Auditing a system that uses AI

We’ve talked before about some of the Quality issues related to the use of artificial intelligence (AI) technology, but mostly in an incidental kind of way.* But the adoption of AI by many companies seems not at all incidental. So it is fair to ask how the introduction of AI is going to affect core Quality functions. How will AI change—say—auditing?

It's a good question, and there have been a couple of attempts to answer it in the past year. Last October, Elisabeth Thaller and Jorge Bravo Carreño wrote an article "When the QMS Thinks for Itself" for Quality Progress. Then just last month the ISO 9001 Auditing Practices Group published a formal guidance paper on "Auditing a Quality Management System that uses Artificial Intelligence (AI) systems." It is perhaps not surprising that the article and the guidance paper repeat the same basic points,** and they are pretty much the same points you would make if someone asked you about the same topic. Nothing here is unexpected.  

What makes this consistency possible is that ISO 9001 is structured to apply to any kind of work. So the basic questions are always the same. What are your processes? What are your risks? What are your training requirements? And are you getting the results that you want?

The article in Quality Progress identifies five main thoughts to keep in mind when planning an audit in an organization that uses AI, and the sugges­tions in the APG guidelines can almost be mapped directly onto the same list. The five main thoughts are these:

1. Identify where the organization is actually using AI. ("AI may be hiding in plain sight.")
2. Use the process approach to identify inputs and outputs in the normal way. ("The process approach still applies—even when the process thinks for itself.")
3. Identify risks. ("Risk-based thinking isn’t optional.")
4. Identify training needs, responsibilities, and authorities. ("Competence and oversight are essential.")
5. Identify results and look for objective evidence. ("Focus on evidence, not the shine.")

And while I'm certainly not going to quote all ten pages of the guidance document in a blog post—you can download the whole thing for free by using the link above—let me quote just a couple of items that fit under each of these five headers, to show you the overlap.

Identify where the organization is actually using AI.

• When preparing the audit, the auditor should:
• Identify whether the organization uses any AI system type for any of its processes within the scope of the QMS.
• Determine the specific function(s) of these processes within the QMS, and
• Recognize assigned responsibilities in relation to these processes. [§2, p.4]

Use the process approach.

• Where AI system(s) are used within operational processes, has the organization ensured that their use supports the achievement of intended results?
• Are such processes monitored, and are changes related to the use of AI system(s) controlled?
• How does the organization ensure the consistency, correctness, and reliability of AI outputs? [§3, cl. 8, p.7]

Identify risks.

• Has the organization considered the implications of the use of AI system(s) when determining its internal and external issues?
• Has the organization identified statutory and regulatory requirements in relation to the use of AI systems, such as those related to data privacy and information security? [§3, cl. 4, p.4-5]
• Has the organization determined and addressed risks and opportunities related to the use of AI system(s)? [§3, cl. 6, p.6]

Identify training needs, responsibilities, and authorities.

• Are adequate resources available to maintain and update AI system(s), including access to technical expertise, data quality management, and cybersecurity support?
• Is the organization able to demonstrate that personnel who use, manage, or oversee AI system(s) meet the determined competence requirements? [§3, cl. 7, p.6]
• Does top management take accountability for ensuring that the AI system(s) within the QMS support its effectiveness and achievement of intended results? [§3, cl. 5, p.5]

Identify results and look for objective evidence.

• Are customer satisfaction trends evaluated for potential impacts arising from AI-enabled processes or interactions?
• Is internal auditing addressing the effectiveness of AI-influenced processes?
• Is information related to the performance and effectiveness of AI systems an input into management review? [§3, cl. 9, p.8]
• Are nonconformities or complaints involving AI system outputs recorded, analyzed, and addressed? [§3, cl. 10, p.8]

There shouldn't be anything shocking in this list. One of the strengths of ISO 9001 is precisely its flexibility. And that flexibility comes from seeing all kinds of work through a lens that highlights certain common features. AI is just one more topic viewed through the very same lens. This is why I said above that you would have come up with the same points if you had taken the time to work through the standard in detail.

But most of us are too busy with our day jobs to do that, so it's convenient that the Auditing Practices Group has done it for us. The next time you have to audit an organization that uses AI in its management system, check out the guidelines document to see if any of the suggestions help you. 

__________

* In this post, I asked what the word “Quality” really means in the context of AI. In this one, I discussed how AI tools might be able to enhance training records. 

