---
id: "workshop:learning-and-pivoting"
type: "InnovateUS workshop"
title: "Innovating in the Public Interest: Learning and Pivoting"
url: "https://innovate-us.org/learning-and-pivoting"
date: "2025-05-01"
duration: "60 minutes"
language: "English"
series: "Innovating in the public interest"
instructors: ["Anita McGahan"]
recording: "https://www.youtube.com/watch?v=GhBfGqWHsFw"
transcript: "automatic, from the recording"
source: "InnovateUS (innovate-us.org)"
licence: "CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org"
---

# Innovating in the Public Interest: Learning and Pivoting

InnovateUS workshop · 1 May 2025 · 60 minutes

Just about every cross-sector collaboration project hits a brick wall at some point. What does it take to get past that challenge while staying true to public priorities? We will talk about how strong public-private partnerships adapt, learn, and pivot while keeping long-term goals and public accountability in focus.

## Learning goals

- **Anticipate and Address Roadblocks** – Recognize common challenges in public-private collaborations and develop strategies to navigate them effectively.

- **Adapt to Changing Needs** – Learn when and how to pivot initiatives while staying aligned with public-sector goals and community priorities.

- **Strengthen Collaborative Resilience** – Develop skills to manage setbacks, rebuild momentum, and sustain cross-sector partnerships over time.

## Speakers

- **Anita McGahan**, Senior Research Scientist, The Burnes Center for Social Change

Part of the series [Innovating in the public interest](https://innovate-us.org/workshop-series/innovating-in-the-public-interest).

Recording: https://www.youtube.com/watch?v=GhBfGqWHsFw

## Transcript

Automatic transcript of the recording: names and terms can be misheard, and the times can be a few seconds off. Each time links to that moment of the video. The transcript alone: https://innovate-us.org/learning-and-pivoting.txt (plain text) and https://innovate-us.org/learning-and-pivoting.json (every segment with its start and end second).

**Anita McGahan**

[0:03](https://youtu.be/GhBfGqWHsFw?t=3) Welcome to Innovating in the Public Interest, partnering for public good. My name is Anita McGann. I'm a senior scientist at the Burns Center for Social Change at Northeastern, and today's session is the fourth in our series on learning and pivoting. In general, our workshop series is about supporting you as innovators that want to make lasting change that improves lives, and we've been talking over the course of several workshops that are now posted online, if you're interested, but we've been talking about this idea of how to get started and how to be effective in actually implementing change initiatives, especially change initiatives that involve AI. So far, what we've covered is getting started, sort of phase one is around how do you even conceptualize what you're trying to do and get started, build your team. How do you achieve some early wins, make sure you have the evidence you need to get past skeptics and to persist with your idea.

[1:08](https://youtu.be/GhBfGqWHsFw?t=68) And then now we're in a sort of second phase of our thinking about getting going. So how do you actually set up an organization? How do you make sure that organization can support innovation? And now how do you learn on the run? In other words, how do you set up an organization so that you can adjust and learn and be effective as problems start to be revealed to you and opportunities crop up? The next couple of lectures in this series on growing and scaling organizations will come up on May 15th and September 26th, and we're going to be focusing especially toward the end of our workshop series on gauging your interest and teeing up a series of in-person workshops that we can do in September on implementing these ideas, and you'll hear more about that in just a few weeks. So please stay tuned and please return for a growing and scaling.

[2:09](https://youtu.be/GhBfGqWHsFw?t=129) But for now, let's talk about learning and innovating where we are now. So we've talked about building your team and narrowing your focus and thinking about how you're going to bust myths and break through to actually tie your agenda to a redefined understanding of the public good and create some early successes for everybody who's involved. We've talked about winning early, which means using gaffers tape and Band-Aids to actually sort of stand up some sort of a prototype or a launch project so that you can demonstrate that what you're doing is going to have a big impact and you can build a shared understanding of what it is that you're actually trying to accomplish. And it's only after you go through all that kind of phase one work that we've recommended that you start to actually do the work in the way in which your ambition led you in So last time we talked about three different types or styles of innovation. the first place.

[3:10](https://youtu.be/GhBfGqWHsFw?t=190) One was workflow redesign. The second was strategic innovation at the level of a unit. So workflow redesign occurs entirely within a unit. It's very well controlled. It's within a set of activities that may already exist in which you're trying to do is make them better, make them more efficient or higher quality, improve them. Strategic innovation is also within a unit, but there are trade-offs now that you're dealing with strategic innovation. You're saying, we're going to accept a little bit more, let's say risk of organizational problems in exchange for a big boost in efficiency, something like that. So there are trade-offs that you're willing to accept when you do strategic innovation. And then we talked about structural innovation, which involves doing something completely new or blowing up an existing organization and basically hitting the reset button. So we're going to continue to talk about organization design. And the core message from today's workshop is that you want to design your intervention So you want to make sure that your organization is set up in a way that allows you to adapt for learning.

