Praxis Product Management

It’s March 2023, and like every other point in history, a lot is happening in the world. Presently, two macro-factors capture my fancy:

1. The AI-of-everything kicked off by the ChatGPT unveiling in December 2022, and
2. The Fed’s years-long raising of interest rates to present levels (4.75%) that have not been seen since 2005.

Don’t look now—but I believe these factors combine to create an inflection point not seen since the post-dot com bubble.

Why this matters.

Generally, I write about the process of product management. Which typically covers how to do something (run a QBR, define a product management cohort), the why behind something to be done, or the notes between the two (Case Studies, The Importance of Vision).

Ultimately this ends up being a lot of rapping about rapping. I believe it’s infinitely more valuable to understand how to craft a vision than to write about the Vision documents I’ve drafted and the market insights that led to them. Don’t get me wrong; I genuinely enjoy this aspect. But that’s the act of living/doing product management. It’s not my hat as an author.

Why? Cause I think it’s typically harmful for product managers and leaders to draft off product management work external to your context. When I craft a Vision—I know all the roads we didn’t go down, and I know the weaknesses and warts, the exception to the rule. I would never bring this vision or strategy to another company. It’s bespoke and created in response to the present conditions and future outcomes—the current company is looking to achieve.

I don’t want to examine product work. I want us all to improve at creating product work—and building teams and processes that deliver product outcomes.

But I’d be remiss if I left you with the impression that Product Leadership is all about the process—and doing the work. Or the belief that just because each strategy needs to fit the current context—that there aren’t some gravitational forces we all, regardless of industry, operate under.

Well, gravity exists, And knowing what those forces are—and what they are not. What trends are noise or product positioning, and which are disruptive? This requires judgment.

It is not a question of fact. Both Amazon.com and Walmart looked at the same internet adoption data in the late 90s; Amazon saw a rapidly growing disruptive market—Walmart saw a small distribution channel that wouldn’t be material to the company for at least a decade. Both were right.

Both companies exercised proper judgment for their context—and both leadership teams made near-fatal mistakes. Amazon over-expanded categories when capital was cheap; Walmart waited too long to treat e-commerce as a pillar of their business.

These decisions rest on judgment, the combination of leadership’s beliefs about the future marketplace, the organization’s values, and capabilities. It’s not simply facts and data. It’s not about merely operating the “process of PM”—and calling fact or fiction on decisions that reach your desk. It’s about pulling insights from the world (external to your building, physical or virtual) into your company and team and transforming that into a perspective about the future.

As a requirement, Product Leaders—must have a perspective on the future.

So since I believe both that judgment is important, and we’re at an inflection point. I’m publishing my perspective on what’s going on in the world as both a reference and the thinking you need to be doing as a product leader about the impact of these shifts.

Our inflection point.

The web didn’t reshape the world. Not really. But it was a necessary pre-condition to the disruption that’s occurred since. What did disrupt the world was:

1. Wireless broadband
2. The smartphone
3. The “internet” (web and internet technologies)

A fourth honorable mention is big data/ML—which brought about social media at scale.

But setting that aside, you take the internet—and you trap it in its original form, tied to a computer accessed via a computer browser—what you’ll have is the Bloomberg Terminal for knowledge workers. A tool for the wealthy to do more, a bit faster—when stationary.

And as a result, the previous predictions were generally wrong. The prior predictions were all about the information super highway—connecting with strangers—instead, many of the stupid and invisible hurdles to life got less stupid (Uber v. Cabs), and we created new categories of celebrities. We developed new ways to connect with (and ignore) our friends and family.

What we got was not second life—what we got was life aided by a supercomputer.

Similarly, when looking at AI¹, the breakthrough, in my view, is not the technology—but AI coupled with the other factors that are already underway.

And here we’re at an inflection point.

Institutions, from small to large, have already done much of the heavy lifting of breaking down data siloes. Data moved to the public cloud, and prominent cloud technology stacks to the tune of half a trillion dollars in 2022.² This shift has trained a workforce of engineers to take advantage of and migrate data and technology to take advantage of new vendor offerings from Lambda to Kubernetes.

Insight: Companies are “wired” to unleash new technology, such as advanced language models on internal and proprietary data sets.

Augmented Reality (AR/VR) is still next year’s big thing. I spoke with a PM and Design manager who worked on Hololens in the early days—and one common attribute was how new and different the workflow was in creating assets for AR as opposed to 2D screens. You notice this a bit when shopping. When you can take a chair you’re considering and plop a model next to your dining room table, AR works wonderfully and answers so many questions. Yet very few products have this capability. Generating AR assets differs from snapping a photo and putting it online. Many companies can’t afford the cost structure to create assets during item setup. From inventory to the product page is a human-intensive process, even to get the photos for which customers make buying decisions.