** Partly this outcome is unsurprising because the same person—Elisabeth Thaller—was the lead writer for the article and the chair of the committee that wrote the guidance document. Chris Paris published an article taking the APG committee to task because most of them have no special background in AI. But it appears from her LinkedIn profile that Thaller does have at least some background.        

           

Sampling bias

A friend forwarded this cartoon to me a couple of weeks ago. It seems to come from the website Sketchplanations, drawn by Jono Hey. You can find the original here.  

Of course statistics are central to the Quality discipline, so we need to understand how sampling works. And I think the cartoon makes a serious point in that regard, rather deftly.

It's also funny, and high time too!

            

Can you be TOO optimized?

There's an old story about a wise man that has some relevance for bloggers as well.

Once upon a time there was an old man renowned for his wisdom. Whenever a discussion got tangled and tempers rose, he always told a parable that resolved the question perfectly. Finally a little boy asked him, How could he unfailingly pull out a story each time that summarized the issue so well? "Ah," said the old man, "you don't understand my method. I don't ransack my brain looking for a parable to match each new conversation. Rather, I introduce a topic of conversation for which I already have the perfect parable."

Blogging is a little like that. We bloggers are sure we have the right answer, whatever the question. And by the time you read a post all the way to the end, we've usually made the point pretty convincingly. It almost looks like we're right about everything. (Or I hope it does. 😃) But then, ... who picked the topic for the post in the first place? We did! So of course we're going to pick topics where we think we can win the argument.

But I've been thinking recently about a plant I encountered a few times during my career, where things seemed slightly off-kilter but I could never tell why. I didn't work there, and they weren't part of my normal responsibilities. But I visited a couple of times, to support audits or for other reasons. The audits turned up findings, as audits always do; and the findings were always addressed in the usual way. Somehow, though, I never got the sense that we uncovered anything more than surface issues. It felt like there was something more basic that we were missing.

This plant was located in a town I'll call Riverville, so I can refer to the place as the Riverville plant.

Recently I've been formulating a hypothesis that might account for part of the situation there. So I want to describe the hypothesis, and ask whether you think this could pose a problem in real life? I won't go into a lot of details about how things were at Riverville, if only so that I don't identify it too plainly. But let me know if you think the circumstances I describe could cause some related mis-fires.

Then if it turns out that I'm all wrong, I can go back to questions where I (think I) already know the answer in advance. 😀  

The Riverville plant

Visiting and auditing

The Riverville plant had been in operation for many years. The people were deeply competent, and committed to their work. Also, because these things happen, the plant had undergone multiple changes of ownership throughout the decades it had been in business. Sometimes the new owners ran it as a standalone operation. But occasionally the new owners wanted it to integrate into a larger ecosystem of other businesses that they had also acquired. When I knew them, Riverville had been recently acquired by Octopus Enterprises (not their real name).

Despite the acquisition, though, the plant never felt like the other Octopus plants that I had seen. When visitors started to quote a lot of the normal Octopus manage­ment terminology, I thought I could see the Riverville old-timers smile to them­selves, as if to say, "Sure, we'll play along until you go home. But that's not how it really works." When I audited, I kept turning up forms that I'd never seen at any of the other Octopus plants. "Don't you use the standardized Octopus forms?" "Oh sure, sure, ... mostly we do. But that one's an exception because of the special way we handle [topic] here. It's authorized, though. Let me show you the procedure document!"

And there was always a procedure document to back it up. There were so many procedure documents, in fact. Sometimes I thought it would be an interesting project to collect all of them and paste them together, to try to get a picture of how the whole plant functioned. But of course I couldn't. As I said, I never worked there regularly, so in the normal course of things I would have had no business even raising the question. And in the absence of a total breakdown of their system—which never happened during the time I was familiar with the plant—I could never have made an argument to justify the time and effort. Besides, mostly the folks at Riverville were able to provide the right management deliverables back to Octopus headquarters when asked, and mostly they were able to get their regular work done in a routine way.

Fully optimized

As for those forms? Oh, the forms were awesome! Someone had put a lot of time into these. There were links to data sources on the internal network, so that when you needed to fill in fields 5-15 you just clicked a button on the form. Then you added a little more data by hand and clicked another button, and right away the form routed itself to the next person in line. None of this was managed by some overall data management tool. It was all hard-coded into the forms with URLs that pointed to archive repositories where project information was stored in suitably-structured Word and Excel files. 

Rebuilding the engine in flight.
This is not recommended.

