Over the years a number of companies have approached me with a seemingly simple question: "How should we use AI (or ML, or data science)?"
Do you see the problem there? That question treats "using AI" as a foregone conclusion. That way lies tears. Tears and lost money.
I've found it more helpful to first ask whether AI will address the problem at hand.
And that's why I've assembled this guide.
If you're curious about whether AI can help you, or if your CEO has declared that the company will "do AI" and then left it up to you to make that a reality, these six questions should help you on your way.
These are same questions I ask as I suss out whether a company has an AI-shaped problem.
(Just a note: while I think what I've written here is solid, you'll want to refer to the disclaimer at the end of this page.)
I'm Q McCallum. I help companies navigate AI matters such as risk management and due diligence on the one side, and AI research and strategy on the other.
My main website has more details.
Tap or click each question to see the explanation:
1. Are you trying to impress someone?
AI is hot (again) and investors are throwing serious money at AI startups. So if your goal is to convince some VCs or potential customers that you're all about the hip technology … Be my guest. Spray-paint "AI-powered" onto your product. Cash is king and I respect your hustle.
You can also skip the rest of this guide, then.
If you're actually trying to build something for real -- something that needs to function properly and yield business value -- read on:
2. Are you trying to automate something?
AI is all about automation. Specifically, it is automation based on finding patterns in data. That makes it good for:
- predicting things ("what should this house price be?")
- classifying things ("is this document about dogs or about cats?")
- grouping things ("here's a pile of toys; please figure out which ones are similar and put them into buckets")
- finding anomalies ("show me which credit card purchase looks weird")
- generating text or images (like Midjourney or ChatGPT)
If your problem fits one of those shapes, you're in luck! AI may just help you. Continue reading.
If this kind of automation won't solve your problem, though, AI is not for you. You can skip the rest of this guide and save your money.
3. Do you need to automate existing business rules?
Are you trying to automate firm, unambiguous business rules? Skip the AI and build some custom software. Those if/then statements and for() loops power just about every business out there, with good reason.
On the other hand, do you have records of past activity? And do you want to mine those records for patterns in order to uncover what those business rules might be? You're probably looking at AI.
Emphasis on "probably":
4. Does it require a lot of nuance or human expertise?
While AI beats software as far as automating fuzzy concepts, it's not perfect. Some situations simply call for the expertise and nuance of an experienced person.
In those cases, AI is a liability. It will give you answers, sure … but those answers will be wrong. Probably a lot more often than you'd like.
So if your problem requires too much of a human touch, AI won't do the entire job. At best you'll be able to build something to assist your human team members. Which is great. But if your plan was to develop AI in order to replace people on your team … no.
5. Will you face legal or public backlash for this?
AI has developed a weird reputation. Sometimes people love it. Sometimes they think it's creepy. And once in a while, they love it until they find out it's being used in creepy ways.
(Remember when Target got called out for predicting customers' pregnancies? Or when insurance company Lemonade claimed it was analyzing customers' "non-verbal cues" to detect fraud? That sort of thing.)
Even if AI has looked like a good solution thus far, ask yourself how people will feel when – not "if," but "when" – they find out what you're doing. Is the answer "not good"? Then it's time for you to hit the off-ramp.
If your use of AI is above-board, read on:
6. Do you have access to data? Enough data?
AI works by finding patterns in data. You'll need access to a fair amount of data related to whatever you're trying to automate or predict. This can be past customer transactions (if you're in retail), flight arrival times (airlines), call records (telcos), whatever.
If you have a dream but no historical data, then building a custom AI solution simply won't work.
If you do have a decent amount of data, though … you're all set:
If you've made it this far, then you're ready to give AI a try.
What's next on your AI journey?
Your company's first AI project.
You can head over to "How Do I AI?" for my ideas on how to approach your company's first AI projects.
Additional reading material
I post longer-form thoughts about business and AI on my blog and at O'Reilly Radar.
Consulting services
For more personalized assistance, you can contact me to discuss a consulting arrangement. My work covers AI risk management and due diligence and also AI strategy and research.
If you're trying to start, restart, or evaluate your company's AI work, I can probably help.
Click here to open a contact form:
Hello! I'm Q McCallum.
I help companies navigate AI issues such as strategy, due diligence, risk management, and research.
I've assembled this one-pager guide to help you on your AI journey.
Where to find me: blog | newsletter: Complex Machinery | Bluesky | Mastodon | AI risk management & due diligence | AI research & strategy consulting
Disclaimer: This guide doesn't constitute consulting advice, nor does reading it establish a business relationship between us. Use at your own risk.
Copyright © 2024 Q McCallum. All rights reserved.
Header image credit Jigar Panchal on Unsplash.