Past, Present, and Future of Lean and AI

Gene Kim, DevOps leader and co-author of Wiring the Winning Organization, joins Tyson Heaton, LEI Executive Director of LeanTech/AI, and Art Smalley, Toyota veteran and LEI advisor, for a lively, enlightening, and wide-ranging conversation about artificial intelligence. They discuss how AI is forever changing the work of individuals and organizations, and explore the good, bad, and ugly of AI today and tomorrow.  

One thing they all agree on: the use of AI is picking up, even among lean practitioners. Up to 80% of people in the lean community I suspect still have little to no understanding of AI (“What’s a model?”), estimates Art. But there are examples that the other 20% across industries have started running with it. Tyson mentions, “incendiary events,” like an interaction with a hackathon, that are fueling some of this momentum as individuals return to their work with new capabilities and an appreciation for the technology.  

“I've coded with chief medical officers, chief science officers, CEOs, CFOs, people in like revenue cycle management and healthcare,” says Gene. “And just to see these people all solving problems by themselves, you know, as they're using coding agents. It's just amazing to see.” 

Discussion highlights from this often-amusing podcast include: 

• AI finds things that humans alone will not: At a personal and corporate level, AI is uniquely useful at analyzing and finding desired information in any form (data, images, etc.) within unstructured data, says Art.     “One of the big angles of this LLM play into functions out outside of ops —  whether you're in purchasing HR, legal, finance ,ops, whatever — they're all sitting on mountains of unstructured data... Apple, of all companies, has this problem, too. There's unstructured data, structured data, and types, and suddenly we now have a way to ingest a lot of that and make it searchable in a way that was not before.” 

• Like lean standardization, AI removes waste from work that can open space for creativity: “Folks who are new to lean and standardization, they see a standard as potentially restrictive and limiting the creativity and limiting the potential,” says Tyson. Mature lean users, however, recognize that standardization removes constrains and allows them to create and experiment to produce better results.     “A decent chunk of [knowledge work] is basically pattern understanding, pattern synthesis, and pattern projection to get work done,” he adds. “And if you can take and offload that or pair with AI, which it's really good at, it unleashes your creativity.”  

• Negative reaction to improvements made possible with AI: Gene tells the story of an individual who would do many hours of Excel work each day on a highly specific “horrendous” task, and then AI applied a macro that could do it for her. “Her reaction was anger. She's like, ‘How dare you? I didn't ask for this.” AI removed something she held “near and dear.”     Tyson says a path to overcome this natural human reaction to someone’s work being removed is a better foundation of understanding of value-add vs. non-value add vs. incidental work from the customer’s perspective. 

• The fear of AI taking jobs is real for some individuals and in some companies: Art relates AI improvements back to early lean improvements, when removing non-value-added piece of work from someone’s work was viewed as taking away their job; they did not recognize the lean improvement was freeing them to think of the other ways they could be adding value to the organization.     Tyson says it’s critical to help rewrite people's identity as creators and improvers rather than lathe worker or spreadsheet specialist. “It's also kind of a tragedy that so many people and so many organizations aren't afforded the ability to remake their identity as an improver of their own work processes and workflows because of either the leadership structure or the behavior, the cultures of the organization.”  

• Overcoming resistance to AI: Pair coding/programming (when two people code together) and learning by doing together has real value in getting individuals to embrace what’s possible with AI, says Art. “If there's a respected person in the organization, sometimes there's peer learning that occurs. There's different ways people get influenced, different directions. And peer coding seems to me to be a very natural way of transfer of information that's often mutually beneficial, and you get a better result, learn by doing, low risk, high reward.” He adds the interest in AI goes from incremental to exponential.     Gene agrees, and points to examples of increased activation among individuals learning it and just doing it with respected members of their technical community. “We're now in a universe where it is easier to just do the experiment than it is to sort of like describe to someone else and try to justify the experiment. It's like, no, just open the laptop and try it, and we're all going learn something.” 

• AI weaknesses and role of humans: “It is difficult to overstate just how useful and expert these things are,” says Gene. “So that's on one hand. On the other hand, like these things are also like summer interns. They are prone to doing the most stupid things, right? They delete data, they don't follow instructions very well, they sometimes get confused, right? They forget things. So your job as a human right now —  kind of in this weird time where we know what how powerful the AIs are — sometimes it takes a lot of vigilance and supervision and judgment to make sure that are we really getting closer to the goal.” 

• The dangers of falling in love with AI: Art says AI has given him a sense of control, wonder, and fun in building things again that he has not experienced in years. Gene also says that solving problems — his own and those of others with AI — has been exhilarating and that he hasn’t “been this productive in a long time, maybe even ever.”     Tyson notes, however, that the way he’s working, too often task-switching and working on multiple projects while AI delivers outcomes, “was not leading to the result I wanted either in like the depth of work around a component of work that I was doing or my attention span. I think I'm getting a little shorter, and I think I'm losing some depth... I need to countermeasure the way that I'm working to be able to make sure that I'm kind of maintaining myself as a person of depth, as being able to dig in deeply into these issues that are important or into these work threads that are important, rather because otherwise I spend a whole bunch of time cleaning up the slop that came because of my poor custody as a workflow manager.”    Gene says that the critical personal constraint for AI has gone from the number of tokens a user has and the number of agents to how focused we can be on one thing. “Attention, focus is like the most precious constraint and constrained resource we have.”