In fact, the forms were the key to my recent hypothesis. I think the Riverville plant was too optimized to change! They had worked together for so long that they had optimized every single transaction in a large and diverse factory. But they optimized them all according to the outlines of the old Riverville management system. If a new owner let them run themselves, and just showed up at the end of the year to collect the profits, there was no harm done. But when Octopus Enter­prises tried to make them integrate with the Octopus manage­ment system, the only way they could do it was to run two systems in parallel. They couldn't stop using the old forms, because everyone relied on them. They couldn't stop using the old data­bases, because the forms pulled data from them. They couldn't stop working in the old way, because everyone in the company depended on every step of the old system. To replace the old Riverville system with the new Octopus system would have meant rebuilding the airplane engine in flight. 

I was never able to dig deeply enough to prove that the Riverville personnel were running two systems in parallel—I mean, as opposed to adapting the Octopus system by adding "just a couple" of the old Riverville forms—but it would make sense of my general observation that somehow things were a bit off. As we discussed last week, running the Octopus management system as a Potemkin system would have been a perfectly natural response, if the Riverville management felt that they were being forced to make changes that didn't help them any. 

The problem, of course, is that working in two parallel systems takes twice as much effort as working in one. Also, if you have data in two places, it is almost impossible to keep them both consistent. Whatever options Riverville faced, this one cannot be called The Easy Choice.

End of the Bronze Age

My thinking about Riverville has been supported in an interesting way by an article that appeared in LinkedIn around the end of March. The author is Marco Nutini, a risk-management expert from Brazil, and the title is "1177 BC Called. We're Not Listening." 

It's a fun article, and not long. Read it, if you have a few minutes. Nutini starts off summarizing his thesis with admirable brevity:

In the late Bronze Age, the Eastern Mediterranean was a marvel of interconnection. Egypt, the Hittites, Mycenaean Greece, Ugarit, Cyprus — linked by trade routes, diplomatic marriages, and supply chains that moved copper, tin, grain, and luxury goods across vast distances. It was, by the standards of the era, a globalized world.

Then it collapsed. Not from a single cause, but from all of them at once.

His point is that the highly interconnected world of the late Bronze Age was optimized for current circumstances. Assuming that the world stayed at peace, that the rain fell reliably, and that no major population centers were wiped out by earthquake or volcano—assuming all that, the world ran just fine. But when those assumptions failed—when drought, famine, and earthquake came, when cities revolted and the Sea Peoples were landing on every shore—the finely-calibrated economy of the late Bronze Age could not withstand the shocks. It fell apart, collapsing into the Greek Dark Ages. Optimization for normal times is not the same thing as resiliency in trying times.

Admittedly, having your factory bought by Octopus Enterprises shouldn't be as devastating as having your seacoast raided by pirates, or your island explode. But either way, I think there is a risk that over-optimization can make you fragile and unable to adapt when circumstances change.

If I'm wrong, feel free to tell me so! 

            

"The absence of alternatives ...."

We've talked at some length about how companies learn to improve. But in many ways it sounds like a difficult process, requiring a lot of persuasion over a long period of time. So it must be tempting to wonder if there is anyway to shorten the exercise. 

Can you force a company to get better?

Henry Kissinger, of course!

You can certainly supply strong motivation. Years ago I worked for a small startup that had no special interest in certification. We knew our technology was good, and that's where we focused our attention. Then one day our largest customer told us that they had adopted a new policy: after January 1 of the next year, they were going to cancel all contracts with any supplier that had neither ISO 9001 certification nor a solid plan to get it. I remember our CEO announced to us in a Town Hall meeting that our new strategic plan required us to work towards certification with all deliberate haste. As he laid out the details, he quoted Henry Kissinger: "The absence of alternatives clears the mind marvelously."

Grigory Potemkin: Don't
let him plan your ISO
9001 implementation!

Does it work? Well, we got our ISO 9001 certification, although by the time we finally achieved it our market had changed and we no longer relied so heavily on that one customer. (The tech market is notorious for turning on a dime.) But there's a risk to this kind of forced compliance. If you are not very careful with the implementation of the new measures—ISO 9001 regulations or whatever else they might be—your organization could settle for a superficial or "Potemkin" compliance in which the new methods are used to create an overlay over top of the old ones. One day every year, at the time of the audit, everyone follows the new methods and uses the new terminology; then for the next 364 days, work is done like it always was before. This is not an effective approach.

