About AI product management
Marco van Hurne
I build AI companies | Data Science Strategy | Data Governance | Certified AI Compliance Officer | The Future Group | Beyond the Cloud | Author of the Machine Learning Book of Knowledge
Buckle up, buttercup!
Todays blog is about AI Product Management
First things first, let's set the record straight. There's a big difference between a PM who specializes in AI (the AI Product Manager extraordinaire) and a PM who's just using AI to crush it at their job. In this short text, I am focusing on the former - the crème de la crème of AI Product Managers.
Let’s find out.
Before we start!
If you like this topic and you want to support me:
AI Product Managers work on three main flavors of AI-powered products. You've got your Applied AI products, where AI is the key ingredient. Where the AI needs to be built to make the magic happen. Think of things likeNotion AI and LegalZoom's Doc Assist.
Then there are AI platforms, the heavy-duty tools for developing and deploying machine learning models
AI products are a whole different thing compared to your typical software products. The core principles of product management still applies, but AI PMs need to be truly data focused. They've got to know the ins and outs of data collection, analysis, and deployment
Let's compare it to "traditional" software development
Hold on, I am putting on my lecturers' hat and adjusts my glasses.
Generative AI is subset of deep learning, and it uses ML models to create "new" content based on existing data. Focus on the quotes people, and check my latest article on the death of creativity thru AI monoculture.
Whether the product is based on machine learning or deep learning, you should know that AI products development is a whole different thing than traditional software development.
And here is why:
First, development stages. Traditional software usually starts off with a functional spec and goes through design, coding, testing, and release phases. But with AI-based products, it's all about performance accuracy
Second, the people involved in development. Traditional software products are built by cross-functional teams
Third, data dependency is another big one. AI-based products rely heavily on massive amounts of data to train AI models. The data is what ultimately determines the product's functionality and user experience
Recommended by LinkedIn
And fourth, the user experience is where things get really interesting. Traditional software has predefined user interfaces and workflows, but AI-powered products like chatbots, tax advisory tools, or slide generators offer user experiences that adapt based on a user's behavior, or preferences, and of course on the users' inputs. Customer satisfaction with AI products is all about the perceived accuracy and trustworthiness of the data. If you want to know more about designing Used Centric AI products, check my article on it.
Fifths, interpretability. That is another key difference. Software products are generally more explicit and easier to understand, so users can easily grasp the cause and effect of a particular action. But AI products that are based on deep learning models, are trickier to explain. AI PMs need to help the AI understand the user's intent and context. They also need to make sure the user isn't left scratching their head about how the AI arrived at its decisions. For further reading, check my article on observability.
Sixth, testing. That is different too. With software, you've got a predetermined testing plan and test against set inputs and outputs. But when testing AI products, you're often evaluating performance based on unseen data, edge cases, and potential biases in the data. Businesses have their own evaluation metrics to test the robustness of an AI product or system.
Seventh (Next time I'll be using numbers again), let's talk about risk. AI systems are inherently riskier due to their non-deterministic behaviors. Algorithms using different paths to arrive at an outcome can pose risks from regulatory and public perception standpoints.
Last but not least, adaptability. AI-based products are typically designed to learn from new data and user interactions. This makes them grow (and not the other way around - synthetic data causing model collaps), and thus they can create new content or recommendations on the fly. Compare that to traditional software products that can remain relatively static once deployed.
So there you have it - a quick inside look on AI Product Management.
If you want to know more. Check out these links.
Well, that's a wrap for today. Tomorrow, I'll have a fresh episode of TechTonic Shifts for you. If you enjoy my writing and want to support my work, feel free to buy me a coffee ♨️
Think a friend would enjoy this too? Share the newsletter and let them join the conversation. LinkedIn appreciates your likes by making my articles available to more readers.
Signing off - Marco
Top-rated articles