Platform Strategies in the Age of AI

In case you missed it, platforms are now the rage in technology strategy. To be current means that you need to be developing a platform to capture multisided markets of producers and consumers to exploit and capture network effects that ordinary pedestrian business models cannot hope to attain. And, unbeknownst to all but a few that closely watch this space, according to a recent report by The Center for Global Enterprise there are now some 176 Platform Companies throughout the globe who as recently as last year were classified as ordinary product and service providers.

Who knew?

For example, did you know that cars are no longer a means to get from point A to point B? Nope. Instead they are a meme that can be stuffed with all kinds of random value in the form of infotainment apps to distract the driver to the point where they can no longer manage the vehicle; hence the need for driverless cars, which are not really cars but mobile entertainment platforms. Smart homes are no longer a collection of programmable appliances but rather a community of z-wave connected nodes capable of deep learning the Zen of your den, sensing your every need and tilting your recliner to a perfect 42 degrees. Even your Fitbit can be anonymously affiliated with those of your BFF’s on Facebook just to let you know what a kind of a lay-about you’ve actually turned into.

The reclassification of data to amplify a trend is nothing new. After all eggs taste so much better when you call them oeufs, as in “I’ll have the oeufs en menrette”; how do you think we went from zero to a bazillion dollars in cloud based revenues overnight?

It sounded better than calling them hosted technology services.

To help us better understand the phenomenon of platform-based strategies, the Center for Global Enterprise has created a useful taxonomy. By their definition there are four types of platform-based strategies that share a number of common attributes. They consist of innovation, integration, investment and transaction platforms that facilitate the orchestration of matching, interaction, compliments and eco-systems in multisided markets. This facilitation is often technology based as in an integration of devices, networks and information that promote transactions or the creation of an eco-system of value creating compliments. It is further assumed that to be considered successful these platforms would create a privileged competitive position due to positive network effects that result from the interaction of suppliers and consumers that make use of the platform’s underlying capabilities.

You getting all this?

Maybe an example would be helpful.

The most frequently cited example of a successful platform is Apple’s iPhone as it has all the components that most academics agree need to be present to satisfy the platform strategy paradigm: consumers, producers and providers which, in the case of the iPhone, are mostly familiar to us. Consumers, that would be you and me, make use of the capabilities afforded by the providers, orchestration, communication and governance, in the guise of operating systems, networks, devices, by invoking third party applications provided by producers. Network effects result from the transactions and interactions that are produced through consumer generated activities on the platform creating adjacent opportunities for parties so engaged such as Facebook, Instagram, Pokémon, etc. What these eco-systems amount to in conventional strategy speak is a system of cross subsidization where all constituents across the entire spectrum enjoy some benefit in the form of cost reduction or revenue augmentation through participation. And increasingly, if you’re a technology company, if you can’t make money from some unrelated subsidy you won’t be making any money at all.

Currently, there are a number of players looking to extend this metaphor into more specialized use cases. For instance, General Electric’s IoT platform called Predix is looking to become the platform for intelligent devices and device intelligence combining data communication, collection and analysis using standard interfaces and protocols. Similarly, a number of players are busy trying to create platform plays around blockchain implementations upon which eco-systems of value creating entities could coalesce. And, if you have been paying attention, you would notice that most of the major players in AI have been making noise that their efforts are directed towards making an AI platform where they would provide the orchestration, governance and technology to support a larger eco-system of value creating entities capable of serving a host of unique consumers.

First Comes Voice

If you look beyond the acknowledged major AI players there are at least a dozen or more open or proprietary offers that are merchandized as general-purpose AI platforms. As such, they expose through API’s a specific set of well or loosely integrated AI capabilities that address specific potential needs. As mentioned in a previous post,( see The Next Cambrian Explosion – 2016) there are currently at least a dozen recognized AI disciplines/code bases including natural language processing (NLP), deep learning, semantics, inference engines, graphing systems, ontologies, etc. If you consider for a moment the 20 most popular mobile phone applications, games, location services, shopping, etc. none of them directly expose the user to any of these capabilities although all of these applications could potentially benefit from incorporating some of them.

Earlier this year Google began to promote a message called AI- Frist as a means of suggesting where the developer community might begin to focus its efforts. Upon closer inspection, Google’s message seemed to be focused on very narrow application of AI to existing offerings that being natural language processing. It would seem this notion is not lost on any of the major players as nearly all of them are placing an emphasis on seamless, robust and infallible natural language processing as the corner stone of their AI platform strategies. Amazon has followed suit with the introduction of its Alexa Skill Kit, IBM has lavished a good deal of PR on Watson’s NLP prowess, Microsoft has done the same through its Cognitive Services offerings and the list goes on, perhaps reflecting a common intuition about first mover advantages when it comes to NLP. After all, if NLP becomes the primary interface to the consumer and the primary means of acquiring knowledge of those consumers who wouldn’t want to be in-line for a piece of that opportunity? But this begins to have important implications for consumers, providers and producers, the primary platform constituents, when it comes to how AI may or may not become a platform going forward.

When it comes to general purpose AI offerings, the major players seem to be betting that platforms that offer tightly integrated AI services will win out over best of breed offers especially as capabilities improve. And there seems to be a case for this. Why would an application in need of AI services go to IBM for NLP, Microsoft for deep learning and Apple for ontology engines? But there is a larger question that application producers would have to answer before reaching this decision. Why would they expose their customer information to any of these players on an exclusive basis when owning the customer interface would pretty much ensure ownership of the customer? And if not ownership of the application’s consumers certainly ownership of the application provider as knowledge of their customers improves the level of service their end users experience. Pretty sticky stuff.