We've talked about this risk before (for example, in this post and this one) and it is always real. When Robert Cole summarized research on American companies adopting Japanese quality methods in the 1980's and 1990's, one study in his pool focused on six second-tier automotive suppliers who were required by the major OEMs* to adopt Total Quality Management (TQM). Of those six, only one adopted the methods in a way that made a lasting difference, or that was fully integrated into the normal way of working. Four suppliers adopted the new approach "rather mechanistically, with the methods not being used regularly or to their greatest potential." And one supplier did nothing at all. Cole describes this nearly-perfect statistical distribution dryly as "a full range of outcomes," and points out that in many cases it was only years later that these suppliers finally put in the effort to upgrade their operations.** 

Naturally this is only one study. It doesn't prove that if you try to force improvements on your suppliers, you'll have only one in six odds of success. But it does remind us—what we already know—that improvement is never easy. It does suggest that if you want your suppliers to improve in some respect, you should work with them, stay engaged, and support them on their journey. As we've seen, external support always makes a difference. And in the long run that support may be more effective than giving your suppliers an offer they can't refuse.    

__________

* OEM = Original Equipment Manufacturer. In this context, it means General Motors, Ford, Chrysler, and so on. 

** Robert E. Cole, Managing Quality Fads: How American Business Learned to Play the Quality Game (New York, Oxford: Oxford University Press, 1999), pp, 127-128.       

           

Know your own procedures!

Who needs to understand the procedures in your department?

Well, I suppose you do, for a start. The other people in your department should understand them too, or anyone else doing the same work. The people who interact with your department should understand enough that they can work with you. If you work with the public, then they need to know enough to do their part: the Information desk is here, and the Returns counter is over there.

How about your boss?

Of course. The boss established all the procedures, so of course he understands them.

How about his boss? How about top management?

Umm ... I guess. Maybe not in detail, but overall. Why? What's your point? 

You'd be surprised.

In a sense, maybe it shouldn't be a surprise when top management doesn't know the procedures that run their own company. Most of their day is spent doing other things. Still, it can be awkward. These are the people, after all, who decide on the company's strategic direction. But what if the new direction points somewhere that the procedures aren't designed to go?  

It sounds like a frivolous or captious question, but it has been known to happen. Robert Cole tells the story that in 1983, Ford's CEO instructed the company to improve relations with their supply base, as a key to improving their suppliers' performance.

"Based on this, [L. M.] Chicoine [Ford's vice-president for purchasing and supply,] publicly called for increasing the number of long-term contracts (greater than one year) with Ford. To his surprise six months later, he found there was no change. Because Ford rules called for extensive bureaucratic approval for any contract greater than one year, no supplier was interested in tackling that."* 

Sometimes the disconnect between management direction and corporate behavior is more subtle. Another story from Cole tells of ongoing discussions in the early 1980's between Intel and one of their major suppliers, the Japanese firm Kyocera. Intel argued that they got poor service and poor terms from Kyocera, compared to Kyocera's other customers. 

"The Japanese managers responded with a riddle to Intel queries: 'To Kyocera, the customer is always king. But reliable kings have reliable servants.' The Intel managers finally figured out that the point of the riddle was that although the customer (Intel) might be king, it nonetheless had to act in certain ways to produce reliable supplier behavior. Above all, that meant no order cancellations, level production, and an overall predictable environment for suppliers."**

In the story about Ford, top management genuinely did not understand what steps were required by their own procedures. In the story about Intel, top management did not understand how their normal behavior was perceived by others. But both cases represent a kind of disconnect between the strategic direction that top management wanted to go, and the organization's ability to get there.

I have tried to think about my own experience, and can't recall any disconnects as significant as these ones. But I have certainly seen how easy it is for top management to miss realities on the ground, because the people around them shield them from inconvenience. Once I worked for a company where the management held periodic Town Hall meetings, to get feedback from those of us in the trenches. At one of these meetings, someone complained that the IT department needed more resources, because whenever his PC broke it took three weeks to get it fixed or replaced. The vice-president who fielded the question genuinely didn't understand the issue: he explained that whenever he had a problem, IT always fixed it the same afternoon!

This phenomenon—this disconnect—is why it is so important for top management to get out of their offices and understand how the work is done. It is why John Seddon says that his first step when working with a new client is to send the management staff out to the front office to watch a single order come in, and then to track that order through its whole life-cycle. It is why Taiichi Ohno invented the Gemba walk.

The terminology differs, but the basic idea is the same. Before you can make decisions about what to do next, you have to understand what you are doing today. And that's not always easy. 

__________

* Robert E. Cole, Managing Quality Fads: How American Business Learned to Play the Quality Game (New York, Oxford: Oxford University Press, 1999), p. 112. 

