If artificial intelligence conjures images of super-intelligent robots in your mind, you’ve been watching too much sci-fi. While we haven’t yet created machines that behave, think or feel like humans, we have created learning algorithms that are extraordinarily effective at pattern recognition and that can solve complex problems like beating the best humans at the game of Go and autonomous driving. These algorithms and neural networks are already impacting the way we think and feel (Facebook), the content we see and consume (Netflix, Google), and what we buy (Amazon), and soon they’ll be integrated into everything from healthcare to education to transportation: improving processes and learning as they do it.
Over the next few years, most, if not all leading companies will use artificial intelligence and machine learning to improve their efficiency and productivity. Just as cloud computing and mobile technologies have been integrated across the board, AI will be the next differentiator between the businesses that surge ahead, and those that are left behind. AI will redefine what we see as “normal.” This is the premise behind Element AI, the company we co-founded in 2016 and that is already proving this hypothesis having raised a $102M series A this year to build a platform to help leading companies and governments make the AI-first transition.
In shifting to an AI-first mindset, companies will solve problems in novel ways, transforming business processes, and the way products and services are engineered and delivered in the process. Rather than applying AI to improve a system, these companies will redesign such systems from the ground up with AI at their core.
They will ask themselves: if a self-learning, fully-automated AI system can be built to solve this problem or provide this service, then what would the ultimate user experience be? And if an AI system can deliver this ultimate user experience, then what data sets would be needed to train it? Does the data currently exist? If not, what is the product needed to build these missing data sets, and how would it evolve over time to reach full automation?
The key is in building a product that provides enough initial value to users that their interaction with the product helps build the data set required to train the AI. This, in turn, improves the product experience, initiating a virtuous cycle or positive feedback loop. The more the product gets used, the richer the data set, the better the AI, the better the user experience. Over time, the users of the product will become significantly more productive and the product more automated.
The companies that adopt this AI-first product approach will introduce new models that previously weren’t possible: both because they will have superhuman abilities, and also because they will dramatically reduce the cost of providing certain services.
A concrete example of a company embracing AI-first principles is Mindbridge, a startup that helps auditors to more easily detect potentially fraudulent or at-risk transactions. Their system looks for patterns and flags aberrations, classifying transactions in terms of their risk level, which is then verified by human auditors. The AI learns as it goes: as more transactions are verified by the auditors, the predictions become more accurate. Ultimately, the AI will become so accurate that it will detect potential fraud before it happens— something humans simply cannot do. While today, Mindbridge sells software to auditing firms, in the future — once their AI has reached maturity — they could sell an insurance product to companies guaranteeing that they’ll never have to deal with corporate fraud again.
Because taking an AI-first approach means greater accuracy and efficiency over time, it also holds transformative potential in areas of society where we have massive inequality. In fields like law, education and healthcare, which are currently costly to do well, AI can be used to democratize and personalize services. For instance, with dramatically increased efficiency through pattern recognition and natural language processing, AI can significantly reduce the time lawyers need for case research. While at present, junior lawyers spend hours poring through past cases, a company called ROSS Intelligence is building an AI that quickly finds relevant cases using natural language searches. Their AI learns from every interaction the case researchers have with the system, so as time passes and the AI becomes more accurate, the system will extract the relevant content in real time based on nothing more than a description of the current case. In this way, the people doing case research become significantly more productive and legal fees are reduced, making justice accessible to everyone.
An example of how AI can transform the healthcare space is likewise seen in Imagia, a company that uses AI to better diagnose cancer by looking at numerous sources of data: images, scans, ultrasounds, personal history, family background, etc. The technology is not meant to replace doctors or technicians, but gives them a second opinion by presenting recommendations based on a large volume of data that has been analysed by other doctors in the past. This also helps speed up the treatment process: by making more accurate diagnoses, fewer tests need to be run, patients can be treated faster, the system is unburdened, and the costs of helping people with cancer plunges.
Other potential uses for AI in healthcare are limitless, because diagnosis is really about pattern recognition and determining the right protocol is about simulating potential outcomes. Because accuracy is based on data, human doctors only have access to what they have time to read and the patients they see. The AI, however, can draw from vast banks of information that would take a single human hundreds of years to process. Also, for people living in inaccessible or rural areas, access to inexpensive care can be made available by using video conferencing coupled with AI diagnostics. And while human interaction will always be important, we can use the AI to make the nurses, doctors and other caregivers exponentially more productive, knowledgeable and focused.
These are just a few examples of ways an AI-first approach can transform our society, and it really takes a change in mindset to think about how AI will affect every market in the future. When an entrepreneur comes to us with an idea, we like to have a conversation about how AI could impact their chosen market and business model. Given the potential AI has to transform industries, create new experiences and impact society for the better, all entrepreneurs — whether they are AI researchers, software engineers, product managers or business people — should be tapping into the potential of this new technology.