News

Building generative AI skills in your organization: Tips from recent publications co-authored by GOF’s George Westerman

February 15, 2024

With a daily abundance of new information on generative AI, managers and leaders may be seeking reliable and streamlined guidance on using gen AI to raise productivity.

In two recent articles published in Harvard Business Review, the founder of the GOF George Westerman and coauthors offer up-to-date practical advice on two crucial AI challenges for businesses:

(1)  Selecting gen AI tools that work for your business.

(2)  Helping train workers to adopt and maximize gen AI’s advantages.

Both challenges will likely need to be addressed simultaneously. Businesses will need new technologies and new skills. Here is a synopsis and preview of recent GOF-related work addressing these key challenges.

“How Generative AI Will Transform Knowledge Work”

By Maryam Alavi and George Westerman

Leaders may realize that their organizations could benefit from gen AI. The next step is to help employees find opportunities to use it successfully. 

Maryam Alavi (Professor of IT at George Institute of Technology) and Westerman explain how gen AI can help knowledge workers, by boosting productivity and freeing time for more meaningful aspects of work. Knowledge workers are a category of worker highly likely to be affected by gen AI in years to come.

“Knowledge work primarily involves cognitive processing of information to generate value-added outputs,” write Alavi and Westerman. It includes both structured and unstructured tasks. Previous digital tools have been good at structured tasks. Now AI is making inroads on unstructured ones as well.

Drawing on current studies, Alavi and Westerman observe that gen AI can provide three main benefits to knowledge workers. It may help to communicate and teach these benefits to employees:

  1. Reduce cognitive load—Gen AI can summarize long documents, locate crucial information rapidly, and create routine communications, like brochures and personalized emails.
  2. Boost cognitive capabilities—Gen AI can help with idea generation, divergent thinking, critical thinking, and access to knowledge across an enterprise.
  3.  Improve learning—Gen AI shows promise as a learning tool that can personalize feedback on the job, serving as a digital tutor, trainer, or mentor.

While employees may need help seeing these benefits, businesses will also need to create the right conditions to promote, advance, and reward innovative uses of gen AI. Alavi and Westerman propose taking four steps:

  1. Define policies and assign responsibilities—Businesses need to set policies on AI to reduce risks, such as the potential for privacy violations and for misleading AI hallucinations.
  2. Encourage experimentation and innovation sharing—Employees should have chances for peer learning, times when they can teach and critically discuss new ways they are using gen AI.
  3. Celebrate wins—When employees use gen AI well, “make a big deal,” and “celebrate the innovations and the innovators,” write Alavi and Westerman.
  4. Don’t wait—Start using or experimenting with gen AI now, to begin building gen AI knowledge and skills that are likely to be necessary for remaining competitive in any industry characterized by knowledge work.

“Find the AI Approach That Fits the Problem You’re Trying to Solve” 

By George Westerman, Sam Ransbotham, and Chiara Farronato

As enthusiasm for gen AI grows, businesses may be overly eager to use gen AI for diverse tasks, including ones for which it’s poorly suited. “Wielding a (generative AI) hammer, everything starts to look like a nail,” write Westerman and coauthors.

To counteract this impulse to reach for generative AI whatever the task, companies should consider a broader range of advanced analytics tools and select the one that fits their problem best:

  1. Generative AI reduces cognitive load for working with long texts (like new regulations) and creates “new data, images, video, text, or sound.” But it is susceptible to hallucinations and requires “huge amounts of computing power and data.”
  2. Traditional deep learning is good at ingesting “large volumes of complex data to learn patterns and relationships.” But outputs are hard to explain, meaning deep learning may be a poor tool where regulations are strict.
  3.  Econometrics, statistical methods for discovering causality, is relatively cheap and low tech. Its outputs are also repeatable.
  4. Rule–based automation. Such automation follows “if-then” models. These systems “can be understandable, interpretable, and transparent,” but require expertise and are inflexible to change.

To select the right tool, Westerman and coauthors urge leaders to ask five questions:

  1. What is the cost of being wrong? Gen AI can be useful at creating “close enough’ answers,” but other tools may be better when error is unacceptable, such as in air-traffic control.
  2. Do you need to explain the decisions your model makes? Some tasks—like issuing loans and selecting job candidates—require explainability. In those cases, econometrics and rule-based automation may work best.
  3.  Do your models need to generate the same answers every time? Econometrics and rule-based automation give repeatable answers but may not be accurate, flexible, or powerful enough. Deep learning can be repeatable but also changes as the data on which it trains updates. Generative AI changes its outputs for each new context.
  4. Does your data have a source of truth? This question may matter most if you’re considering deep-learning analytics tools. Such tools require large and well-labeled data, without which these tools may be “no better than flipping a coin.”
  5.  Does your training data reflect the conditions under which you’ll operate? AI and other deep learning tools need to be trained on data that is reasonably comprehensive and unbiased. For example, investment data must reflect good and bad times, and medical data need representative samples of men, women, and races and ethnicities.

* * * 

Developing an AI strategy may seem daunting. These frameworks may help–with selecting analytical tools, and developing new skills and innovating. 

Back to All News & Research