** Ibid., p. 113.       

           

Management by travel

Last week I wrote about Florida Power and Light learning Total Quality Control from Kansai Electric in Japan, and I mentioned that "FPL and Kansai set up an extensive program of travel and visitation, so that personnel from each company could build personal contacts with their counterparts." It was a single bullet-point in an article largely about other things, so many readers might have missed it. But the technique is a powerful one.

Years ago, I worked for a Bosch subsidiary called ETAS. ETAS had run as an ambitious startup for some years by then. They were beginning to expand, with sales and development offices outside of Germany. But they wanted to preserve the unique company culture that had brought them their early success. Now, culture requires communication, and the most effective communication is always in-person. Nothing else replaces being there. So the company made a strategic decision for several years to invest in business travel.

This meant that the office in California where I worked had regular visitors—from the other ETAS office in America, and from the home office in Germany. We had workshops to bring together our engineers with their counterparts from other product lines. Likewise for the sales force; likewise for the customer service team. The global Process Owners for each of the business processes in our Quality Management System visited regularly to check how the systems were working for us. Nor was the travel all in one direction: there were events held in Germany by this department or that one, and our people were regularly invited—expected—to attend.

The organizational structure was international, as well. For some years my immediate supervisor sat in Germany. (I was still in California, so we were separated by nine time zones.) Three key positions in our office's management structure were filled by ETAS personnel who relocated from Germany. On the other hand, ETAS's Global Director for Operations and Finance was an American. International conference calls were routine, and so was the travel.

Not my passport. I downloaded this
picture from Wikipedia.
But you get the idea.

One of my topics in the Quality department was internal audits. The ETAS program assumed that for each internal audit there should be at least one auditor from somewhere else, to avoid the risk that a local auditor might get jaded by familiarity and stop seeing nonconformities. So I routinely led audits in our other American office, and in Germany; and we routinely invited my German colleagues to come audit in California. I didn't realize quite how routine the travel had become until one day the Passport Control officer in the Stuttgart Airport had to search for a blank page to stamp and finally told me, "You come here a lot."     

So did it work? Yes! It worked very well.

The net result of all this travel is that we really were able to work with our colleagues from other offices and other countries as if they were just down the hall. It was not unusual for me to chat with one of our engineers and hear him casually reference recent conversations with our German top management—referring to them by first name. We discussed common concerns, asked and gave support on common projects, all as if we were working in the same building. Of course there were occasional misunderstandings arising from imperfect translation or differing legal backgrounds; but they were easily managed, because we had built up friendship and familiarity by working together.

Ultimately, as the company continued to mature, we got to the point where it was no longer as critical as it had once been to reinforce the cultural framework; and we no longer poured quite so much money into travel. The savings certainly made a difference: there is no question that our travel-based management strategy had been very expensive. At the same time, it is also unarguable that it helped tie together a collection of far-flung offices into a single operating unit. 

Do I recommend the same strategy for every company? Well of course it depends. Not every company has multiple offices in different countries; of the companies that do, not all of those need their offices to work together seamlessly. As a result, many companies don't need anything like this approach. And I'm sure that some companies which might otherwise benefit from it, can't afford it. If the margins in your industry are too narrow, that limits what options are available to you. 

But it is still true that the best communication takes place in-person. It is still true that no electronic tool ever quite takes the place of being there. That's what makes travel and visitation such powerful tools for building a culture and binding a company together. 

           

Action and reaction

How do companies learn?

We've discussed this question on and off for years, and mostly it's not easy. Two years ago, in the middle of a series of posts on Boeing, I wrote about data falsification scandals at Toyota and suggested that it can be phenomenally difficult for a company to learn new attitudes. Two weeks later I backed off to say that it's possible but not easy. And just last month I wrote about Robert Cole's study, Managing Quality Fads, about how American companies learned quality methods from the Japanese during the 1980's. So yes, it's possible but not easy. 

That conclusion just brings us back to the original question: If it's possible, then how? And here it can be useful to look at success stories—companies that did learn a lot from others, and improved their operations accordingly. One advantage of Cole's book is that he gives a lot of detailed data about the period he describes, including a number of case studies. Of these, the first big success in adopting the new quality methods was Florida Power and Light (or FPL).

What made their success possible?

Cole sets the stage by explaining: "FPL [began] its quality improvement ... in 1981. In 1985, it became the first large American company ... [to start] learning directly from the Japanese. It entered into a deep exchange relationship with Kansai Electric and other ... Japanese companies[,] and worked closely with Japanese professors...."* But there were a number of special factors that supported FPL's quality initiative.

