How to use AI in your business in 2019 — AI sweet spots
This posting may contain affiliate links for products and services. I only recommend products and services that I use or would use. If you click through and make a purchase, I may receive a commission (at no additional cost to you). Thank you for your support in this way.
Step one in figuring out how to use AI in your business is to know what AI can and can’t do today. David Petersson lays out 2019’s AI sweet spots for enterprise decision-makers.
As we begin 2019, AI technologies are attaching themselves to more and more aspects of our lives. If you doubt this, just ask Alexa. The symbiosis makes sense. The tons of data we and our man-made machines generate today have proved to be a wonderful training tool, turning a class of technology we don’t understand all that well into crackerjack pattern-detecting systems. Today, AI can:
- Recognize faces
- Recognize objects
- Recognize malicious behaviors
- Recognize language patterns and be used in translation
- Manipulate images
In the future, AI technologies will do many more deeds — good and bad — previously done by humans.
Still, for CIOs and business leaders immersed in how to use AI in your business, it is crucial to understand what AI can do and what it can’t. A lack of understanding will result in unrealistic expectations, wasted time and budget on ill-conceived goals, and missed opportunities to automate tasks that otherwise would require an enormous human workforce.
This column explores how to use AI in your business by reviewing areas where AI can help companies operate more efficiently and effectively; the analysis is supplemented with some examples of AI software targeted to solve or improve specific enterprise pain points.
AI in your business: Statistics
Machine learning grew out from big data and quickly carved out a role in analytics. We used classifying algorithms to, for instance, separate spam messages from legitimate ones. We did this by converting the words to numbers the computer could understand, and then measured which types of words were most relevant to spams and which ones appeared mostly in innocent messages. These classifying algorithms evolved into being able to classify people based on their choices and what they liked, a skill put to use in the infamous Cambridge Analytica case.
Soon, this data could be used also for “predictions,” i.e., once you have classified people into certain groups, you could also take action based on those classifications. For instance, CrystalKnows uses AI to classify people into 64 different personalities and then uses that data to assist marketers and sales people to better craft their messages to their leads.
But now, we have AI that helps us to understand statistics, turning numbers into narratives that non-data scientists can grasp, or that can highlight important information that should be given. For example:
- Automated Insights Inc., founded by former Cisco engineer Robbie Allen, uses a process called natural language generation to turn big data and charts into human-sounding narratives, making them much easier and faster to comprehend. Gartner predicts that “by 2020, natural language generation and artificial intelligence will become a standard feature of 90% of modern BI and analytics platforms.”
- Salesforce uses its Einstein AI technology, to surface actionable data to sales reps, allowing them to focus on the leads that have a higher chance of converting.
- The Boston startup Laudio uses AI to analyze multiple data sources to improve staff retention; it can detect when a member should be praised for a job well done or when they need support.
AI in your business: Conversation
Remember the chatbot bubble? Chatbots were supposed to replace apps and human agents to a large degree, but many of them failed as they were not smart enough. The problem in any AI learning is that the system does not really know what it is dealing with. An AI bot can identify cats, yet it does not know what a cat is, or what an animal is in general. AI’s comprehension is very limited and “narrow,” while for full understanding, we would need “general AI.”
Anthony Macciola CIO, Abbyy
Initially, to let AI engage in a meaningful way in any conversation, the system would watch for specific keywords. Of course, this system is very limited due to the many ways we humans tend to express ourselves, and it is very difficult to predict all variations. In those instances, the system either has to throw an exception or refer to a human operator.
Still, this does not prevent AI from becoming an expert in its own specialized fields.
Unbabel, a Lisbon-headquartered startup, developed AI technology to remove the language barrier. Using Unbabel, companies have managed to serve clients all over the world in 28 different languages — in one instance, 15 staff members could take on tasks that would normally require hundreds of staff members.
So, even if AI still can’t completely replace human workers, its sweet spot lies in augmenting them. According to Kaitlyn Lloyd from Automated Insights, “The best way to implement a valuable AI strategy is in close partnership with humans, with the AI serving as augmentation to human intelligence.”
Lloyd’s argument for how to use AI in your business is that humans have the ability to relate on a personal level; we have strong emotional intelligence and communication skills. Artificial intelligence, without human teaching, can’t learn these soft skills — at least for now.
Use AI in your business for mundane tasks
It’s 2019 and we still don’t have autonomous AI-driven vehicles. The ones that are operating continue to be subject to errors and fatal incidents. It has become evident that relying solely on AI is not a short-term answer for self-driving cars, and companies have moved to using LiDAR or inter-vehicle communication systems to cope with the shortcomings. Billions of dollars have been devoted to this industry, which faces many bumps in the road ahead.
Thus, when you set off for investing in AI for your business, it is imperative to have a correct AI strategy. As Anthony Macciola, CIO at the text scanning and optical character recognition software company Abbyy, put it: “The right way to implement AI is to first drill down to the core use case you are wanting to optimize.”
The wrong way to approach AI? “It is to think it will magically fix your organization. In fact, AI-enabling technologies shine a spotlight on inefficient processes and systems,” Maccioloa said.
In other words, in plotting how to use AI in your business, it is best to start from the mundane tasks and inefficiencies you face and build up from there.
AI takes A/B testing to new level
As mentioned earlier, AI has the capability to automate many of your tasks. Ascend, software from Sentient Technologies, is a good example of how to use AI designed to address a specific pain point. Remember A/B testing? The logic there is to test two versions of your website (or email) and see which one works better, so eventually you can keep the one that’s performing better.
Ascend uses AI to take A/B testing to a new level, letting the system try many different combinations automatically over time and gradually find the best combination. It uses a technique called “adaptive evolutionary optimization,” which is based on Bayesian and traditional statistical techniques to predict the performance of the winning versions.
As seen from the previous example, not only addressing the correct pain point is vital for a successful AI strategy, but so is the type of AI approach taken by your provider.
Consider the approach taken by Endor, which sells predicative analytics software co-developed at the MIT Media Lab by co-founders Alex Pentland, the Toshiba Professor of Media Arts and Sciences, and Yaniv Altshuler, a former MIT postdoc. The company’s aim is to predict customer behavior, but in order to build its AI, Endor uses the Social Physics theory: the science for understanding human crowd behavior based on big data. This way, the company not only knows what data to train the machine learning algorithm on, but also what parts of that data is most important and how it should be used.
And finally, as you build a strategy for how to use AI in your business, keep expectations realistic. Abbyy’s Macciola pointed to the mistaken notion some have that AI will completely replace knowledge workers. Echoing Lloyd’s argument that people have emotional intelligence that today is lacking in machines, he also underscored the fact that people have more advanced training and understand the company’s core values in conjunction to the technology processes better than machines at this point. “The skilled worker will always be vital to the success of the company,” he said.
AI is awesome. It’s doing things that previously would need plenty of man-hours and extra costs, and because it can automate so many things at a fraction of the costs, it is opening new opportunities for innovation and creativity.
But AI has its limits. We are still far from artificial superintelligence and, in many cases, AI still needs human assistance. It’s safe to say that AI’s sweet spot is in taking on the repetitive, mundane parts of your business and incenting humans to focus on what only they can do (for now): Be creative.