[4:25](https://youtu.be/GhBfGqWHsFw?t=265) and pivot and adjust when you hit unanticipated problems, or you find that your work is adjacent to a really big opportunity, and if you just change a little bit, you can really take advantage of that. And I want to unpack that for you and talk a little bit about what that looks like and how you actually accomplish that. So that's our agenda for today. It's to think about innovating as part of the organizational design process. So to get going, I wanted to remind you of what we talked about last time about why we have organizations at all and exactly where these different levels of innovation are relevant. So organizations are hierarchical by nature. They involve teams, multiple people that are unified in some way that is designed to create a whole that's greater than the sum of the parts.

[5:26](https://youtu.be/GhBfGqWHsFw?t=326) As I said, workflow redesign, which is pervasive across organizations, is opportunity established that happens within established processes. It's kind of win-win. When you're doing workflow redesign, there are no losers. You're basically sticking with your current performance output metrics, your current goals, and you're just trying to improve the workflow. I can tell you as a university professor, there's about a zillion ways within universities that we can improve workflow design, and that's because many of our processes were designed even pre-computer, much less pre-AI. And if we just automate some of the stuff we do, we can get much better performance. It's quicker, it's more accurate, it's more efficient. That's an example of workflow redesign. Think of the allocation of students to classes. We used to do this on paper. If we can automate that, if we can do that in a smart way, that improves the lives of the students, the administrators, the teachers, everybody wins.

[6:29](https://youtu.be/GhBfGqWHsFw?t=389) That's an example of workflow redesign. I think innovation is a level more ambitious in the sense that you're really reconfiguring workflows to achieve really significant advances on some level of performance, like an order of magnitude improvement in efficiency, but there's a compromise that's involved. So somebody somewhere in a process has to give up a little bit of some kind of performance in order to get this big boost in some other area. I'm going to give you lots of examples of these, by the way, in just a second. And then structural innovation is organizational change at a very broad level where some units may have to go all together, other units may get created. There's a big change in performance objectives. You're essentially setting aside the mandate of one work unit and amplifying the mandate of another and trying to achieve a kind of wholesale reset of what the organization's overall structure looks like.

[7:39](https://youtu.be/GhBfGqWHsFw?t=459) So we've pulled these different approaches to organizational innovation into these three And you'll remember that workflow redesign involves point solutions to problems. different buckets here. We're talking about a two-year horizon for payback, maybe a 10 or 20% productivity gain, which is huge. But the structure stays intact and the measures of what is valued, like efficiency or accuracy in assigning students to classes, those measures of performance stay intact. In the middle lane on strategic innovation, we're talking about system change, changes in the ways that values are trading off against each other, maybe more efficiency, but maybe more risk or even lower quality in some areas. But huge productivity gains become possible, 200 to 500% productivity gains with AI. We're talking about a couple of years to deploy and maybe a 10-year payback here.

[8:40](https://youtu.be/GhBfGqWHsFw?t=520) And then on the right-hand side, we looked at structural innovation. This involves 300% or more productivity gains, five years, up to five years to deploy, 20-year paybacks. It's really fundamental change. You're giving up some units of activity here, but you get a ground up resetting on your priorities as well. As we get going on this, I want to remind you why we tolerate organizations, even in the first place, given that communication and coordination are so difficult. I've done a number of studies on firefighting and policing and healthcare delivery, in which this fundamental question comes up, which is why do we need organization at all to do certain tasks? Maybe we would be better off just letting some individual make a choice, kind of a classic example is setting up the schedule for garbage collection.

[9:43](https://youtu.be/GhBfGqWHsFw?t=583) It turns out that everybody wants their garbage collected on a Monday. And if you try to come to a compromise, you can't easily do that because everybody just wants Monday. So better to let one person just design the system and plan for reorganization of the days of the week every once in a while to make it fair here. By giving the schedule to a committee, you create a lot of onerous meetings and there's no obvious way to resolve people's differences and preferences through meetings. And so the thing just takes on a life of its own and becomes burdensome unnecessarily. So you really only want organizational processes to take over on some initiative when it's worth it here. Organizations are essential for taking inputs and translating them to outputs when we need multiple levers of expertise, when there are multiple interests, when we have complex values that need to be weighed against each other.

