This month Capterra launched our Artificial Intelligence Software category.
I’m pretty excited.
This is the perfect time. Demand for artificial intelligence is growing. The market for AI will reach $70 billion by 2020, according to BofA Merrill Lynch. And by 2035, AI could cause annual economic growth rates to double and boost productivity by nearly 40%, an Accenture artificial intelligence report predicts.
What Is Artificial Intelligence?
Alright, first things first. Let’s review what artificial intelligence is. Like any emerging technology, intelligent people (mostly nerds) disagree. “Distilling a generally-accepted definition of what qualifies as artificial intelligence (AI) has become a revived topic of debate in recent times,” wrote Nathan Benaich, a tech investor at Playfair Capital. “AI is not one technology. It is in fact a broad field constituted of many disciplines, ranging from robotics to machine learning.”
But generally speaking, AI is an umbrella term for technology that does non-rote computing, aka, the stuff only humans can do right now. Today, for the most part, computers do what they’re programmed to do, no more and no less. Artificially intelligent computers do more than their humans program them to. Within AI there are several, somewhat overlapping, subcategories. They include machine learning, deep learning, and neural networks.
Right now the main activity in artificial intelligence centers around machine learning.
Machine learning is when a computer algorithm improves itself.
An algorithm is just a complex formula. Inputs and outputs. A machine learning algorithm takes inputs, delivers an output, and then adjusts itself based on how accurate the output is. If you’ve ever responded to a tagging suggestion on a Facebook photo, you’ve interacted with a machine learning algorithm.
If you accepted the suggestion, you patted the algorithm on the back. But if you chose the second or third option on the list, the algorithm tweaked itself based on your feedback. You made the algorithm just a little more accurate.
What is Artificial Intelligence software?
There are two main types of artificial intelligence software: “pure” and “applied.”
Pure AI software is essentially the algorithm itself. It’s up to you to decide what data sets to apply it to and how to train it.
Applied AI software uses artificial intelligence to automate existing work and to do new work, which is now profitable through cheaper and more powerful AI.
Natural Language Processing (NLP) enables AI to respond appropriately to human speech. Siri, Cortana, and Alexa are chatbots powered by NPL AI. The machine learning comes in when these assistants tweak their own algorithms in response to your behavior. If you say, “Thanks” or nothing at all, nothing changes. But if you repeat yourself, or try the command with different words, they slightly adjust their settings to respond better the next time.
In the B2B world, applied NPL generally takes the form of a chatbot or automated (not menu-driven) phone attendant powered by a machine learning algorithm. These machines route your users to the right data or people on your team. The promise is automated call/email routing that gets better and better over time (mostly) on its own.
At CES 2017, “voice-controlled AI assistants [were] everywhere,” according to MIT Technology Review writer Jamie Condliffe.
But there are other business applications for chatbots. Facebook launched the Messenger Platform in April 2016, where brands can build NPL-powered bots for Messenger. By November, CIO Journal was reporting that Messenger powered at least 33,000 chatbots, including Mastercard’s bot, “Kai.”
Even CRM software is getting artificially intelligent. Salesforce Einstein uses email, calendar, and social data to predict when your prospect is statistically most likely to open an email from you and respond positively, and recommends a 20-minute time slot for sending your prospect that follow-up email.
What’s next for AI software?
According to Condliffe, 2017 will see companies adding NLP AI “into as many pieces of hardware as possible.”
Gartner predicts that by 2019, startups will dominate Amazon, Google, IBM, and Microsoft in the market for artificial intelligence software by offering industry-tailored chatbots that excel at niche tasks. For IT customer service, more than 10% of hires will spend most of their time writing scripts for robots. Almost a third of leading companies will see AI eat into their revenues.
By 2020, organizations will be able to improve their chances of success four times over by incorporating human factors engineering best practices into their artificial intelligence projects. Also by then, a fifth of companies will have employees whose full-time job is monitoring and guiding neural networks.
An AI checklist for businesses
Here are some things to do to prepare for ubiquitous AI software:
Look around your business for opportunities to use AI to cut costs or ways to provide more value to your customers. Ask yourself whether a bot could cost-effectively reduce your average time to resolution or help upsell your customers.
Amazon’s “Just Ask” feature on the Echo is an example of upselling based on customer data. Alexa offers users daily promotions and special deals based on their buying history, delivery address, and shipping and payment preferences. Customer Service expert Richard Shapiro calls it “a game changer.” Could AI replace human insurance assessments? Could it help marketing generate more leads or sales close more deals?
After you come up with some potential use cases for AI, send RFPs to smaller AI vendors for specific projects. Of course, you can find some in our Artificial Intelligence Software category.
When hiring, the skills you’ll need center around creating, updating, testing, training, and retraining neural networks. Look for candidates who want to learn bot scripting and development and who show an aptitude for internal communications and process articulation. Train your workers in algorithm testing, content acquisition, and data employment in artificial intelligence projects. Also, look for human factors engineering aptitude.
So, do you need artificial intelligence software? It depends, of course, on what parts of your business can be automated, how much it will cost to automate them, and how much additional profit automating them will yield.
Here’s what’s certain. Bots are learning to handle more parts of your business autonomously every day. And every year artificial intelligence software gets cheaper and more widely available. Now is the time to start looking around for things to automate, getting proposals from vendors, and training up the talent on your team.
To get started, check out our new Artificial Intelligence Software category. And let me know in the comments what parts of your business you’d like to automate first.