Insight: In categories like shopping, generating AR assets is a separate and labor-intensive workflow relative to other product setup tasks.

Content fragmentation is the present and future. The internet was supposed to be the one great content pipe. Google was to organize the world’s information. But each day, we’re further away from one market to rule them all (The Everything Store in Amazon parlance) to multiple markets (Ex. Shopify as a platform). The open web newspapers have primarily been replaced by two extremes—content farms and paywalls. Similarly, the blogosphere has been replaced by the newsletter subscription. Finally, the ending of the cable bundle wasn’t replaced by the one streaming app to rule them all—but instead by an ecosystem of streaming apps. Podcasting has shifted to… you get the point. Media companies of days old didn’t limit distribution to annoy customers but instead to protect content value.³ I had a front-row seat to this working in Fashion ecommerce—control of who gets your product lines and in what colors (Place) is your negotiating leverage for price and promotion. Control of content holds with corporate data, too⁴—companies will begin to value their proprietary customer information. I surmise that Nordstrom has the world’s richest data set for the buying habits of fashion-forward customers.

Insight: There is not one universal corpus of data to mine—data has boundaries and access to one data set, often prevents access to another (Example. No entity will have unfettered access to both Bing and Google’s corpus, a contract with Google would preclude an agreement with Bing and vice versa).

The cost of capital has increased. In business school—the standard discount assumption was 10%, which says you can earn a 10% interest rate on cash without trying, so anything you do needs to clear the 10% you can make on liquid assets. While a rule of thumb, the thinking is simple—corporations of significant size have resources working to put money to work—even cash—that can surely earn a 10% return. We everyday consumers couldn’t do this—because we don’t have people working to put our money to work for us. In essence, when the fed raises interest, everyday consumers also have a higher return on capital available for near zero risk in treasury bills. At the time of writing interest rate is 4.75%. Effectively, half the way to the target of 10% capital without taking on any risk. When returns with zero risk are this high—it changes how people invest. It makes people view risk much more skeptically. Over the last 20 years, S&P returned 8.19% annually. But sometimes that number goes up—sometimes it’s down. The previous two years have been down years—by a lot. Now you can earn more than half that return without taking on any risk of losing your original value. Two things—this makes cash far more valuable—and reduces interest in speculation for long-term returns.

Insight: Corporations will shift focus to profitability—investors will value dividends—long-term growth plays will discover more avarice investors.

Implications.

Based on these inflections, here are a few implications.

AR/VR headset becomes a thing—and AR shopping becomes normal. I’ve been bearish on AR for well over a decade. All new forms of interaction have a chicken and the egg problem—lack of disruptive content means low adoption of hardware—and the low adoption of hardware means limited R&D investment in disruptive content. Given that the primary value proposition was a content-driven use case in gaming, I didn’t think AR would reach the adoption levels necessary to support sustained content investment. With AI, I believe the cost of asset development materially decreases. As a result, I think AR headsets will develop a material base in about five years, growing to, let’s say, 10% of the active PC install market in 10 years. In 2022, 9.7 million AR headsets were shipped.⁵ AR shopping using a mobile device, which exists today, will continue to be the most prevalent AR use case—shifting how people buy fashion and furniture.

Why? Progress in deep learning models/AI—will reduce the cost of asset development, including but not limited to believable 3D models generated from product imagery and AR avatars.

Organizations rapidly shift to AI everything. Since this is already happening, I’ll instead highlight three constraints on how they will do it—

