A recent study by Accenture found that artificial intelligence (AI) could boost profitability rates by 38 percent by 2035. That would provide a $14 trillion boost to 16 U.S. industries, including professional services, information and communication, construction, and healthcare.
As more people become aware of what the technology can do, more workplaces are looking at how it can help them create a more efficient and profitable business. But achieving these grandiose profit goals in your business won’t come easily. To be an effective tool for growing businesses, artificial intelligence must be researched and deployed properly and responsibly.
First, it’s important to understand what the term artificial intelligence (AI) means and doesn’t mean, as it has become a catch-all for many types of software and automation. In its simplest form, AI usually refers to software that can learn, adapt, and then take an action based on its own decision making. If you are curious if the software you are using incorporates AI, I’d recommend checking out this great flow chart from MIT Technology Review that can walk you through it.
There are a number of algorithms that fall under the artificial intelligence umbrella. Types of artificial intelligence include capabilities like natural language processing and machine learning, which can then be split further into areas like deep learning and reinforcement learning. These are all terms you will hear thrown around that describe how the algorithms train, improve, and output their results. But let’s put the terms aside for now, and focus on the matter at hand: how to incorporate AI into your business.
Making AI work for you
AI has a lot to offer. Considering how it could help improve your business, learning more about it would be a smart move. As the technology continues to improve, you need to monitor how it can affect your industry to stay ahead of your competitors. If you are looking to bring AI into your business, there are a few things to keep in mind as you go about developing a strategy.
Don’t apply AI for the sake of applying AI
Superficially applying artificial intelligence can be a waste of money and a drain of resources. Don’t bring AI into your business just because you want your company to appear on the cutting edge. AI can be used to help sell why your product is different, but don’t let that be the sole driving factor for adopting it. “People are either all about AI, and really try to show [the AI] to the extent that they are over the top because it’s important as a differentiator, or it’s entirely hidden from the end-user,” says Rob May, CEO of AI company Talla.
Study the basics of AI first
Before approaching a company to help develop artificial intelligence solutions for your businesses, have a basic understanding of what AI can do and where it is helpful. Some great resources are:
- edX AI courses: High-quality online courses created in partnership with prominent universities.
- The Algorithm email newsletter: A great weekly newsletter covering the latest in AI news and delving into the latest AI research.
- Cognition X’s newsletter: A fantastic daily summary of AI business news.
- The AI at Work podcast: A great addition to your morning commute, this podcast brings on industry leaders to talk about their views of the AI industry, where it is headed and how you can take advantage of the technologies.
Solve a problem with AI
Make sure you have a concrete problem or area to which you believe artificial intelligence will improve your business. “A lot of times AI gets adopted in the area where there’s the most forward-thinking executive that wants to try AI,” says May. “That may or may not be where AI can have the biggest impact on the business. There’s not really anybody going through their income statement and balance sheet and looking at the opportunities for AI to have the biggest economic impact.”
Be conscious of algorithmic bias and biased datasets
Just because an algorithm is making the decisions does not mean it’s not biased. Detecting and removing algorithmic bias can be difficult. The decisions and results AI algorithms make aren’t necessarily without favoritism towards one group or another. If it is being trained on historically biased data or on a dataset that isn’t diverse, it could have biases ingrained into the software.
Does your AI need to explain how it came to a decision?
Making an explainable AI algorithm is difficult. If your application requires the software to be able to tell the user exactly why it made a choice, make sure to know this upfront. “It’s a big, big concern for people because companies are worried about a situation where they don’t fully understand why things were done, or why decisions were made,” says May. Be honest with any companies you are partnering with in regards to developing or applying the AI so they can tell you if the explainability you need in achievable.
Now that you know what to be aware of when developing your AI business strategy, here are some areas where you can look at applying artificial intelligence to your growing business.
Predicting what the world wants
Since algorithms learn on past data and experiences, one application of AI systems is using past purchases or market data to predict what will happen next. This is called predictive analytics. This could assist you in forecasting what type of car a customer might want to buy, or even when they would be shopping for a new car.
