Machine learning is an exceptionally powerful technology, and its applications are virtually limitless. Here are five of the more surprising — and impactful — ones we’ve seen lately.
Over the last few years, AI and machine learning (ML) have quickly transformed from specialist technologies with high barriers to entry into ubiquitous business tools. Today, consumers and organizations interact with both every day — whether it’s directly through communications with chatbots or indirectly as they act on the outputs of ML models.
Use cases like process automation, customer service automation, and CX personalization aren’t just widely known and accepted — they’re table stakes for modern businesses. But most organizations are still just scratching the surface of what’s possible with ML.
Leading innovators are constantly pushing the boundaries of what can be achieved with AI and ML, creating powerful solutions to business challenges that have persisted for decades. Here are five surprising business problems that cutting-edge ML use cases are helping to solve today.
ML point solutions help thousands of organizations make decisions about how products are priced, where customers are targeted, what ideal customer journeys look like and even what entire marketing strategies should look like.
Currently, all that those individual solutions do is help businesses optimize against a single specific business goal. But slowly, organizations are broadening their use of ML, and looking at how it can be harnessed across different domains and solutions to accelerate learning and decision-making at scale.
Advanced decision factories use ML to offer customers a diverse range of options as they interact with an organization and its services. Models continuously learn which options — say price sensitivity, convenience, value-added services, etc. — are working, and scale those up and out to drive better results in more customer interaction.
It’s a real-time approach to learning that eliminates the time it takes to turn customer insight into action. Everything a company learns from every interaction is immediately applied in future decisions to improve experiences and drive higher customer engagement.
Creative tasks like content creation and product design are time-consuming and require specialist skills to do well. But as ML-powered content creation tools mature, the way businesses approach them is set to transform significantly over the next few years.
Today, content creation aids have broken through to the mainstream — finding their way into the biggest creative suites, including Photoshop. While completely AI-generated content may not quite be enterprise-ready yet, ML models have become incredibly strong at summarizing existing content and creating variations on existing designs.
That’s having a huge impact on how leading businesses manage creative workflows. If we take written content as an example, today, if a human creates a report, AI can easily summarize that content into atomized content for use in other channels. That helps save a huge amount of time and money, and removes a lot of the opportunity cost incurred in content strategies.
With ML models automatically repurposing, summarizing, and reimagining content, teams don’t need to choose how they want to present content or prioritize creation for specific channels. They can simply roll out tailored versions across all channels as necessary — helping them learn a lot more about what works and what customers find engaging.
Enterprises use A/B testing to assess everything from the effectiveness of email subject lines to how well products are positioned on store shelves. Managed well, it’s a powerful tool, but long feedback loops have always been a barrier to its effectiveness.
In the past, businesses had to devise their options, test them in live environments, gather their data, then process that data before they could apply any lessons. Depending on what businesses were testing, that process could take weeks. And in a lot of scenarios, the insights could be completely outdated by the time businesses acted on them.
Today, a lightweight form of reinforcement learning known as contextual bandits are making real-time testing and optimization between multiple options a reality. These algorithms take a list of options, serve them to customers, then learn from engagement data in real time and automatically apply what they learn to dynamically deliver the options customers respond to most.
Crucially, they don’t just give businesses a snapshot view of preferences at the time of testing. Advanced algorithms continuously test and try options, so businesses can be confident that they’re serving customers what they want to see as their preferences change.
Every business would like to do more with less — whether that’s achieving higher growth at a lower cost, or making limited innovation budgets go further. But traditionally, optimizing the utilization of highly constrained resources has been very challenging.
Take Kittilä Airport, for example. It’s a small airport in the north of Finland, with just 12 spots for airplane parking. But during peak months, it sees 58 flights arrive every day — 70-80% of which arrive in a single four-hour window. To accommodate all those flights, parking plans need to be carefully calculated, but the number of variables involved makes that an extremely tough challenge.
Today, the airport is using an ML optimization model to automatically build the best possible parking plans. It uses daily data and considers every possible variable to generate optimal plans that make the most of the available spots.
The same principle can be applied to virtually any constrained resource or asset. By looking at all the variables surrounding that resource’s use, and considering the needs of every party vying for it, ML models can give you an objective view of how to allocate that resource to deliver maximum value for the organization.
Millions of businesses use ML to make routine decisions about their operations every day. But when it’s harnessed at scale, ML can also provide answers to some of the biggest strategic questions about an organization’s future.
By harnessing data from across an organization, businesses can use ML models to simulate the impacts of huge shifts, so they can make informed decisions about major strategic initiatives. For example, if a company aimed to reach a zero CO2 goal, it could use models to accurately predict the outcomes of different decisions, such as moving to an all-electric vehicle fleet, or overhauling its supplier portfolio.
In scenarios like that, ML is helping to bring informed objectivity to decisions that are often driven by gut feelings and best intentions, rather than hard data. It’s a huge shift in how strategic decisions are made, and has the potential to transform strategy changes from leaps into the unknown, into calculated, well-forecasted organizational evolutions.
As businesses become more comfortable with common, accessible ML capabilities, it’s important not to lose sight of the fact that ML is an extremely powerful technology, and its applications are virtually limitless. At the same time, it’s not flawless technology — we’ve seen many examples where the use of ML can raise ethical questions or where there are issues with bias in data sets. But if you’re careful and considered in your deployments, you surprise yourself with what business problems you can crack.