1. Reducing other tech investments and use AI to bolster existing flagship product offerings. I believe mobile app development’s number and cost structures will be under particular scrutiny. I think mobile app development will continue—at a reduced investment level. Why? Due to the increased cost of capital, the focus on AI integration must be self-funded from the corporate project portfolio.
2. Utilizing proprietary data sets. The AI of everything will be a threat to the open web. The promise of opening up a corpus to web crawlers was the possibility for visitors to follow. Now crawlers are just injecting content for learning—where a visit is not the outcome of a search. I believe we’ll see an increase in paywalls. Why? The decrease in traffic from search visitors funnel reduces the appeal of ad-funded business models.
3. Making stupid things less stupid. For the next five years or so—a lot of the progress from AI will look obvious in retrospect. Think: more like wheels on a suitcase than human on the moon.⁶ Let’s say something as simple as a decent customer service agent. “The computer, we are being told, will replace the decision-maker, at least in middle management. It will make, in a few years, all the operating decisions—and fairly soon thereafter it will take over the strategic decisions too.” The Effective Manager, 1967. The computer didn’t replace work. The computer did many things; we can say most assuredly it made some foolish things about work (making overhead projector slides)—either less stupid or obsolete. I expect that trend to continue.
4. Asymetrically. The integration of AI is going to be asymmetric. Just as we’re in the digital revolution in much of our life—but find ourselves stuck in the dialup era the moment we need to interact with the government or, too often, healthcare. AI will profoundly disrupt a few dimensions—but leave others untouched. I think personal use of AI will become very rich and capable. Siri will be for more than the weather and trusted as a first-hand source too. Additionally, Ecommerce (fashion, for example), Social Media content generation and ranking, Media recommender systems (think Netflix, Spotify), Document creation/Knowledge management workflows—and human business processes are going to get a ton of investment—everything else I think is going to have very little over the next five years. Why? There isn’t going to be an AI switch. Compute costs; AI computing will cost richly, in both expense and skilled labor. Both will preclude AI from going to everywhere all at once. It will go to critical franchises and business processes. Over the next two to three years, we might see it deeply integrated into a broader set of uses (think Alexa microwave) before clawing back to a more stable equilibrium in the next five years as costs align.
   

Finally… AI will unlock new business models. I’ll call this one corpus brokering. Everything from academic journals to stock images, industry forecasts/reports, and books—will become valuable corpus for AI training and information retrieval. These corpuses’ rights and usage rights will become an essential part of the Value Net of AI technology. For example, AI doesn’t know the most streamed song on Spotify or the influencers on black Twitter. Each corpus will have its purchase license—and the expiration of rights. So copyright litigation is going to heat up.

And this is the beautiful thing about the future—it’s not out there; present actions and variables constantly rewrite it. The movie business looked more robust than ever—then COVID. The gym seemed dead—then recovery. Our product judgment is based on insights, so it’s constantly rewritten. It’s the process of acting on our beliefs that rewrites the future—and if you’re successful, the end turns out a bit more favorably for the things you care about.

And with that, I’ll leave you with two more nuggets about the implications of AI and interest rates.

First, privacy advancements have reduced many of the financial free lunches of ad-funded businesses. The combination of AI of everything, which will increase the value of your data and the economic underperformance of these ad-funded businesses, will put pressure back on privacy progress. Ad dollars were efficient at getting you to part with your money in ways unseen—and they did it at the cost of user data, your data. With budgets crunched due to rising interest rates—time will tell if the enterprise privacy, security, and identity initiatives brought about through CCPA and GDPR continue to garner investment. Similarly—if through the focus on the economy, consumer privacy policies continue to progress. How AI is implemented will also impact consumer privacy—how much computing is in the cloud and how much is on the device. Given that on-device computing requires more capable devices, the privacy divide between affluent and less well-off denizens will increase.

Second—the AI arms race, is really going to accelerate the reconciliation of our techno-dependence and our warming planet. Learning models and their impact on climate was the firestarter for Timnit Gebru’s firing from Google.⁷ This was two years before ChatGPT. As companies look to respond and integrate AI into everything—front and center will be the cost, in both human capital and cloud/services expense—what too often will be absent is the compounded climate cost. Wouldn’t it be great if we zero-summed our climate spending as enterprises, just as we might zero-sum our budgets? If we’re having more climate impact from this—what climate-impacting activity will we stop to pay for it?

It’s important to recall that climate change is the issue of our lifetime—and each of these tools will also be used to define the battlefield of who bears most of the cost in a changing world.

1. Author’s note: basically, everything we’re calling AI is really good ML—but for the sake of comprehension over clarity, I will stick to calling it AI. ↩
2. https://www.statista.com/outlook/tmo/public-cloud/worldwide ↩
3. https://www.theverge.com/23621907/streaming-tv-boxes-roku-amazon-google-apple-nvidia ↩
4. https://mailchi.mp/fdac1ee69bec/benedicts-newsletter-452264 ↩
5. https://www.idc.com/promo/arvr
   For reference, PCs forecasted about 280M units in 2023 https://www.idc.com/getdoc.jsp?containerId=prUS49918522 ↩
6. Can you imagine that it took close to six thousand years between the invention of the wheel (by, we assume, the Mesopotamians) and this brilliant implementation (by some luggage maker in a drab industrial suburb)? And billions of hours spent by travelers like myself schlepping luggage through corridors full of rude customs officers. Worse, this took place three decades or so after we put a man on the moon.
   Antifragile. p188. Nasim Taleb. ↩
7. https://www.wired.com/story/prominent-ai-ethics-researcher-says-google-fired-her/ ↩

---