Recap by Nerissa Sardi, recent VP of Marketing, Medici

Led by: Gabi Zijderveld, CMO, Affectiva; Dawn Anderson, VP Marketing, CureMetrix; Leanne Marshall, CMO, Yoti

Nerissa Sardi
Gabi Zijderveld
Dawn Anderson
Leanne Marshall

Wading Through the Hype

There is a lot of hype, B.S. and clutter out there with companies saying they have AI, deep learning, machine learning, and computational metrics. There is also a lot of concern and skepticism surrounding AI because of personalization; Bigger companies are using it in ways that might be perceived as underhanded.

Dawn from CureMetrics provided the example from her company which uses AI to help radiologists better prioritize the 40 million mammograms they review each year. Previously, her company was using AI to predict sales opportunities and hospital staffing needs.

With so many companies integrating AI into their products and processes, how should we look at marketing it?

There is also a lot of corporate pressure for marketers to talk about AI and machine learning because it is such a hot topic. There is a concern that corporate analysts and leadership are enamored with the idea, but end users generally don’t care about it or don’t see its value. For example, Centene uses AI for back-end claims and processing, but it doesn’t have direct benefit for the customers.

Dawn Anderson-  If we are using AI capabilities in our marketing, we have to understand and be interested in using it to improve the experience for the direct-user.

Finding the Right AI Solutions

Several big vendors offer AI solutions to companies. Vendors that roundtable participants have used include:

  • GumGum (for agencies)
  • IBM Watson (with mixed ratings from roundtable members in the room)
  • Microsoft AI
  • Alexa
  • Live Person
  • HighSpot

In terms of measurement and tracking, several roundtable members are using vendors like:

It’s a very saturated space with chatbots and virtual assistants currently taking up a lot of attention (example, for customer service.)

Before securing an AI vendor, first, consider how the solution will be helpful, then plan for governance and outcomes. For example, most companies are pursuing cost-savings by implementing back-end machine learning.

Specific Use Cases:

  • ZIP Recruiter uses tons of AI inside their company and in their customer-facing products.
  • Spoke offers healthcare communications software for companies.
  • One CMO has partnered with a UK university to develop back-end machine-learning models for cost-savings. And another is discussing a collaboration with MIT on machine learning for recruiting pipelines. Academic collaboration and partnerships help solve complex niche problems and make outcomes more affordable and faster for start-ups. Such alliances also provide credibility for your solutions and may provide for future customers, based on successful collaborations.

Fixing the Bias in AI & Machine Learning

People mistakenly think that there is a ‘general’ AI out there, but this is not the case. AI needs to be highly specialized and with specific data developed for it.

Gabi Zijderveld highlighted the need for combatting the systematic bias that is endemic the development of machine learning, highlighting the work that Affectiva has been doing in cataloging eight million faces across the globe, continuously adding representative data to improve accuracy.

Gabi Zijderveld– “There are many examples of biased machine learning out there, including how AI developed for Law Enforcement is proving to be very inaccurate in identifying people of color. People build what they know. If your AI is built by a 20-year old in California, or cataloged by a small team in Ukraine, then you will get a biased end-result.”

Dawn Anderson highlighted how breast density in women can present itself differently in different countries or ethnicities.

Dawn Anderson– “If the range of options is not programmed into the machine learning, the outcomes can be severe. In the process of developing AI, be sure to create a lot of data models to represent the many different types of users you may have.”

Speed vs. Best Practice

There is a lot of legitimate concern about the quality of AI, and the speed at which companies are implementing it in the absence of good best practice, regulatory oversight, and transparency. Marketers need to think about the permission requirements for GDPR, as just one example. If your company is utilizing end-user AI data, then you must be transparent about how it is being used.


It is still the early days in the implementation of AI. CMOs should be curious about how the brand shows within the AI journey. There is a dark side of AI, and concerns for how it might be used against us in the future. There is an immediate need for best practices, regulations and alleviating bias.

At the same time, there are lots of positive things about the future of AI. Marketers can leverage it by highlighting the positive benefits for customers (end-users,) but need to be prepared to answer questions.