For example, since 2004, Walmart has been using this type of data to learn which products it should stock up on in cities in the path of hurricanes. Famously, Strawberry Pop-Tarts were one of the biggest sellers in these regions, resulting in them sending extra supplies to necessary stores. AI can take things to the next level, helping you tailor your marketing, purchasing, and even your products to what the market is likely to demand. This provides you fuel for making big business decisions.
For example, despite Wal-Mart’s early use of predictive analytics, they believe they have a long way to go. “From 1 to 10 in our use of data, I would say we’re probably about a 2,” said Douglas McMillon, CEO and president of Wal-Mart, in 2017. “We use data to improve in-stock and replenishment. We don’t use data to personalize.”
While predictive analytics have existed long before AI, as IBM puts it, “Predictive analytics has busted out of its data science shell.” The techniques are becoming accessible to a wider audience that isn’t trained in the intricacies of data science. Improvements in machine learning specifically have been one of the keys to this. Machine learning finds patterns in data and then uses those patterns in some way. For example, websites that focus on creating digital content like Pinterest learn from what you pin, like, and interact with on the platform. Based on this information, the platforms learn what will be most likely to keep you on a website, which content is similar to the things you have been looking at and which advertisements would be most tempting for you to click.
Streamlining recruiting and hiring
You can use AI for all parts of the recruiting and hiring process, from sorting resumes to actually conducting interviews. Websites like Arya claim they can search for, recommend, and engage potential candidates for you using artificial intelligence, saving you the trouble of searching for qualified candidates.
Online recruiting service Ziprecruiter offers software that learns from your preferences in candidates and recommends new candidates based on that feedback. HireVue says it will even perform interviews and tests of candidates. No matter the part of the hiring process, there is probably an AI-powered tool that can assist with it in some way.
Be wary of some of the more extravagant claims in this area though, as the effectiveness of some of these types of services is still being debated. The service Human says it can pick up on someone’s emotions in an interview to help you determine the best candidate for the job. Human emotion is obviously a highly complex thing for AI to understand, and relying on software to make emotional judgments can be difficult. “I think what’s really hard about hiring in AI, is that your data set that’s input is a resume or LinkedIn profile. It’s not anything about how the person really works,” says May. “It doesn’t capture temperament, personality, work energy, and all these other different things.”
In situations like this, being conscious of AI bias is crucial. Even though it is an algorithm, it can still unintentionally project bias and make unintended conclusions based on past data, perpetuating things like long-standing diversity issues.
Smarter robots on the factory floor
The robots on car factory floors, which you have likely seen pictures of, have traditionally been robots with pre-planned paths. They can’t change paths when presented with a new situation. Now, by combining AI and robotics, they more adaptable.
This type of AI is called machine vision, which allows devices to be guided in their decision making by capturing and processing data they take in about their surroundings. This could mean stopping if a human gets too close to prevent a collision, or even deciding how to pick up differently shaped objects. These abilities, combined with some decreases in the cost of robots, also mean they can be applied to more business scenarios and needs. The result is an increase in robot adoption globally. As of 2016, there were 74 robots per 10,000 workers worldwide according to the International Federation of Robotics.
Robots are obviously most applicable to the manufacturing industry, which is one of the three industries Accenture predicts will see some of the greatest financial impact from the technology. The other two industries are information and communication and financial services.
AI customer support, at your service
You have probably noticed an increase in chatbots that offer to help as you shop online. Ecommerce businesses hope these AI chatbots preemptively work you through any issues or questions you may have, rather than take up the time of a human worker.
May says that people who use his company’s product, Talla, make their customer service representatives more productive in two main ways. One is to offer a natural language-based self-service widget on websites that customers can use to find answers to their support needs. And the second is to help human agents with things like answer retrieval for certain incoming questions, or with some automation tasks related to ticketing and case management.
By deploying these types of tools, you can reduce the workload for human customer service representatives, and resolve problems before they become customer complaints.
AI at work
More business applications for AI are popping up every day. As you consider AI solutions to your business, make sure to start with a business problem AI can help solve. Artificial intelligence will continue to move further into our daily home and work lives, so now is the time to learn, understand, and embrace it.