[10:54](https://youtu.be/GhBfGqWHsFw?t=654) So the answer is that organizations knit inputs into outputs through the implementation of routines, routines and what we're calling workflows. And what these workflows and routines do is they normalize the ways that values are brought into action. So let me give you a couple of examples here. This is relevant for a conversation about innovation. With firefighting, half of casualties are to firefighters that run into dangerous situations. Casualty risk has to be managed. So fire departments, for example, in the United States and Canada, get insurance, there's technical training, there's a lot of conversation, a lot of work that's done to figure out what's the right command structure that is appropriate for sending firefighters to put their lives at risk in different situations. Once those routines are settled, once those conversations happen, and the organization is established, and the processes are set forward, then we normalize sending firefighters into burning buildings, and we accept the risks because they've been vetted, they've been managed, and we need to be able to deploy firefighters into firefighting situations without having each time to have a conversation about whether it's worth it.

[12:20](https://youtu.be/GhBfGqWHsFw?t=740) We've preset those criteria, and those criteria are organized, and therefore the organization can take action. So what we've done is we've translated values into a set of processes that may involve some compromises but which are acceptable to the communities that are relevant, including the firefighters themselves and the communities they serve. And we can move forward now with the deployment of firefighters as a result. Same idea can happen in policing. When is it appropriate for a police officer to use force? You think through that problem in advance, you make the trade-offs, you think about how to mitigate them, you come up with rules and establish procedures, and then you allow the police to do their work. In healthcare delivery, the risks are extensive, but we have protocols, we have procedures, and individual physicians and doctors are not held accountable for bad outcomes if they can show that they were rigorous in following procedures that were pre-vetted and accepted as best practice and agreed upon by the community.

[13:27](https://youtu.be/GhBfGqWHsFw?t=807) So the idea of organization is it supports the taking of action when there are these complex trade-offs and when expertise needs to be interconnected to move forward here. And these trade-offs occur because efficiency doesn't always deliver quality. Quality may be trading off against reliability, reliability against safety, and so on. You know, transparency. If you have procedures that expose who's done what, then you can see where the rules were followed and that becomes the basis for accountability. You're not accountable for a bad outcome if you've followed that procedure and that procedure is transparent and somebody can track what's happened in the case of a bad outcome here. And all of this involves learning. So we're going to turn now to this issue. I could spend, by the way, an entire course on this topic.

[14:29](https://youtu.be/GhBfGqWHsFw?t=869) In the university environment, I've taught, you know, semester-long courses on how organizations are designed to manage these trade-offs and how learning needs to be incorporated into So I'm going over a lot of material in a very short period of time, but I want to give you process design. kind of the basic idea here. Organizational learning enables the establishment of these processes and the achievement of, the deployment of an organization's resources to, to, create value for the stakeholders that are relevant here. Now, I want to make a distinction between learning at the level of an organization versus individual learning. Learning at the organizational level is built into workflows, routines, habits, and ways of doing things. For example, we would say that firefighters have learned that firefighters have learned that insurance is needed to protect firefighters from catastrophic harm.

[15:34](https://youtu.be/GhBfGqWHsFw?t=934) We have learned that healthcare workers need checklists to cover drug interaction. Each individual that's involved in firefighting and healthcare may also have a lot of learning. In fact, it's their expertise, their learning, their skill at a particular domain, and their uniqueness in their having learned that skill that makes them valuable to the organized effort. But those skills and unique capabilities are different from the organizational learning that's built into the routines. And we see this sometimes after incident reviews in policing and firefighting, for example, when somebody will look at a situation and say, what can we learn from a school shooting? Or what can we learn from a catastrophic fire about how to improve our organization design so that we can do better next time? Organizational learning is learning that's deployed in the organization structure and the routines that constitute the workflows.

[16:39](https://youtu.be/GhBfGqWHsFw?t=999) So when we're looking at workflow redesign, and think of AI here, if you think of AI in this context here, we look at workflow redesign that involves AI, there's this kind of fundamental challenge that's new, that we have to confront now, which is that AI is AI because it involves unobserved learning, machine learning, that where there's programmatic advance inside some sort of a training algorithm that guides the AI that is not observable. In fact, it's the lack of observability that allows the machine learning to occur at scale. So there's this challenge, this new challenge that arises with AI enabled implementation of innovations. Think first at the level of workflow redesign of how we understanding internally how these workflow processes are trading off the different values that are important in our processes.

[17:51](https://youtu.be/GhBfGqWHsFw?t=1071) So thinking first about organizational learning in the workflow redesign problem, there's tremendous upside potential for AI. Estimates suggest that about up to 40% of efficiency is lost with each layer of hierarchy in an organization, and that the benefits of better coordination can be really worth it. So what does this mean? If you can do something personally without having to rely on an organizational project that surrounds the task, organizational process and problem, you probably should do it because coordinating with other people is time-consuming, onerous, bureaucratic. Turns out it's really hard to communicate effectively.