Now it might be possible to put in place safe harbor agreements where the AI services would have absolutely no access to or knowledge of the application’s consumers but exactly how realistic is that? Given Amazon’s behavior would you believe it if they told you they didn’t want your customer data if you decided to make use of their Alexa Skill Kit? What if they offered their entire AI platform for free for just a peek?

Many established consumer brands recognize the threat of losing their identity to intermediary NLP providers and have chosen instead to pursue an independent chat bot strategy often cobbling together voice based AI interfaces using open sourced technology.

Then Comes Data

Next in the hierarchy of application producer needs would be some form of deep learning focused on customer insights and knowledge. This would most likely take the form of data clustering, graphing, pattern recognition, etc. One could assume that most of the easy insights about any particular application’s customer base have already been discovered (see Sis, Boom, Bah! – December 2015). Therefore any new discoveries would likely be associated with some form of deeper learning from larger untapped data sets.

Deep learning implies that there is a potential to discover latent opportunities evident in the behavior of existing customer data or domain knowledge. But for many shops that option has already been thoroughly explored. Getting beyond what is already known about any given customer by any given application provider would require not just some but all of the customer’s information from all of the applications and all of the devices they employ. In the case of mobile devices the information gleaned from an individuals device would have to be concatenated with search, merchant services, location services, banking and credit services, ad services, social media, etc. This data from all of these various sources would then have to be volunteered to an AI provider to distil. Think of a company like Acxiom but only on a vastly larger and scarier scale. Assuming you could navigate the trust issues involved collecting all of these transactions from millions and millions of consumers, numerous application producers and platform providers, a machine would then be trained on it to discover potentially promising commercial insights.

Even if the processing power were infinite and free, a point we are rapidly approaching, most of the opportunities for most of the participants would lie well beyond the threshold of diminishing returns. But assuming discoveries are made, how would they be monetized by the AI shop that makes them and who from the thousands of platform players who contributed data would benefit from an insight so discovered?

Where to now?

What happens to the technology industry at the end of Moore’s Law? What happens when productivity improvements can no longer be wrung from circuit size and speed? Can the likes of Intel, Dell, IBM and HP envision a second act? Do dinosaurs actually become birds?

What we have witnessed over the last few years is the steady decline of pure information technology players being slowly replaced by savvy technology users employing costless, often open sourced, technology components whose business models and revenue streams are predicated on siphoning subsidies from other industries and unrelated business models (see Social Subsidization and Diminishing Returns – March 2015). How does Facebook become a leading technology company when its business model is predicated on advertising? How does Amazon become a leading technology company when its business model is predicated on retail transactions? Why is Uber considered one of the largest technology companies going by merely exploiting asset re-allocation? Where in their economic models are the investments in R&D? How many technology patents do they secure every year? The answer is it no longer matters.

Pure information technology companies are evaporating and they are being replaced with companies who no longer possess deep technical knowledge but rather the knack to produce technological arbitrage of revenue from formerly successful business models.

While it is possible that major AI platforms will emerge it is equally as likely that they won’t or at least not in the guise of the platforms that strategists recognize today. If they do emerge there is a strong probability they will be open sourced, where specific customer and domain knowledge can be retained by customer facing entities and not given over to a larger pure AI player or, worse in the case of really big data and deep learning, trapped between multiple AI partners.

Should this occur there is a case to be made that AI will become more of a forward integration strategy where AI players will acquire customer facing and domain knowledge in order to siphon off the revenues of a specific capability (retail) or a specific industry (healthcare). We can see this already writ in the published intent and acquisitions of companies like Amazon and IBM. It would not be too far fetched to assume that technology based firms with deep AI capabilities might pursue industry wide penetration strategies seeking to achieve substantial returns based on the revenue arbitration of existing businesses. The legal industry whose exorbitant fee structure and codified domain knowledge might make an excellent candidate, as would pharmaceuticals, life sciences, insurance, education and host of others.

Apple, Google, Facebook and Microsoft may still pursue a general-purpose personal agent approach hoping to “own the customer” but the likelihood is that customers will opt to own their own personal assistants (see Who’s Zooming Who – October 2015) rather than accept one that has been provided through the subsidization of commercial interests.

If they don’t join you, beat ‘em

While platforms may continue to support the cohesion of multiple parties to a complimentary value creating capability the same set of incentives may not be available to players seeking to make a go of platforms when it comes to AI. If NLP becomes the dominant interface, and it likely will, nobody is going to be downloading apps and looking at mobile ads. So, consumers won’t be the driving force that binds the platform players together; business buyers will be in the driver’s seat.

And for business buyers, who are successful incumbents in their own industry, the stakes are way too high to casually engage with technology partners who could inadvertently redraw the boundaries of their industry to their own advantage. This is particularly true for industries that continue to rely on asymmetrical information to support a lack of competitive practices and pricing including but not limited to doctors, lawyers, financial planners, insurance providers the list goes on. The advantage the current set of players in the AI space possess is the opportunity to thoroughly and completely disrupt these incumbents and radically re-write the rules of these industries.

Well by no means easy, firms like IBM might drop the aspiration of becoming an AI technology platform provider altogether and spin-out an industry stomping, revenue siphoning 800 lb. gorilla of a healthcare subsidiary. Then turn its attention to the legal industry and spin-out an 800 lb. gorilla of a legal subsidiary.

Lets face it, at this point for a shop like IBM, the prospects of becoming an industry specific dominating entity are a whole lot brighter than becoming an generic also ran AI platform company.

Would that make IBM a lot like Uber? Maybe. But they could still call themselves a technology company.

Graphic courtesy of StockPhotoServices
Copyright Carlo Leopoldo Bezerra Francini