Get the Most Out of AI Todayartelligence
A conversation with Airbnb’s Theresa Johnson highlights three tangible ways AI can help companies.
With all the clamor about the potential of AI, what are companies actually getting now for their AI investments?
Fast-improving technologies are fascinating but frustrating. When technologies progress quickly, executives can feel that they are on a treadmill with the carrot always dangling just out of reach.
They know they are making an investment — they can see the money and effort disappearing — but the benefits can seem like they are perpetually “coming soon.” This is particularly true for a technology like AI, which lends itself well to flashy demonstrations and science fiction. But working in labs and movies is quite different from working in production within organizations.
However, some companies are already seeing tangible benefits from their investments in AI. One example is Airbnb. A conversation with Theresa Johnson, product manager at Airbnb, highlighted three ways AI can help companies and is already helping the online marketplace and hospitality company.
AI is helping organizations access unstructured information. Finding answers would be much easier if all information were structured in neat tables with a limited set of defined values for each column, all nicely filled in with no missing data. Boyce and Codd would be so happy!
But no company exists in this highly normalized world. Instead, data isn’t structured because it is inherently unstructured or the costs of adding structure are prohibitive. For Airbnb listings, some structure is easy — logging a property’s number of bedrooms, number of bathrooms, square footage, amenities like a pool, etc. The difficulty is that, while these data points help, customers can’t narrow down the sheer number of listings to a manageable subset of possibilities using just this easily structured data.
Instead, Airbnb is using image processing to find similar listings, even if attributes haven’t been coded or customers can’t articulate exactly what they want. Johnson gives the example of using AI to “extract what’s in a photo plus hundreds of other signals to determine, in a human way, other listings that are fundamentally similar.” AI is helping customers get information the company has but that is in formats that were previously tough to access.
AI is helping organizations learn faster. Information takes time to diffuse through organizations. Just because one person knows something doesn’t mean others do. Another “air” company, Airbus, found that AI helped it quickly diffuse what one person learned to others in the organization.
But AI can also help the initial learning too, not just the diffusion of existing knowledge. For example, fraud prevention isn’t glamorous but is critically important for practically every organization. A huge difficulty is that the legions of fraudsters adapt rapidly ― too rapidly for any defender to keep up. Johnson describes the problem of “predicting and finding the fraud of tomorrow when it doesn’t look like the fraud that happened yesterday.” As a result, she says, fraud “is one huge area where we’re using more artificial intelligence techniques.” A particularly nice aspect of this current application is that this benefit is quantifiable, since it results in “actual dollars saved in terms of charge-backs prevented.” AI not only helps organizations diffuse knowledge but is helping create the knowledge to diffuse.
AI is helping organizations offer new services. Yes, AI can help organizations refine and optimize processes, saving time and money. In the above examples, humans could find similar listings manually or would eventually learn to spot new fraudulent techniques; it would just take a long time and be expensive. But AI is also presenting opportunities for new offerings.
For example, Airbnb is known as a platform for matching property owners and potential guests, but AI can help the company expand its offerings. Johnson says she is “super excited about AI as applied to financial services.” Airbnb recognizes that its hosts are small business entrepreneurs and that guests may need help managing funding. Johnson describes how AI can help with both sides — for hosts, “using artificial intelligence to help with lending decisions [and] with financial management,” and for guests, “understanding the resources needed to travel and to provide new AI-based products and services based on their financial habits and patterns.” AI is important for offering these types of services at scale. The technology helps organizations not only optimize processes but provide new services. (And stay tuned in September 2018: MIT SMR and BCG’s forthcoming second annual report on AI offers much greater evidence of the differences in applying AI to expenses versus revenue.)
These examples illustrate several ways AI can be used right now by a company that is actually using it. But there is more Airbnb can do, and it isn’t stopping with these use cases. Johnson notes that “there are many, many teams across Airbnb that could take advantage of these AI techniques, but each couldn’t devote the resources to individually spin up these models, nor would it be efficient as a company.” Recognizing that AI applications apply broadly, Airbnb has formed an internal AI lab to help apply AI throughout the organization. As Johnson observes, “Why do the same thing across seven different teams over and over again? Let’s create a model and scale it out to all the teams that have different use cases for it.” Our September report on AI will show that Airbnb isn’t alone in moving forward, finding more ways for AI to help.
A key word here is helping. Any technology alone can’t do these things. But AI is helping organizations like Airbnb now. These aren’t tomorrow’s AI. They are today’s AI. What can AI help your organization do today?
This interview was conducted as an input for the forthcoming MIT SMR-BCG collaborative research report on artificial intelligence, publishing in September 2018.
SOURCE: MIT Sloan Management Review