AI and Product Management
Discussing product management in general and what is different when managing AI products
What if you have an idea of a product based on AI? You should not throw all your resources into the development of that product with just a vision in mind.
It doesn’t matter whether you as an entrepreneur will handle this yourself or you will delegate this to someone else. You should pick your next actions wisely. Here is where a product strategy can help.
This post might be helpful for business owners and product managers willing to deliver a new product that is based on AI or anyone else interested in machine learning and product management.
This post has the following sections:
- Introduction
- Why having a PM?
- Product Management
- AI Product management
- Conclusion
1. Introduction
Companies’ awareness of AI and machine learning maturity level are increasing right now.
A lot of companies either use machine learning and data science in their everyday work or have launched an AI-based product.
Such a machine learning project or product needs to have a clear strategy to be successful. This strategy can be developed and maintained by a product manager.
In case you weren’t interested in machine learning before and you want to know more about it now, I’ll leave you this link. It is a bit outdated at the moment, but it covers the basics of machine learning in a fun and easy way.
2. Why having a PM?
To deliver a product, you need to provide enough value, reach the right market for your product and do this at the right time.
At the start, you might have a set of assumptions of how the product should look and who is that product for. Usually, those initial assumptions are wrong.
When building a new product you will compete with similar products and/or old ways of solving the same problem. To convince people to start using your product, you need to provide them with a lot of extra value.
Finding that extra value, the right distribution channels, and a market to start with is no trivial goal.
Those who have read about Lean Startup would say that it’s fine you just need to experiment as much as possible. But Lean methodology doesn’t tell you that every experiment costs money and time.
It’s not enough to fill in a Lean Canvas to make it. There is always a lot of exploration and prioritisation that is necessary.
Taking care of those tasks and other ones a product manager can save a ton of money and time for the company.
3. Product Management
The goal of a product manager is to make money for the company by delivering/growing a successful product solving customers’ problems.
A product manager is usually responsible for almost every aspect of a product. They are not idea generators. Their role is more about making decisions in a state of uncertainty.
There are different tools and techniques a PM should be familiar with. For example, data analytics, wire-framing, customer journey maps, customer development interviews, market research, and many more.
But tools and methodologies change constantly and the most important thing for a product manager is to develop a specific mindset that will help them focus on the right things:
- Keep in mind your company’s goals, but solve customers’ problems
- Distinguish facts from hypotheses
- Learn your product before investing in the development
Exacts functions and responsibilities of a product manager can differ in different companies, so here are the areas of a product a typical product manager usually owns:
If you want to learn to manage products, check out Go practice simulator — one of the best interactive courses for product managers and product analysts. It is available in both English and Russian.
4. AI product management
An AI product manager is a product manager whose expertise lies down in the field of AI.
An AI product manager is supposed to have the same set of skills and expertise as other PMs. However, machine learning models often operate with sensitive information and have some limitations which a product manager needs to understand.
If one wants to deliver an AI-based product they should take into account the following:
- Know your domain
As we discussed before, you always have a limited amount of time and resources and it is important to understand the field from the inside to decide which hypotheses to invest in and what to put on a roadmap first.
- Understand when you should use machine learning and when you shouldn’t
Having expertise in machine learning we may want to solve each and every task with it. But people don’t use products just because they have some sophisticated technology inside. They use products that satisfy their needs and solve their problems better.
For example, we want to develop a matching system that would match investors with startups. We might want to train a bunch of different models for them to process all data and make decisions on their own. This system would be very complex and hard to develop.
Wouldn’t it be better to provide users of the platform with several predefined filters? It’d be faster to develop and would give users control over the process of finding what is best for them? Of course, we can still assist them with machine learning.
- Understand the ethical side
Machine learning models sometimes use personal information as input, for example, images from smartphone cameras. This data is highly sensitive and you have to take into account the time and resources required to secure your users’ data and prevent it from lost.
Another aspect of machine learning models that often becomes a topic of discussion is its tendency to have different biases if not trained properly. These biases can be based on race, gender, accent, etc. It is not a product manager’s responsibility to prevent models from that. But it is their responsibility to make sure that the product is not causing someone to be treated unfairly.
- Understand interpretability
Neural networks are infamous for being a “black box”.
There are domains where it is okay to have a well functioning black box. We don’t really need to know how the model decides if there is a cat in the image as long as it does it correctly.
However, sometimes we can not rely on the model exclusively. In this case, we want to visualise what the model pays attention to and how it makes its decisions.
Imagine you are developing a tool to help doctors determine diagnoses. It would analyse the patient’s data and symptoms to return the diagnosis as an output. A doctor wouldn’t want to make any prescriptions without understanding what exactly has led the model to that diagnosis.
5. Conclusion
AI product management has some field-specific requirements. So my advice for a PM starting a new AI product would be to stay focused on customers’ interests. Keeping them in mind one will be able to find an answer to every question.
I’ve shared my personal opinion. Feel free to let me know what you think is also important for an AI product manager.
As always, I hope this post will be useful for someone. Please let me know if you have any questions. You can also reach out to me via LinkedIn