[18:53](https://youtu.be/GhBfGqWHsFw?t=1133) There's a lot of evidence to this effect here. So why do we tolerate organizational hierarchies? It's to enable the transfer or the integration of expertise between individuals into an organizational process that supports its deployment here. So you're coordinating among experts. So thinking of firefighting again here, you may have different experts on a firefighting team. Someone may know a lot about certain kinds of building materials. Other people may be great at managing certain kinds of chemicals that could be deployed against a fire. There could be hazmat people and so on. And you have an organizational protocol for when each type of capability or each person's expertise can be deployed in a particular situation. And you're gonna have other organizational protocols that describe how that person needs to report on what they're doing and how that information is gonna be exchanged with other experts and how other experts can use information about say chemicals that are deployed in a particular fire incident to make decisions about hazmat deployment and things like that.

[20:11](https://youtu.be/GhBfGqWHsFw?t=1211) So the hierarchy is creating a standardized process that enables the expertise to be deployed. If AI allows the chemicals person to do the hazmat work too, let's say, it may be worth it to take that problem of mutual deployment and just leave it to that one individual. So there's a level at which AI can actually diminish the need for organizational process. And thinking through when that's safe, when that's appropriate, when that's efficient, when that's worthwhile in workflow redesign is one type of learning that we need to confront as a community and prepare for when you're thinking about workflow redesign. So you may find as you deploy a workflow redesign project that you no longer need the organizational to handle something, that something could be delegated to an individual now because the AI can actually support that implementation.

[21:14](https://youtu.be/GhBfGqWHsFw?t=1274) I was just talking not too long ago with a series of executives that were involved in major construction project deployment. It turns out AI software is really good at deploying scaffolding equipment, can make really great choices about when scaffolding needs to be put up in a particular domain. And so the organization no longer needs a scaffolding decision-making team to make those decisions. They can just get rid of that because the AI can do that more effectively. So coordinating insights between experts can change, the nature of the workflows can be, as you implement AI, can reveal that the efficiencies extend all the way to eliminating certain workflows or work processes altogether here. So there's a series of learnings that need to be considered there.

[22:18](https://youtu.be/GhBfGqWHsFw?t=1338) I do wanna share with you that organizational learning is easy to underappreciate. One of the exercises that I do with my students in the classroom, and I wish I could do it with you here, but I can't do it on Zoom, but let me tell you about it, is to set up the students in five chairs as I've got represented here, A, B, C, D, and E. And I give them each a piece of paper and the piece of paper has a couple of symbols on it, like a division sign or an ampersand or a percentage sign or something like that. And the five of them together without talking to each other have to determine which of the signs they hold in common, but only the person in chair number A knows that that's the goal. The others don't know what the goal is. And this represents a hierarchical system where the leader is the one that knows the goals and then everybody else on the team has to contribute the knowledge they have or the information they have, but they don't know what the overall overarching task is.

[23:21](https://youtu.be/GhBfGqWHsFw?t=1401) How many minutes do you think it takes the average team to figure out what the task is and then accomplish the goal by passing notes to each other? You wanna take a guess in the chat? If you had to do this on your own, you could probably do it in 30 seconds. Let me see what I got in the chat here. 10 minutes, that's an underestimate. 30 minutes is an overestimate. It's usually about 12. It takes about 12 minutes. 17 is a good estimate. It takes about 12 minutes for these five people to do something that a single individual could do in 30 seconds. So it's just a illustration of the onerousness of having an organizational hierarchy. And that means that if you're going to have a workflow that depends on hierarchical exchange of information, that it better be worth it. It better be worth it to have that organizational process. Now, if it's worth it, the challenge of workflow design is represented by this model, which it dates back to the 1970s.

[24:36](https://youtu.be/GhBfGqWHsFw?t=1476) It's called the learning curve. And what the learning curve basically says is that the cost of doing some piece of work, like let's say deploying a checklist in a surgical center to make sure that everything is being considered as a patient comes in, let's say for a knee replacement or something like that, we'll work through all these checklists. The costs of implementing that checklist process go down as your volume accumulates. So the more experience that you have actually doing something, the more efficient the organization becomes at doing it each time. You and I may share an experience on airplanes. I don't know if you've ever seen a pilot work through a checklist on takeoff. I have a friend who's a pilot and I said to him, you know, you must go on 10 flights a day and you're a professional pilot. You must have gone through this checklist a million times. How come you still work through the checklist every time you pull a plane onto the runway?

[25:37](https://youtu.be/GhBfGqWHsFw?t=1537) Don't you have this memorized? And he said, the problem is that I have it memorized. I want to force myself to actually pay attention to every single thing on that list to the point of actually physically making a check mark to make sure that I'm taking nothing for granted every time I take off that airplanes. Because the trade-off is that if I'm wrong, even once people's lives are at stake and that's not something that I could live with. So, you know, you want to think about how we're going to get down a learning curve with an optimized workflow process as you introduce new technology like AI into a system. What we're doing is we're intervening on routines and habits and processes and standardization. We're intervening to disrupt the historical negotiations and meetings and checklists and reports and accountability frameworks that everybody has developed to optimize this workflow.

