There’s a weird dynamic in the current housing market in Tucson, Arizona. Almost all listings fall into the following categories:
- on the market less than 21 days, or
- on market greater than 100 days.
It turns out that the vast majority of houses in category 2 are listed by Opendoor, the iBuyer real estate company that IPO’d in 2020 via a SPAC. The company did $8 billion in revenue in 2021, albeit with negative $.5 billion EBITDA.
The idea of Opendoor houses being on the market for so long makes zero sense to me, in light of the core capability that underpins the entire business model:
The entire premise behind Opendoor is that the company can reasonably forecast future market prices. Reasonable confidence in a forecasted future resale price allows them to back into a purchase price today so they can buy a house quickly, do some minor renovations if necessary, then put the house back on the market and quickly capture a profit.
Set aside the forecasting algorithms that are the backbone of the Opendoor model, the company should have some variation of dynamic pricing to adjust prices for active listings in a way that reflects the current, and forecasted, market conditions.
But that’s not what’s happening with Opendoor listings, at least in Tucson. These houses are sitting on the market, getting stale, in a slowing market.
In 2021 Zillow shut down its ibuying business in a blaze of glory, admitting that it “could not predict market prices with sufficient accuracy to support the business model and the amount of capital at risk in inventory.”
I’ve seen similar concerns about used car dealer Carvana, the Opendoor of the car category and another AI-first business.
All of this jumped out at me as I was reading The AI-First Company: How to compete and win with Artificial Intelligence. (Ok fine I’m generously skimming it bc hello 😵💫)
The big takeaway from the book, and observation of companies like Opendoor, is that being an AI-first company creates enormous competitive advantage, or has the potential to do so, but is really hard to execute well.
Today we look at what artificial intelligence might mean for companies in the business of producing & processing livestock, meat & dairy or the companies who provide services to producers & processors.
Some key concepts from the AI-First Company:
“The first wave was tools that brought physical leverage, think rope traps and spears. The second wave brought intellectual leverage in the form of the printing press, calculators, computers.
The third wave is artificial intelligence tools that provide decision-making leverage. These tools are affording us an entirely new form of intelligence that gathers, processes, and communicates information to make better decisions. We’re learning better and faster as we see these decisions play out.”
“When we change what we use to learn, we can change
how fast we learn.”
That might be my favorite point, and quote, from the whole book.
The entire purpose of using AI is to learn more/faster/better in order to drive better decisions and outcomes. Full stop.
“Data Learning Effects are the automatic compounding of information.”
Data Learning Effects are where competitive advantage is created, in the
automated, systematized process of accelerating learning from data.
“AI-First companies put AI to work, prioritizing it within real budgets and time constraints.”
“AI-First companies make short-term trade-offs to build intelligence in order to gain a long-term advantage over their competitors.”
Ok so this is great and all, but how does an established company become an AI-accelerated company?
“The steps to building a data learning effect with intelligent machines are:
(1) capturing a critical mass of data,
(2) developing capabilities to process that data into information, and
(3) feeding that information into a computer that runs calculations over data to learn something new.”
The author also translates that into English:
“Get lots of data, process it into something useful in terms of making a decision, and create a system that automatically generates more useful data.”
You can think of those 3 steps as the Data –> Decisions value chain.
Just like a Cattle–>Beef value chain of cow-calf to stocker to backgrounder to feedyard to packer, or a Hog–>Pork value chain of farrow to wean to finish to packer.
“The value of each part of the chain depends on the other parts of the chain, e.g. commodity data requires a high degree of processing and augmentation by the network to turn into a valuable asset whereas differentiated data requires less processing and augmentation to turn into a valuable asset.”
That statement also holds true for Cattle–>Beef and Hog–>Pork value chains, doesn’t it?!
Some ‘so whats’ for established companies in livestock, meat, and dairy, and their service providers:
(1) Be on the lookout for the AI-First business model, the emerging competitor in your category. Every category will have them, some already do.
(2) Sometimes what’s marketed as Artificial Intelligence is just a smart person with a laptop and mad Excel skills. Beware of the difference between AI-first and AI-as-window-dressings.
(3) A key distinction is whether a company is building an AI capability to drive its own operations and a company building an AI capability to drive its offering to customers.
Two contrasting examples:
AI-first: Opendoor. AI is central to their business model. If their AI capabilities do not meet the task, the entire business model falls apart.
AI-friendly: John Deere. AI is an addition to their business, or maybe it’s a core capability for a secondary part of their business, their software business. But if John Deere’s AI capabilities are not up to the task, less than 10% of the company’s revenue is at risk. They’re still going to manufacture and sell incredibly high-quality farm equipment.
(4) Each of the 3 steps is challenging: capturing more data, processing the data, and systematizing. Different categories within livestock, meat and dairy have bottlenecks in different places; some segments still lack sufficient data capture, while some segments simply lack the automation to consistently process that data into decision-making horsepower, while still other segments have all the building blocks without the necessary changes in management styles to harness AI.
(5) It is really hard to successfully launch and scale an AI-First company, but I think it might be harder to transition from a non-AI company to an AI-Friendly company, because it’s not just the mechanics of creating the system to get decision-making tools,
its retooling how you make decisions.
Shifting to be an AI-accelerated company requires retooling how you manage, or at least what you manage to. It requires rethinking SOP’s, in some cases. It requires rethinking culture, hiring, and training. It requires building entirely new capabilities. It requires time and trial and error, all of which require a massive philosophical leadership commitment to doing business differently.
(6) “The future is here, it’s just not evenly distributed.”
That familiar phrase applies to this whole conversation because there are companies investing now AND capturing the benefits of AI to drive better decisions, improved operations, improved customer relationships, etc.
I can think of companies across every single category within livestock, meat & dairy that are poised to win as AI evolves. They are the companies hiring data scientists and software programmers right alongside hog buyers, corn buyers, meat sales teams, or any of the other standard roles involved in the business of producing and processing.
Side note: if I were an ambitious 20-something just starting my career, I wouldn’t even entertain the idea of working for a company that doesn’t have at least 1 data scientist on staff. It’s a proxy for how forward-thinking the company is, IMO.
(7) Warren Buffett only invests in businesses with an “economic moat”, which is a business’s ability to maintain competitive advantages over its competitors in order to protect its long-term profits and market share.
Moats are valuable precisely because they are difficult to replicate, which is wildly true here.
What might seem odd is that we started with how AI-First models are not living up to the hype, then dove right into why there’s reason to be hype-ish about AI and its potential impact in the industry. That’s because this is one of those areas with a lot of potential, a lot of promise….but we have to be realistic about the challenges to realizing the potential. Not everyone is gonna win in this game.
Will Opendoor or Carvana be able to survive long enough to improve their algorithms? I have no idea.
Does the outcome of these early AI-first business models change the fact that AI-First models are likely to emerge in every industry, with increasing efficacy, in the next 5-10 years?
Lastly, I assumed that I didn’t need to offer the caveat that adopting AI does not replace the need for humans, but if your knickers are in a bunch after reading this….well, welcome to 2022, my friend. It’s hard to imagine a world in which the power of data and Artificial Intelligence decreases in the future, especially as we get better at capturing data, building systems to automate Data Learning Effects, and building processes and business models to leverage those Data Learning Effects.
There’s no going back. And the odds are good that while some are trying to decide whether or not they like the idea of Artificial Intelligence, their competitors are building their moat, brick by brick.
Artificial Insemination (AI) has massively accelerated the genetic improvement in livestock, and now it seems likely that Artificial Intelligence (AI) will provide the next acceleration, in the form of decision-making improvement.
The other AI.
What a time to be alive😉