Explainable AI & the boardroom conversation

  • Danny Coleman
  • June 19, 2020

AI is a somewhat underspecified term and can mean many things to different audiences.

However, in business, boardroom execs today are still seeking answers to what AI will truly deliver to their business. Many have spent millions of dollars, yet few have seen significant real-world impact and ROI.

One exec recently remarked “In the current climate, I will only sponsor AI investments that deliver action and impact. Seeing insights before going into production is what will truly move the dial for AI. That way, we as business leaders can make informed decisions based on data that is relevant to our business objectives”.

The concerns and questions we often hear are the same regardless of which industry exec you happen to share time with. Here are some salient points of view:

  • What is our AI maturity curve compared to our competitors?
  • Will regulation hinder our opportunities to scale AI?
  • Bias is prevalent in humans. Can AI really solve this or provide better insight?
  • Implementing AI without knowing the outcome is high risk
  • How do we ensure the AI in production today remains relevant in the future?
  • Simplicity of using AI is essential and not to be blinded by the science
  • Most AI vendors lock you in – we cannot risk being tied to one vendor
  • We seek AI offerings that scale and not point solutions
  • We need software and services that business users can execute
  • Creating AI assets has huge upside – how do we constantly evaluate them?

At Chatterbox Labs we’ve built our XAI platform around 4 pillars:

  • Explain
    • We reverse engineer any text, mixed (numerical & categorical) image AI models to explain outcomes without the need for training data.
  • Trace
    • We can trace any AI model to ensure there is no decay or drift in the AI model. This process takes place pre and post production to ensure AI model relevancy.
  • Action
    • This process was built to ensure you can action upon the pertinent data outcomes of any AI model to ensure business relevancy.
  • Fairness
    • To explain and action bias for every datapoint. Identify systemic unwanted bias via data, model and action in a fairness scoring system.

We believe this approach will answer, validate and alleviate concerns boardroom execs have about their current and future AI investments.

If you would like to find out more, please get in touch below.

Back to blog

Get in Touch