[26:39](https://youtu.be/GhBfGqWHsFw?t=1599) And we've got to think about what we can learn as AI is implemented to retain the strengths of those workflow processes while we learn how to further make the ones we need more efficient and how to set aside the ones that are no longer needed here. So you want to think about how you're going to embed learning in your structure as you launch your project, even to do a workflow redesign. And that means constantly measuring the efficiency, measuring the quality, thinking about what the values are that are embedded in that workflow design through the heritage processes that have occurred, measuring, measuring, what are we learning? What are we learning? What is the expertise that's being expressed by the AI? How are we going to deal with hallucinations by AI?

[27:39](https://youtu.be/GhBfGqWHsFw?t=1659) How are we going to validate that the learning that's being deployed is appropriate here? Lots of workflow design is built on trust, but how do you trust the machine? You can't, you have to validate it. You have to be careful. You have to very carefully understand how the machine is acting in the expression of AI outcomes with regard to those underlying values and make sure you're sticking with what the workflow is designed to deliver here. So AI may outperform the established experts or introduce new expertise, or even eliminate the need for some types of workflows, but there has to be validation. What is that validation going to look like? What are you going to be learning? What's going to be moved from individual learning to organizational level learning? What's going to be moved the other direction? What are the wild cards that are going to be thrown Maybe we've got some piece of AI, let's say, into coordination?

[28:40](https://youtu.be/GhBfGqWHsFw?t=1720) let's say you're doing colonoscopies in a hospital and the AI identifies an order of magnitude more candidates for polyp removal than the endocrinologist might have found on their own, than the pathologist might want addressed. What do you do? You've now got a patient on the table. The AI is saying that there's more candidates for cancer than have ever before been addressed. How do you deal with that? You know, that may be a wild card in that kind of situation. What's the new learning curve that we've got to deal with? All of this is reason to start on values. You've got to think even in a situation what are the underlying goals of the workflow of workflow redesign, and how are they being affected by the AI implementation that you may be advocating or any organizational redesign that you may be advocating? All right, I don't have forever.

[29:40](https://youtu.be/GhBfGqWHsFw?t=1780) So let me move on to the next topic, which is strategic innovation. And just to remind you, strategic innovation is thinking about a workflow redesign on steroids. So we're thinking about, you know, within individual units now, you know, changing workflow. Those in a way that put different weights on the outcomes to the point where some level of performance needs to be compromised a little bit in order to achieve a big improvement on performance somewhere else. So for example, in many strategic innovation situations involving AI, you're looking at a 500% potential improvement in productivity, but you have to compromise a little bit on something let's say like quality or safety or risk profile or something like that.

[30:41](https://youtu.be/GhBfGqWHsFw?t=1841) So unlike workflow redesign, strategic innovation involves some compromise and you've got to as an organization be ready for that. So you're weighing trade-offs to sacrifice what used to be valued in a slightly different way And it's not just disruptive technically, and that's disruptive. it's disruptive culturally and it's disruptive analytically and stakeholders may be confused. You've got to think about whether or not the core systems that you have for measuring and evaluating and assessing performance are fit for purpose. Some of them may have to be redesigned. We may need new measures of outcomes, new ways of understanding what's happened. So let me give you an example. Back when I first moved to Toronto about 20 years ago, you saw police on the streets. Police would walk the streets, would talk to the community, members of the community. You would see cars in alleys. You don't see that anymore now. It's, I don't think I've seen a police officer on the streets of Toronto in my neighborhood for years actually, not since the pandemic.

[31:50](https://youtu.be/GhBfGqWHsFw?t=1910) Now policing is very much more about surveillance. It's about getting access to cameras. It's about deterrence through clarity on the streets about what's being surveilled, what's being monitored. And it's about negotiating access to different surveillance techniques and different surveillance materials that may trade against bias. Surveillance is notorious for incorporating bias and for occasionally making mistakes with great consequences here. So what you're doing in this framework is you're giving up, you're asking community members to give up some privacy for a lot better information. You're asking people to give up a little bit of maybe personal privacy in exchange for personal security, for more general levels of safety.

[32:51](https://youtu.be/GhBfGqWHsFw?t=1971) Now there can be unintended consequences of that, elevated levels of bias, of institutionalized bias. And that is an unintended consequence that may need to be addressed. Compelling arguments will arise for preserving the old ways of doing things. People will say, hey, police on the streets can offer judgment. They know the people. They know the community. They know the community's habits and conditions. And that's very valid. That actually is absolutely true. And we have to find some way to learn how to reproduce that knowledge in the current system if we're going to move toward a surveillance-based system or to mitigate it some other way, to introduce some other check and balance in the community that prevents an unacceptable outcome. And we have to learn about what kinds of technology, what kinds of interventions can break those trade-offs and allow us to accomplish what's important.

