The more sophisticated we become in using AI about people, the more human we need to become.
Earlier in the year, I sold an AI algorithm for consumer behavior and decision intelligence. It’s my third algorithm/AI acquisition, and each time the challenge is the same. There is a great underestimation of bias.
Technology magnifies the quality of the people, processes and data behind them. The more sophisticated we become in using AI about people, the more human we need to become.
CX professionals and leaders can put their energy into creativity, imagination and strategy with better use of AI/data. So, let’s look to manage our biases and use the opportunity to develop advantages from them!
These are regular biases I’ve found to create decisioning problems in insight, segmentation and orchestration processes. Remember, awareness is your strongest asset.
Awareness: The most common bias. It tends to interpret new information as confirmation of our existing theories.
Fix: Overcoming confirmation bias begins by using more quantitative analysis. Also, engage an objective perspective to handle and interpret your data from your engagement efforts. Be comfortable and appreciate the people who can oppose your perspectives and respect you.
Advantage: This is an opportunity to create a stronger team dynamic when making decisions. Do this through:
- Developing a good data culture.
- Leveraging team assessments to understand decision-makers.
- Creating a diverse team or collective perspectives. Nosce te Ipsum. (Know yourself)
Related Article: ChatGPT’s Impact on Customer Experience and Marketing
Awareness: Common with investors, entrepreneurs and high-growth marketing leaders. The tendency to dismiss or ignore new research evidence which overrides or undermines an existing decision. A leader so invested in not losing a CX initiative completely ignores the data illustrating the pitfall.
Fix: Be open to abandoning the deep investment. Also, become as objective as possible when considering new research. Define a deadline or have a research or advisory support group help mitigate this effect.
Advantage: Leverage parachute metrics which show you velocity changes. Examples include an accelerated drop in customer satisfaction, campaign engagement or top-of-funnel inquiries.
Placing yourself on notice before your stakeholders do provides a buffer. You can then make sound decisions and better prepare narratives. Top startup founders and enterprise CEOs use these metrics to manage up and down. Failure is part of business and marketing, but how it is handled makes or breaks careers.
Overfitting and Underfitting
Awareness: Overfitting involves an overly complex model which fits the data too well. Many Go-to-Market leaders will recognize this when their AI or statistical models applied to new data sets are consistent but inaccurate on average. Example: Nearly every prospect meets some “qualified” criteria, yet it is incorrect.
Underfitting occurs when a model or algorithm cannot capture the underlying trend of the data. The model is too simple and does not reflect the data well enough. Using the previous example, an underfit model/algorithm produces an occasionally correct qualified prospect.
Fix: Solve overfitting by splitting up data sets, say training vs. testing. Also, cross-validate over such multiple sets. Solve underfitting by incorporating confidence levels and monitoring them.
Advantage: Neither is useful. But they can be great discussion points with your martech, sales tech, and customer data vendors with intelligence capabilities.
Awareness: Outliers are data points significantly above or below the norm or outside the pattern. Relying on such numbers at face value may not paint an accurate picture. Busy leaders and especially those leading CX across an organization have a tough job of translating conversations analog to digital and back again. Outliers are not to be dismissed if they are your customers and are treated equally.
Fix: Tread carefully. Ignoring outliers is sensible in some situations and completely irresponsible in others. Have a complete understanding of the context of the problem. Also, overlay them with any financial, ethical and social constraints. Ensure your team is also aware.
Advantage: Studying these reinforce your team’s purpose. Use it to improve processes and feedback for the model/algorithm. Statistical outliers can uncover new opportunities or the need to reinvest in campaigns.
Related Article: Can AI Marketing Transform Your Business?
The Best of AI Is the Best of You
With good preparation and regular context checks, your organization needs to:
- Develop awareness, education and training plans as great first steps.
- Build teams with diverse perspectives.
- Ensure the inclusion of thoughts, ideas and constructive comments.
- Strive for transparency in the process and models/algorithms.
CX & GTM executives managing these teams should immerse themselves in these conversations. Also, take the opportunity to work with data. These leadership investments build upon the AI and data culture you want to reflect and scale.
Our collection, interpretation and AI application of data will magnify who we are and how we operate. Awareness of biases and iteratively improving upon them will help leaders feel more confident in their marketing AI-driven decision-making support.
We use data to empower us and AI to scale us.