiments are great, but keep an eye on and plan for the ability to scale.
  • AI/ML technology will need access to the relevant data repository. How does any given piece of technology deal with data security? Privacy and GDPR? Biases? Multiple languages?

As always, these frameworks are a work in progress. Let us know what’s missing here, and let us know how you’re playing with all of the machine learning and artificial intelligence now at our fingertips!

If you watched the above video all the way through, you might want to know more about pumpkin toadlets and the “What’s Your Use Case?” t-shirt.

Leave a Reply

Your email address will not be published. Required fields are marked *