[33:59](https://youtu.be/GhBfGqWHsFw?t=2039) You've got to think about external consultation and how much of that is possible and preferable in different systems. So as you go in and try to replace one way of doing things with another way of doing things within a unit got to think about how much external consultation like I've just described, with the community is possible and preferable for understanding how to manage those trade-offs. And where are the compelling arguments for change? And how are you going to get the community to feel confident that those arguments have been considered even when some individuals in the community don't like the choices that have been made? So you've got to think about and learn about how responses can be addressed in ways that enable the innovation to occur, but take seriously the compromises and their implications for how things should be done effectively.

[35:06](https://youtu.be/GhBfGqWHsFw?t=2106) I remember going to a university that was in Asia and being blown away by the technology that had been developed in this university's engineering school for intervening in the systems of aircraft in the air. And the technical team that was behind these innovations wanted to deploy them for airline safety. But before doing that, the regulator very, very sensibly, I think, said, wait a minute, these systems could also take down an aircraft. They could be deployed for good, but they could also be deployed for bad. And we have to think through how we're going to curb that potential, how we're going to stop malfeasance before we can allow this project to move forward. And that kind of intervention is something you need to think about. How are you going to learn before some sort of catastrophic compromise is acceptable?

[36:09](https://youtu.be/GhBfGqWHsFw?t=2169) So we've got to get ahead of that. So we've got to think also about the implications for personnel. Are you going to retrain people? Are you going to replace them? Costs can escalate if we don't realize the efficiencies associated with a potentially high productivity if the cost savings are not realized. AI intervention, So if AI, I read in the newspaper this morning that college level jobs, jobs for recent college graduates are being replaced by AI. What if you're running an organization, let's say a financial institution or some sort of government agency where you've historically assigned a lot of report writing and measurement and analysis to new college graduates that you've recently hired and you want to replace those people with AI. Well, you actually have to think about what am I going to do to fulfill my commitments to recently hired workers that I no longer need?

[37:18](https://youtu.be/GhBfGqWHsFw?t=2238) Am I going to retrain those people? Am I going to replace them? you may be doubling up If you don't do that, on the deployment of capability, mucking up the organizational change and not providing a kind of decent, long-term sustainable career path to people that you previously committed to. So what's the best solution to that problem that needs to get thought through in order for the organization to achieve the efficiencies that are at the heart of the promise of AI. Same thing happens on the customer side or the service side in public agencies. How are you going to actually discern which clients can be served and whether or not there are some people that no longer can be served? This happens a lot in public and private service. You know, think about healthcare. How are we going to meet the core needs of really complex, in patients with really complex conditions when the AI is designed only to address very specialized conditions?

[38:27](https://youtu.be/GhBfGqWHsFw?t=2307) So let me go back to my knee replacement candidate, somebody that comes in for a knee replacement, but that person, let's say, is complex, has complex conditions. They have cardiovascular disease, they have obesity, they have diabetes. The reason they need a knee replacement is connected to these other conditions. The AI is designed to do a super effective job at the knee replacement, but if the underlying complexity is not considered, the knee replacement surgery enabled by the AI may actually lead to much worse health outcomes than if we adhere to the old manual processes. So how do you identify that possibility in advance and learn how to deal with it and to create contingent systems that are designed for that? So strategic innovation, you've got to think about how much instability When do you want resilience in the system? can you tolerate?

[39:29](https://youtu.be/GhBfGqWHsFw?t=2369) When do you want to allow the outcomes to become manifest? Certainly we don't want bad knee surgeries or airplanes to be affected adversely while they're flying in the air, but maybe we can tolerate some other kinds of organizational compromises. Let's say we can afford to allow some students We can let the registration system fail sometimes not to register. for registering students for classes, knowing that those students will then go to the registrar's office and get the problem rectified. It's a remediable problem in that last instance compared to the catastrophic outcomes that would arise from the airplane and knee surgery examples. So you've got to think about resource allocation as well. What are the trade-offs, the implications of the trade-offs for resource allocation? How should we think about shifting values and how they diffuse across units?

[40:30](https://youtu.be/GhBfGqWHsFw?t=2430) Surveillance and privacy, our initial concerns about surveillance and the kind of big brother nature of having cameras every place have somewhat been set aside in many corners. Many of us, our lives are not so much changed that it isn't worth it for us to have that surveillance because of the benefits that it gives us, the conveniences that it creates and the safety that we get from it. So how are the values in a particular community evolving as experience with a new innovation process unfolds? Okay, I'm gonna run out of time before I run out of things to say. So let me keep going here. Let's talk about structural innovation. You remember that this is a situation where one unit may take over for others. You may get this sort of massive amplification of one type of activity at the cost of having to abandon others.

