AI isn’t magic. It's a set of tools that are used to automate simple cognitive tasks, at very large scales. Understand a few basics about AI first and you’ll be much more productive in thinking about how you can put AI to work for your business.
AI often gets hyped up for the purposes marketing and selling clickbait articles online. No, AI can’t “dream” about things. Yes, Terminator is a great flick, but we’re nowhere close to making AI with the general suite of human-level capabilities we see in the movies.
Yes there have been some great advances in AI lately (from amazing robots to game-playing AI). But these sorts of AI are still in the R&D phase, and don’t have commercial applications for small businesses yet. The kind of AI that is useful for businesses today can - when implemented correctly - ape (or mimic) simple cognitive tasks that humans can do. Think of these as simple scoped items you could - at least in principle - teach a reasonable person to perform in a few minutes.
finding desired objects in images like human faces, cars, bicycles, animals, etc.
summarizing the contents of a few paragraphs of text in a few words.
detecting irregularities in a series of financial transactions based on things like withdrawal size, timing, and unusual product purchases.
Now you might say - some of these tasks wouldn’t take a reasonable person only minutes to perform. But remember I said “teach a reasonable person to perform in a few minutes.” Telling someone to ‘mark every face’ in a big stack of images takes a second, but actually having them look through a stack of 100,000 images and do it would take hours (about 56 hours, if we assume they can markup one image every two seconds).
And therein in lies the ROI of AI: it doesn’t just automate simple cognitive tasks like these, it automates them at scale. What would take a human hours to perform can be done by a well-trained AI in seconds.
AI is all about “monkey see, monkey do.”
AI is a monkey-see, monkey do technology. For an AI to learn how to perform any of the simple cognitive tasks listed above, it needs to see examples of the desired outcome.
So, in order for an AI to learn how to
find objects in images, it needs to be shown examples of images with the desired object(s) marked by humans.
summarize a few paragraphs of text in a few words, it needs to be shown examples of paragraphs and human-made summaries.
detect irregularities in a series of financial transactions, it needs to be shown hand-labeled examples.
In other words, to use AI you need data. Quality data and - depending on the application - lots of it. An AI is then ‘trained’ on that data to ape your desired outcome.
Start putting AI to work for your business.
To put AI to work for your business begin with the end in mind. Where are the pain points of your business - both internally and with customers?
Ask yourself questions like:
What do I have? What do you know about your data? Survey the parts of your business that ingest data and ask what kind of data do you collect, and why.
Where are my internal bottlenecks? Do you have any internal bottlenecks that involve low-level and repetitive cognitive tasks that are eating up a bunch of your or your employees’ time? These could include things like cataloging digital information by hand, summarizing text, merging similar information together, etc.
What insights can I derive from customer data? Start with the end in mind - what would you like to understand better about your customers?
What kind of data-driven insights can I provide my customers? What value could your customers get out of insights from your data? What kind of feedback / predictions might make their use of your product more enjoyable?
In our next post, we will walk through a basic AI questionnaire that helps you get started thinking about how you can use AI in your business.