• FPL was in the same line of work as Kansai Electric, but they were not direct competitors. So neither company feared accidentally revealing trade secrets to the other.
• FPL and Kansai Electric used the same equipment—they both used Westinghouse-designed nuclear plants. So FPL trusted Kansai's reported metrics, rather than waving them aside as "apples and oranges." When they learned that Kansai's Mihama plant reported 10% as many quality problems as FPL's Turkey Point plant, they looked right away for management and operational differences that could explain such different performance from the same infrastructure.
• FPL and Kansai set up an extensive program of travel and visitation, so that personnel from each company could build personal contacts with their counterparts.
• Crucially, the CEOs of the two companies built a strong personal rapport. It soon became flatly unacceptable to express anti-Japanese sentiments at FPL.**   

What did they do?

With this background as a framework, FPL adopted Kansai's Total Quality Control (TQC) measures almost intact. The successes were dramatic enough that four years later (in 1989), FPL was the first non-Japanese company to win the Deming Quality prize.

Perhaps more to the point, FPL's success got the attention of other American companies. Soon FPL was giving seminars on quality improvement; and in 1986 they spun off a subsidiary (Qualtec) as a for-profit consulting firm. In the end, Qualtec could not keep up with the demand for its services, so FPL sold it to the Marshall Group in 1995.

What happened next?

Newton's Third Law of Motion famously says that to every action corresponds a reaction, and I'm not the first author to apply the same phrase metaphorically to human organizations as well. In 1989, James Broadhead took over as FPL's new CEO; and after the company won the Deming Prize, he began to dismantle much of the quality superstructure and documentation requirements. He argued that this structure and these requirements were too extensive to be cost-effective, and therefore that they were inconsistent with FPL's overall business objectives.

But it is important to understand that Broadhead did not rip out the company's quality measures root and branch. And FPL's overall quality levels did not suffer under his administration.*** What he did was to integrate the quality measures into normal operations, so that they became "just how we do things" rather than adding them as special steps administered by the Quality Department. As that integration took place, he could also shrink the size of the Quality Department because there was no longer a need for so many extra staff. He found that FPL had implemented a lot of documentation requirements because they were demanded by the Deming Prize; but in fact, no one ever looked at the documents again after the prize committee left. So he eliminated those requirements that had no use. He continued to support benchmarking, self-managing teams, process reengineering, and employee empowerment. So while the American business press described Broadhead's administration as a massive rejection of Japanese quality principles, it was nothing of the kind. He kept what worked—all the measures that reduced downtime and outages and disruption of service—because they worked! All these measures were profitable. He just made them unremarkable enough that they no longer drew special attention.  

Points to take away

The original question was, How do companies learn? So concretely it is fair to ask, How much of FPL's experience can apply to another company?

Obviously FPL faced special circumstances that gave it almost a perfect framework for learning from someone else: for starters, the partnership with Kansai, who was not a competitor and who used exactly the same equipment.

But I think there are three points that have a wider application, and it is important to remember all three.

1. Any learning initiative—really any change initiative of any kind—has to have strong and consistent executive support. The forces opposing change are always powerful; so without reliable pressure from the top, it is likely that nothing will happen.
2. 
   The more widely you spread information about the initiative or the new way of thinking—and the more personal you make it—the better your success will be. One of FPL's big successes was their travel initiative, that brought so many of their employees face-to-face with their counterparts at Kansai. Once they made friends, it was easier to talk about problems at work—and to listen to the answer.
3. Unless there are external measures to guarantee the change is permanent (for example, if one company has bought another), there will be a backlash in time. Don't waste effort trying to fight or avoid it. The better strategy is to make sure that all the most important parts of the change are fully embedded into normal practice by the time that the backlash arrives. Let the new managers pull down some posters or eliminate other symbolic reminders of the change initiative. But make sure the new methods themselves are so routine and so profitable that no one thinks of changing them.   

If you remember all three points, is that enough to make it easy for your company to learn? Probably not. But they are sure to help.

__________

* Robert E. Cole, Managing Quality Fads: How American Business Learned to Play the Quality Game (New York, Oxford: Oxford University Press, 1999), p. 66. This whole story comes from the same work, pp. 66-71. 

** If I remember the early 1980's correctly, FPL's attitude could not yet be reliably assumed across the country as a whole. 

*** So James Broadhead's career at FPL does not foreshadow that of (let's say) Harry Stonecipher at Boeing, a few years later.