[41:36](https://youtu.be/GhBfGqWHsFw?t=2496) There's a sort of fundamental organizational or structural redesign that occurs with structural innovation. You're setting aside entire units of activity to replace them with others. One of the classic examples that's happened in a lot of small towns is the complete and total shuttering of libraries, local libraries. Of course, this hasn't happened everywhere, but in some communities, local libraries have shut down, even police stations have shut down. And in exchange, there's sometimes a reliance on either regional or neighboring large cities, regional communities to get services over the internet. And you can check out a book online as opposed to, and have it sent to you or have it delivered instead of having a local library. So thinking about how that kind of transformation occurs is very, very important if you're thinking about implementing AI or some other technology, for example, that could deliver library services into a remote community.

[42:42](https://youtu.be/GhBfGqWHsFw?t=2562) There may be relationships. Let's think about libraries. There may be relationships. Think of the Canadian North where many communities are very isolated. You're gonna shut down that library. What is the function of that library in the community? Is it only to provide people with access to books? It may be a place that elders gather. No. It may be a place where childcare or afterschool care occurs. It may be a community center, periods of emergency. Gotta think about the relationship between the library, let's say, that you're considering closing and other services that are offered or needed in the community. And those relationships, which I'll refer to as synergies, really drive a lot of organization design. And it's easy to underestimate them. So achieving synergies, it requires massive historical investment, lots of conversations, lots of assessments in communities over time.

[43:43](https://youtu.be/GhBfGqWHsFw?t=2623) We know that most changes in organization design like mergers and acquisitions and divestitures and partnerships are beset with failure, that there's a lot of problems that arise that are easy to miss in your planning. And that changes in these different kinds of arrangements across organizations can cost like civil war, battles internally, fiefdoms emerge, proprietary access emerges. You're gonna take away my library? I'm a librarian. You're gonna take away my job? I'm going to refuse to share the information I have about who comes into the library in the afternoons and which programs are occurring with those stinkers over at the community center. I'm not going to share. And there are lots of examples of fiefdoms and internal battles that'll emerge. I've got two of them here from the private sector. Walmart tried to share its distribution systems between the Walmart stores, the discount stores and Sam's Clubs.

[44:49](https://youtu.be/GhBfGqWHsFw?t=2689) And that was really a very, very challenging. JP Morgan integrated with Chase and people were former Chase and former JP rivals Competitive rivals now are expected to work together while they use the opportunity for their integration to cultivate old grievances and to resolve old battles. It was really a tough problem for people there. Libraries versus schools compete for resources in city government. In federal government, the Department of Transportation may compete with the Department of Health and Human Services for resource allocation in the congressional budget. Competitive battles between different units can impede the sharing of data, the sharing of resources, the centralization of resources, and we know that technology is very centralizing of decision-making authority. Learning how to manage synergies can involve massive organizational questions that need to be anticipated and managed.

[45:53](https://youtu.be/GhBfGqWHsFw?t=2753) Think of a classic AI intervention or analytics intervention, which is to share data across different units. Well, perhaps not surprisingly to you, but surprisingly to me, it is astounding how resistant many organizations within even an established level of government, how resistant they will You know, my data, I went to town to gather it. be to share data. It's the basis of my reason for existence. I'm not sharing with you my proprietary data. It's unsafe. The customer won't want it. It's going to make fraud vulnerabilities and so on. All sorts of reasons and logic for why, let's say, IT services and data sharing should not move forward, for example. There are these weekly all-hands meetings called ground rounds that are held in many hospitals where people from different divisions of the hospital are supposed to share their best practices.

[46:58](https://youtu.be/GhBfGqWHsFw?t=2818) I'm sure you will be shocked to hear that there's a lot of bragging that goes on without a lot of detail on exactly what has been accomplished because there's competition between these different units for authority and status and legitimacy and money within the hospital. So overcoming that kind of resistance to integration and the achievement of synergies through technology is a massive challenge in the implementation of innovation projects. You know, when one unit has to give way to another altogether, you can get these dog-eat-dog battles. I'll never forget hearing this story that emerged from the famous outfitter called L.L. Bean in the United States. This is a historical example of a company that made outdoor equipment, outdoor clothing. And the L.L. Bean procurement officer that was buying, let's say, canoes, I'm making this up, but imagine they were buying canoes from a particular vendor, was supposed to transfer their relationship through a referral to the internet sales division of L.L. Bean.

