From Infinity to 8: Translating AI into real numbers

Like infinity, artificial intelligence is an abstract concept. AI commercials show floating orbs and a sprinkling of fairy dust providing magical answers to our questions—even those we didn’t know to ask. These presentations of AI remind me of an episode from South Park’s second season called “Underpants Gnomes.” In this episode, gnomes collect underpants and make a profit. The question is, how exactly do they get from point A to point B? The business plan is revealed via a slide, of course:

1. Collect underpants
2. ?
3. Profit

AI offers something similar: (1) Collect data, (2) AI, (3) Profit! My goal in this article is to help you be more explicit about Step 2. I hope it helps you make real the incredible AI opportunities that I know are available to your organization.

The first step in getting real with AI is to define it: AI is just maths. Just like there isn’t only one kind of math, there isn’t one AI. Before you run away because of the “m” word, have faith: just as you don’t need to know how to code to influence software design successfully, you don’t need to know math to influence the AI in your desired solution. Focus on the inputs and the outputs, and how you can validate each. In this post, I’ll cover the input and output validation tips, as well as how to make sure what’s in between focuses on the problem you need solved.

Putting the wheel before the cart before the horse

Every day, companies approach my company, Nara Logics, saying AI is one of their top objectives this year, and asking if our platform can help. Scott Cook, founder of Intuit, says, “Success is not delivering a feature; success is learning how to solve a customer’s problem.” In this analogy, AI success is two clicks away. Essentially, AI helps you deliver a feature. To find a problem that is a good fit for AI, I recommend focusing on the “Four Vs” of big data—i.e., the input:

• What data do you have large Volumes of? What actions does it impact now? And what ones could it impact? For consumer companies, for example, customer transactions are large volumes of data. For a manufacturing company, factory production information provides volume.
• Where is there significant Variety in your data? For example, an insurance company offering a few hundred policies doesn’t have volume. However, the variety of customer attributes driving the match to a policy is big data.
• Do you have areas of high Velocity data? IoT is certainly driving this aspect of data for many companies. I also encourage you to think about relative velocity. If an analysis takes you a week and you need to make decisions daily, then you have a high velocity relative to your current capacity. This issue most often affects companies tackling their digital transformations.
• Are there Veracity questions with your data? Do you need to track and verify data initially, or over time? One of our government intelligence customers tracks and verifies sources of data via both automated machine learning and scoring through human feedback.

In addition to evaluating the opportunities in your Vs, take a long, hard look in the mirror to see if your organization is ready to learn. A primary advantage of AI over alternative data analysis methods is continual, progressive, automated learning from data. For example, is your organization ready to respond proactively to possible quality issues, rather than only after issues are discovered? Are you ready to change your supply chain when you learn there’s a summer market for your winter product?

These questions may sound simple, but they can involve re-branding and/or significant organizational and operational changes. If your business is not agile, changes could cost more than the savings or additional sales. To really leverage AI, you need to build a learning organization. For some ideas of what change this could mean, I’m a fan of this Harvard Business Review article on Unilever’s move from insights to action.

Nobody here but us chickens

In an earlier article, I talked through the holy trinity of AI: the chicken (algorithms), eggs (data), and bacon (results). So far, I’ve discussed data and results—or at least the uncooked bacon, i.e., the problem you are solving.

Now, let’s talk about the chickens. How do you figure out the right algorithm? Here are my three pieces of advice for minding your chickens as they hatch:

• AI’s equivalent to the real estate mantra of location, location, location should be lean, lean, lean. Use the lean feedback cycle: “build, measure, learn,” with a doubled-down emphasis on data—which is a by-product of measure, and the requirement for learn in the normal cycle. What’s important here is that we iterate more with continuous experiments and learning in our cycle as we start understanding how our chosen algorithm(s) interact with our data to deliver our results.
• Data now drives actions in our applications; we must become much more data conversant. We need to add another key role in our product development teams: the data whisperer. Notice I did not say data analyst or data scientist. You need someone on your team who is not only talented in analyzing your data, but also understands its etymology and your organization. Cesar, the dog whisperer, teaches dog owners tools tailored to their situations and abilities to tame their pet problems.
• Get real, fast. Because we are excited about the possibilities in AI, we often spend our time on the AI itself and not the rest of the application. Start with a quick statistical approach to serve your application, fake the image recognition with a team of people, or build a simple decision tree to simulate the AI. Make sure the basic flow of your application works for users. Learn from what people want and expect. They don’t have to like your “intelligence,” but they do need to respond to the problem you are solving.

Overall, this practical approach is similar to Marc Andreesen’s mantra on product/market fit. In this case, I call it tech/feature fit. Focus obsessively on the right AI fit for your feature.

This overall framework is not new. Use proven product development tools and frameworks, just add the emphasis on data and respect the new twists that requires. Think of this as a food chain: software is eating the world; software is fed by AI; and AI is fed by data. Know what you eat, and keep it real.