Host Intro (Gene Kim)

All right. For the next talk, to frame this, is about optionality. Before I introduce speakers, Dr. Robert Merton, who won the Nobel Prize in economic sciences, said the more uncertainty you have, the more you want options, which is the right, but not the obligation, to take an action. When you have no uncertainty, you do not need options because you already know all the right answers. But in times of high uncertainty, you want the option to defer your decision until you have more information. In other words, these are conditions where you do not want to make long-term plans.

In the age of AI, where frankly none of us know anything, uncertainty is off the charts. This panel session is all about how organizations might be making decisions in a way that reduces the number of options, which is very un-Ffo. With that, I would like to introduce our two panelists. We have Steven and David. Please welcome.

Can we start off by having you introduce yourself?

David Rice

Sure. My name is David Rice. I am an SVP at Cox Automotive. I run product engineering for a lot of products across the U.S. And I drive a lot of the integrations of how all of the different acquisitions we made over the years come together.

Host Intro (Gene Kim)

And you actually came in through an acquisition how many years ago?

David Rice

I did, 15 years ago.

Host Intro (Gene Kim)

Fantastic. Steven, how about you?

Stephen Fishman

I am Steven Fishman. I am the field CTO for Boomi. That is a connectivity platform leading in all the analyst ratings for API management, iPaaS, core AI, and data management. I have known Gene for a long time, and I had the privilege of collaborating with Gene and my co-author Matt McLarty, who is not here today, on Unbundling the Enterprise.

Host Intro (Gene Kim)

We actually met at South by Southwest even before The Phoenix Project came out.

Stephen Fishman

That is right.

Host Intro (Gene Kim)

Which is crazy. All right, what caught my attention was when you were talking about how organizations might be shooting themselves in the foot because they are doing things in a way that is actually depriving them of optionality. Can you say more about that?

David Rice

If you think about organizational strategy, organizational strategy is usually rooted in either too narrow of options or too large of options. A lot of organizational strategies that can allow you to do anything are not really helpful for focusing the organization and really honing in on what matters.

One of the things I have noticed over the years is that I have asked the question: why is that? A lot of it is driven by the financial models that actually drive organizations. If you are a public company and chasing the quarter, does your strategy really last much longer than a quarter? If you are even a private company where you are chasing the year, your strategy tends to stay within that timeframe.

Host Intro (Gene Kim)

What really caught my attention is: what if there is a hint of an economic slowdown or an earnings miss? What is the first thing that people do? It really bothers you.

David Rice

Yeah. They cut everything around optionality. They cut your platform plays. They cut your hiring. They cut everything that actually gives you choice in the future.

Host Intro (Gene Kim)

And why is that bad?

David Rice

If you think about having a perfect prediction of the future, you just do not have that. You do not know what is coming. AI has proven that faster than really any other technology change. You have got to build the optionality into the organization so that you have those choices, so that you can pivot when AI comes around.

Host Intro (Gene Kim)

What really made my jaw drop was that you are talking about horizon one, two, and three. At the instant of cost control, you cut your horizon three projects, which are the ones that actually were creating future options. That is exactly the opposite of what we want to do.

Stephen, you had also some opinions on this.

Stephen Fishman

I think everyone here is obviously excited about the AI future that we want to create. At the same time, every now and then somebody will say it, but it feels like we are collectively ignoring that that AI future is built on top of a set of decomposed, fully externalized, activity-oriented APIs. Those are the building blocks that allow the agents to have context, to allow the agents to do things. Investing in those things, that is the estate that generates the options that you have to be able to do things: to mix and match, unbundle, rebundle, and compose as you see fit to find market fit.

Host Intro (Gene Kim)

Also often the first on the chopping block as organizations go into cost-control mode. Before we leave this topic, it makes you angry that these represent a certain lack of planning discipline and operational discipline. Can you say a little bit more about that, just to get the angry David out?

David Rice

All right. Yes, it does make me angry because you think about how a lot of operational leaders get in place: they are really good at operating the business, but they are chasing that moment in time and they are trading off on the future. I have an unusual history in that I have been at this company now for 15 years, since we were acquired, and my last company for seven years. So I tend to stay for a long time. I have always had to live with every mistake I have ever made.

So many leaders do not actually have to live with their mistakes because they are chasing two years, two years, whatever. How do you defend cutting off that future when you are with an organization that you are wanting to survive for the next 25 years?

Host Intro (Gene Kim)

One more point, just in case this does not resonate with you: when you go into a hiring freeze in a time like now, which are the first job reqs that go away? AI jobs? AI skills?

David Rice

It is whatever does not fulfill the need of the moment. That ends up being your early career, which sets it up for the future. It ends up being your AI or your future-focused. It really is like, hey, we have just got to build this product to get to this customer to sell this. It just makes no sense. It makes no sense.

