Kevin Buehler

Machine learning teams thrive with continuous education. Fortunately, there are incredible resources available. In 2017, I began diving into Andrew Ng's classic Machine Learning course offered by Stanford. I took it alongside a McKinsey partner who is now a Chief Analytics Officer. I coded my own back-propagation algorithm and got it working over a couple of weekends. Encourage your machine learning team members to define their own learning goals. Get them to try courses from deeplearning.ai, Coursera, EdX, Udacity or one of the many other online offerings. Have them form teams to participate in Kaggle contests. Run your own hackathon. Make it fun, free and a bit of a friendly competition (gamification works). Keep learning!

Let me disagree with the premise. In computer vision, the rise of deep learning models has shifted the dominant approach from traditional feature engineering to automatic feature extraction. Today's top-performing computer vision models, such as Convolutional Neural Networks (CNNs), Generative Adversarial Networks (GANs), and Vision Transformers (ViTs), can automatically learn hierarchical representations directly from large datasets. These models demonstrate that with sufficient data and computational power, it's possible to train computer vision systems that learn to identify the most relevant features for a given task on their own, leading to breakthroughs in tasks such as image classification and object detection.

Consider the team itself. “Assess each team member using the skill/will matrix. For those low on skill but high on will, invest in training. If the will is lacking, consider if their future lies outside the team.” [Response courtesy of GPT-4, after I asked it to consider the struggling team itself as well.] And maybe we can augment the team: “Imagine a team member that works tirelessly, spotting trends humans might miss. That's your GenAI model. Its role? To augment your team's efforts, offering data-driven insights and freeing up human members to focus on creative solutions and strategy." [Also courtesy of GPT-4. I did ask it to be a bit edgy.]

I agree with Dan Lieb’s and Andy McMahon's take. Machine learning has its place, but it's not the panacea for issues of team morale and engagement. Those are fundamentally human issues that require a human touch — communication, empathy, and understanding. Instead of looking to algorithms, we should invest time in really listening to our team members, understanding their challenges, and crafting an environment that fosters real connection and collaboration. Technology can support these efforts, but it can't replace the essential human element in leadership. [Ironically, GPT-4 helped write this response. My prompt: “I think Dan and Andy have it right. Can you please write a response I can post aligned with their perspectives?”]

The most generous acts one can perform often have nothing to do with a particular consulting engagement. For an executive in a new role, offer to provide an introduction to peers in the same role in other organizations, particularly in other industries. If a client is receptive, offer to provide candid feedback as you might for a close colleague, effectively holding up a mirror for them. When an executive leaves or is let go (it happens more than you might think), be available as a sounding board and offer to connect them to others. We have built a talent network of candidates and opportunities for just this purpose, with no cost on either side. The trick in all these cases is to expect absolutely nothing in return. Be truly selfless.

When the COVID-19 pandemic started, the range of uncertainty was enormous. Try using more than one dimension when the range of uncertainty is that large. We used two dimensions to address different aspects of uncertainty: 1. Virus Spread and Public-Health Response (including the effectiveness of physical distancing and the path to vaccine availability) 2. Knock-on Effects and Public-Policy Response (including the effectiveness of fiscal and monetary policy responses). The 3 x 3 matrix that resulted spanned the range of both likely and remote outcomes. “Safeguarding our lives and our livelihoods” was published on March 23, 2020, laying out those scenarios and providing a common framework that proved useful throughout the pandemic.