Product Manager is the Hardest AI Position to Fill
Get a free early copy of my new book
Wondering “what do I need to know about AI?” but don’t have time for courses? Get simple, practical AI advice for product & corporate innovation teams. Get your free copy here.
The myth of AI tech skill shortage
The biggest tech companies — Google, Baidu, Facebook, Apple, Amazon, Salesforce, and Uber — have been hiring entire AI research departments and buying AI startups for astounding valuations.
These behemoths are slugging it out to make fundamental AI advances. Baidu can make millions if its speech recognition algorithms are 1% better than Google’s Speech API.
Your company is not Baidu. Your competitive advantage comes from applying AI to your data and innovating on your business model. Most of your technical headaches will be in ingesting, organizing, processing and storing data — not tuning neural networks.
Qualified AI engineers are in demand — but the shortage is no bigger than any other programming skill
Finding great programmers for anything is hard. It has always been hard. There are fewer AI engineers but also less competition for talent.
The harder position to fill is the AI product manager because they don’t exist — you’ll need to build one.
AI Product Manager: the hottest tech job in 2018
An AI product manager has two primary tasks:
- Defining the AI problem.
- Acquiring the training data.
Defining the AI problem
A well-defined technical problem — regardless of difficulty — is a solvable. A poorly defined technical problem is unsolvable no matter how simple.
The AI product manager needs domain expertise and an understanding of supervised learning to identify solvable problems. She needs to have familiarity with practical AI capabilities and how to filter those with the biggest business impact.
She needs great social skills like any product manager. And she needs a good bullshit detector for evaluating vendors.
Acquiring the training data
Programmers usually write code to manipulate data based requirements from customers or analysts. In supervised machine learning the programmers teach the algorithms how manipulate data by giving them thousands of examples.
This data is called training data and getting it is the most challenging, expensive part of AI.
The success or failure of your entire AI product — and probably the survival of your company — depends on your ability to get the best training data in your industry.
Your AI product manager owns this task.
Unfortunately many product managers have been building web sites and mobile applications. They often have a UX or marketing background — critical skills when goal is customer action.
In AI products training data is the new spec.
In AI a background in data processing or statistics is more valuable than design.
The AI product manager’s skills
It’s hard to define needed product manager skills because there are so many different ways to do the job. Some are great at working with programmers. Some are great spec writers. Others are business domain experts.
And some just great leaders — everyone rallies around them and works to a common goal.
AI is no different.
I will include AI Product Manager job descriptions in our book, but for now consider an example.
Example — Kaggle competitions
Your AI product manager should be able to create a Kaggle competition from scratch. For example, the Instacart Market Basket Analysis.
These competitions look a lot like AI specifications. To complete this goal she must:
- Know enough about AI to generate ideas with your business units.
- Quickly learning a business unit’s needs.
- Have personal skills to navigate organizational politics.
- Have enough statistic skills to design training data or manage a data scientist.
- Be able to negotiate success criteria with the business units.
- Identify new data sources and develop the business case for buying or building them.
Do you know someone who could design a contest like this from scratch? If so I’d suggest buying her lunch tomorrow …
Want a course for AI product managers?
Unfortunately there are no online courses for AI product mangers. I’m thinking about creating one. If you’re interested, please encourage me by leaving a comment below — it would be a big project and I’ll only do if if you’re interested. In the meantime the best AI 101 guide is my new book.