Host Intro (Gene Kim)

If that seems familiar, how many people resonate with some of the frustrations that David is mentioning? Okay, fantastic. Not fantastic. Sorry. Not good. What is anti-applause?

Given the frustration, what is the recipe? What is a better strategy to use during times of uncertainty?

David Rice

You want optionality. That is really what we are talking about here. Why it matters is the interesting thing: unless you have that perfect prediction of the future, you want to be able to pivot to the left or the right. You want to be able to move with the environment. You look at the economic uncertainty that we are in right now, tariffs. I have no idea what they are going to do. Nobody really understands. We do not know if they are going to get pulled back or pushed forward. But it is creating turmoil within the context of my business, which is car buying.

So we have to make decisions from increasing or decreasing behavior. That optionality built into the system from the ground up allows us to quickly pivot and make different decisions.

Stephen Fishman

To speak to development and architects in the audience who might get confused by the optionality term: when we talk about APIs, those are decontextualized things. That is why they have many options for consumption in front of them. The great example is Google Maps. Google Maps has north of $800 million revenue just on the API alone, which was never intended to be a product. Developers found it through the website, did view source, found the API, started calling it at scale, and then called Google and said, we know we are violating the terms of service. Can we pay for this please?

We found that in almost if not every organization that we talked to. We talked about 20 different tech execs from around the world. David was unique in that David was actually intentionally engineering the happy accidents of these big windfalls that all these other companies fell backwards into. It was not just the digital giants. It was Coca-Cola, it was Lowe's, it was all brands. They all fell backwards into it where David was headfirst going for it.

Host Intro (Gene Kim)

To ground this: you believe that these happy accidents are actually deliberate decisions, outcomes of deliberate decisions made around creating optionality?

David Rice

Yes. I think sometimes people luck into them, and Google is a great example where they did that. But you have to be very intentional about choosing to be in a world where you are creating that optionality. It is not just building APIs, because it is very easy to build an API that cannot be used by anybody else because you have not actually externalized it. You have not actually made the decisions from an organization perspective, similar to what Bezos did in his famous email. You have to build APIs. You have to make them externalized. It is the only way you are allowed to connect.

You end up having to make decisions like that as well as making sure that you are building the platform that allows you to expose those APIs in an effective way, that allows you to control and understand scale and how they are reused, and really just set yourself up for that set of choices that you do not yet anticipate.

Host Intro (Gene Kim)

By creating APIs you create these modular structures that create economic value through creating options, through creating choices. That seems very obscure, but it sounds like this is actually something that you are deliberately using in how you make decisions both in strategy and execution.

David Rice

Yeah, and we are kind of unique in that the organization that I am part of was a bunch of acquisitions, but it was a bunch of small acquisitions. We did not have a core base that everything was bought and built into. It is not like Salesforce where you have this big pillar. We bought a bunch of independent things. Figuring out how to actually orchestrate afterwards a platform that you could then create that optionality and connections.

Our strategies for the last really three or four years have all been built on the fact that seven years ago we started building optionality into the core connections of how we flow between our products.

Host Intro (Gene Kim)

Until very recently, this was on one hand exciting, but also seemed just a little bit obscure and academic. We all know that modularity creates huge economic value, but it still seemed just a little difficult to concretize. Stephen, somewhere in the last year I shared with you a formula that blew up my brain, and it sounded like it did the same thing to you. It is from Dr. Carliss Baldwin. What was that formula?

Stephen Fishman

N times K divided by T times sigma. N is the number of modules. K is the parallel experiments. T is the time to experiment. Sigma is the risk and uncertainty. The number of modules, the parallel experiments, and time to experiment are fairly easy to quantify. Sigma, the risk and uncertainty, we all know it is high, but we do not know exactly how much it is. It seems hard to quantify.

But the good news is that you do not need to. You only need to actually understand directionality and order of magnitude. Meaning, an area of high risk: hey, it is probably a good idea to decompose. You only need to know directionality and order of magnitude in order to make the decisions that, hey, clearly externalized by default and decomposed by default are going to play out well here. Because when uncertainty is high, option value is high.

Host Intro (Gene Kim)

Can you say that again because I did not quite follow? N is the number of modules. When you have a lot of modules, you are available to mix and match them as you choose to fit. Amazon in 2001 had one module, 3,000 engineers, and then they had 300 or 1,000. So that is N increasing. K is number of experiments that you can do at one time, parallel experiments, maybe within a module.

Stephen Fishman

Yes. Within a module or within the context of pulling three or four modules together in order to find market fit. APIs are unique little bundles of functions. They are not true products in and of themselves. Bundling them together in six or seven different variations would be six or seven different experiments that you are running.