[48:12](https://youtu.be/GhBfGqWHsFw?t=2892) Well, instead, this particular individual told the canoe vendor, you know, don't talk to this other guy. He's going to say he's from L.L. Bean. Don't talk to him. I'm the person that I want you to deal with here. So there's just stories about how internal, you know, dog-eat-dog battles can emerge. At Blockbuster, the famous retailer that rented videos from storefronts, there was a true belief within the organization among leaders that had come up through the retail stores that retail was valued in communities more than streaming services. So you've got the leadership team at Blockbuster sincerely believing that the retail experience where we used to compete with our neighbors to try to get access to the new releases wall at Blockbuster, that we liked doing that better than being able to, you know, choose from a thousand movies through a streaming service. So you get, you know, these kind of lock-ins to beliefs that reflect historical investments in particular activities that are very, very difficult to dislodge.

[49:25](https://youtu.be/GhBfGqWHsFw?t=2965) So thinking through in advance how to overcome that kind of resistance is very worthwhile. You have to learn what synergies do we want to protect? Well, maybe, you know, we don't want to protect Blockbuster's retail stores, but maybe we do want to protect, you know, the proprietary data from fraud in a particular governmental unit that's being asked to share. So when do we, when do we protect, you know, when do we mitigate? When do we, when do we try to kind of conserve, you know, and hold a bar on what might be a compromise? And when do we want to just, you know, let the cards fall as they may on the irrelevance of some old activity here? Again, it comes back to values. So learning is arising around things like what's our risk tolerance? What kind of cultural change do we want to try to drive here? What kinds of values do we have? You know, how are we going to retain our talent?

[50:28](https://youtu.be/GhBfGqWHsFw?t=3028) How are we going to compete effectively as we go through these kinds of transitions? And why is our organization, why is Blockbuster the best place to do streaming as opposed to Netflix or vice versa? What is the best option for Blockbuster? Maybe it's not to do streaming. Maybe it's to offer different kinds of retail services than only video delivery. What is the organization best positioned to deliver given the trade-offs in the marketplace, among competing services? Maybe the library is better positioned as a community center than the old community center, than a sports arena. If what's happening, if the community's needs are around, you know, elder programs and childcare, do we need that to happen in the sports arena? Should we repurpose a library as a community center? Those kinds of conversations, which involve non, kind of non-obvious alternatives for deployment of old capabilities, need to be addressed if innovation in a library to move all of the book lending online is going to be successful.

[51:46](https://youtu.be/GhBfGqWHsFw?t=3106) So thinking about how you learn which resources can be shared and about true performance and measuring effective research sharing is very important. I'm going to skip this point on the innovator's dilemma. We'll come back to it next time. But it's basically, it reflects this idea that if you've got high performance and stability and efficiency on a well-vetted, well-shaken out model, and you're faced with the possibility of a big improvement in performance, but you've got to go through a period of uncertainty and risk and instability, you've got to think about whether it's worth it. So that's the challenge. Got to think about how much can be gained by moving to a new approach, especially an AI approach that's black boxed, versus the instability and the different types of problems that it may create in the community en route to that outcome. So we've got to, the barriers to technology adoption tend to be organizational and cultural.

[52:48](https://youtu.be/GhBfGqWHsFw?t=3168) They tend to be very important barriers to technology adoption tend to be organizational and cultural. This isn't to say there aren't technical choices that are important, but it is the organizational and cultural challenges that are a massive gateway to successful implementation. And they constitutes bricks walls. They can be roadblocks. They could be structural impediments to change. So what happens if you're doing a change initiative and you hit one of these brick walls? Move to a learning posture, say, you know what? We're not going to be able to close the library. Let's start to measure what happens in that building. Let's start to think about not just the workflow around book lending, but let's retreat to a learning posture on all the other stuff that's going on within those walls. We're not going to be able to make a transition, let's say, to AI based knee replacement surgery.

[53:54](https://youtu.be/GhBfGqWHsFw?t=3234) Let's start to think about all of the different interactions and complexities in a diabetes cardiovascular patient that needs a knee replacement surgery. Let's start to document. Let's write down everything that we need to be considering. Let's think about how we're going to learn so that we can come back to the change agenda after developing more understanding, deeper understanding of answers to the kinds of questions that I've put before you today. All right, let me keep going toward the end here. And we're going to talk next time about growth, different pathways to growth. Many people think about growth internally, but there's other types of growth, like, for example, supporting your peers in other cities and other agencies and other parts of society that may adopt some of the innovations that you want to champion. We'll be then moving to scale and again, talking about our workshops that we're going to be doing in September to support those of you who want additional help.

[54:59](https://youtu.be/GhBfGqWHsFw?t=3299) Thank you so much for your kind attention. And Agata, let me turn this back to you. Could you please turn off the recording and we can perhaps have a little time for a Q&A.

* * *

Original page: https://innovate-us.org/learning-and-pivoting · Licence: CC BY-SA 4.0 (https://creativecommons.org/licenses/by-sa/4.0/), as stated on innovate-us.org
