It seems like generative AI has reached a tipping point, with a majority of firms indicating that they already use it to some degree, and many others stating that they are in the process of adopting it (e.g., see this report).
Yet the uncertainty is high. We don’t know how AI will impact individuals, organizations, or society. As Wharton Mollick remarks in his recent book:
“I can assure you that there is nobody who has the complete picture of what AI mean, and even the people making and using these systems do not understand their full implications.”
Given this uncertainty, how should organizations act in the near term?
Although generative AI is new, it is just a the latest in a long line of technological innovations. So we have some knowledge about how organizations should act in situations with high uncertainty. Based on this, I would suggest five key strategies:
1. Historical parallels. The first is to learn from changes in the past. As Tom Davenport points out, organizations have been mechanizing and automating jobs for centuries. Among other things, we have learned that change is typically gradual, and that if often requires a reconfiguration of several elements to profit from new technologies—it is not enough to introduce a single tool; you also have to change the social system and business model around it in order to reap the benefits.
2. Environmental scanning. New AI tools and upgrades seem to be introduced every week now. It is hard to keep with all of these developments for individual employees. So I think there is a need for an organizational response. Environmental scanning is defined as “the activity of acquiring external information.” Although we can all learn by following developments, I assume that with the speed and complexity of this issue, it needs to be formalized. In other words, that someone (an individual or team) is given a particular responsibility for observing and evaluating what is happening.
3. Scenario planning. By definition, high uncertainty means that we cannot predict very well what the outcome will be. But we can still become “future literate” (to borrow a term from Reil Miller) by considering alternative scenarios and imagining alternative futures. More specifically, leaders need to spend time on this task—among other things, discussing the results of the scanning I described above, and consider the threats and opportunities for the organization.
4. Experimentation. We do not know where generative AI will turn out to be most effective, or exactly how it should be implemented (although scholars are beginning to develop theories about this, e.g., see this paper). In such situations, the right approach is to experiment. One can use tools (such as the one we have developed—Reconfig) to get an initial estimate of the potential productivity increase from automation. Then one can initially do a local experiment by implementing AI, and if it works, scale it up in the next phase.
5. Psychological safety. As I mentioned, the more recent reports suggest that change will be gradual and that augmentation will be more common than complete automation, at least in the near term (e.g., see this report). Nonetheless, there are clearly jobs and processes (e.g., customer support) that are at risk in many companies. So it’s a tricky situation for leaders. How should they do it? I asked ChatGPT for advice, and here it what it suggests:
Leaders must actively involve employees in the development and implementation of new AI processes. This includes seeking their input and allowing them to participate in pilot projects. Such involvement not only helps in fine-tuning the technology to better fit the operational needs but also gives employees a sense of control and ownership over the changes, potentially easing the transition and reducing resistance.
Sounds like good advice.
I will return with further blog posts about topics such as:
- Using ChatGPT for organization re-design
- The impact of AI on organizational structure
- AI and accountability
Stay tuned.
Related posts:
Should there be a computer on your organization chart?
Imagining an alternative future
Nicolay
I like your proposed processes. In my experience, most Stratum V business units don’t have the skilled staff to use these processes. Stratum VI companies are more likely to have the required staff skills, and Stratum VII are the most likely to be able to implement your suggestions.
The book ‘The shock of the old’ by Edgerton is really a very good illustration of the first point about historical parallels.
I started working in 1985 upon graduating as a mechanical/ systems thinking engineer. Much of the automation debates I hear now (industry 4.0 and beyond) seem very very similar to the discussions back then. I’m still wondering about a difference that might make knowledge gained back then largely obsolete … but so far nobody I asked about it sees a relevant difference.