Host Intro (Gene Kim)

Amazon e-commerce could do maybe one experiment at a time and it would probably take them up, which is the T, how long would it take to perform an experiment. It was probably six to nine months. But by breaking into microservices, they drove T down to days. The time and the number are proxies for the cost of the experiments.

Stephen Fishman

Yes. When you can bring that cost down through APIs, which are inherently modular and easy to combine, and decontextualized services, the cost of experimentation goes down and the time goes down, which means you can run a lot more. I think the Bezos quote is, when experiments are expensive and time intensive, you do not run a lot of them. When they are cheap and quick, you run a lot. That is just the way economics works.

Sigma represents many things, but if you know the answer already, the value of this is very low because you do not even run the experiment. You just know the answer already. But when uncertainty is high, the only way you can find the answer is through experimentation.

Host Intro (Gene Kim)

The flip side is also true: when uncertainty is high, rigidity is existentially dangerous.

Stephen Fishman

Yeah. Because somebody will find a better mix than you.

Host Intro (Gene Kim)

David, what do you think? Are we articulating what you have had in your head for 15 years?

David Rice

Yeah. I wish I had that formula 15 years ago.

One of the things that we talked about before, Gene, is this idea that AI changes this a little bit, but it changes it in an exponential way, and that it reduces the thing on the denominator of time. If you think about the possibility of not only changing the time but also reducing uncertainty, one of the key things that I always think about is MVP. Who has worked out what the MVP needed to be, built out to it, then missed the deadline, and then all of a sudden you were willing to compromise on what was in the MVP? So clearly you were wrong about the original MVP and you redefined it.

One of the things that AI potentially allows you to do is redefine what that is. I have not got the acronym quite right, but it is multiple variants of your product. Think about if you build optionality, if you build your capabilities as modular sizeable, and then instead of actually driving to implement a version of a minimal product, you actually implement multiple variants of the product. You roll it out to several different customer groups at a time. You see how they use it, how they adopt it, as opposed to thinking about it as, oh, I have got to get it right.

Host Intro (Gene Kim)

This brings up something really interesting that David said to me several years ago, where the technical work is hard work, but it is well-defined work, and finding product market fit is harder work because it is ill-defined. It is a process of trial and error.

Something that I was thinking about when you said it was another jaw-dropping moment. The MVP as defined by Lean Startup assumes that the coding development part is the long pole. That is why you have to be so judicious about the MVP. When the cost of production of code goes down by 10 or 100x, how does that change the notion of what MVP is all about?

David Rice

The limiting factor then becomes your customers adopting your ideation and figuring out what it is that they want. The capabilities that you have then exposed mean you can further drop down the cost of MVPs by creating the optionality and then leveraging AI to connect everything together and produce more variants.

Think about your customer base: generally speaking, customers use a small subset of whatever application you give them. Wouldn't it be nice if they could use the application exactly that they want? That is the bundling of your capabilities.

Host Intro (Gene Kim)

To pick and choose like a buffet.

David Rice

Right. Exactly what they want.

Host Intro (Gene Kim)

One thing that David has, a luxury that a lot of organizations do not have, is sample size. We do not talk about in the formula that your ability to keep the time low is by how many different people you can actually bring into the experiment process in order to find that market fit. They have the unique opportunity of having millions and millions of customers engaging every day.

In the time remaining, is there help you are looking for and a last piece of advice that you would want to give?

David Rice

For me, one of the huge things about being here, Gene, was realizing that there are so many people in the exact same place. I am trying to drive transformation from an AI perspective with 5,000 engineers across the world. I am trying to figure out how we actually fully evolve into whatever the next iteration of the full product development life cycle. So anyone who is leaning in and doing experimentation in that case, I want to hear about it. I want to learn about how those experiments go. I especially want to learn about the failures.

The main piece of advice I would give is optionality is something that, similar to security, similar to resiliency, can now be decided to be a core table stake. Every time we build software, we are in a point where the pivot and change of the technology with AI allows us to completely change the time that we predict from a development perspective. Making sure that we put that in as a default expectation and build that way as an industry is going to explode the possibilities in the future.

Stephen Fishman

To follow on what David said, the one thing I am looking for help with is: at Boomi we have a large agent marketplace. We have got thousands of agents in production by use in all of our customers today. A lot of that is focused on agents for developers to use in the process of automated testing, probing, and all sorts of things that developers would want in the process of actually making solutions.

What I am looking for is organizations. I talked with the folks at Tricentis here; I have been in conversation with them for a while to help because we are an open-by-default platform. If you have agents that you are ready to actually bring to a wide group, we would love to talk to you.

The last thing is the title of the presentation. The one point is that because decomposition makes it easier to change, it is technical credit, not technical debt, because you have actually created the space to make change easier, which is the opposite and the flip of how debt makes it harder.

Host Intro (Gene Kim)

Thank you, David. Thank you, Fish. Thank you, Gene.