The AI Product Process with Microsoft Product Leader

This week, Product School hosted Angelo Liao, Product Leader at Microsoft for a special #AskMeAnything session. In this talk, Angelo answered everything AI products: from launch, to measurement, to implementing AI features within a larger product line.

Meet Angelo

Angelo Liao, Product Leader at Microsoft

Angelo is a knowledgeable Product Leader who works specifically with AI to create amazing and helpful products.


For five years he worked on Product at Microsoft. His most recent focus was building the AI Coach product portfolio. PowerPoint Presenter Coach was the first AI Coach at a major tech company and Angelo shipped it. The product listens to your speech and uses your video feed to help you prepare better for your presentation as you rehearse.

After this successful product, he thought up the AI Interview coach which he partnered with LinkedIn to bring to market. He also previously worked on the developer platform for Microsoft Edge where he worked as an editor and thought leader during the development of the web authentication API for browsers. After Microsoft, Angelo joined MetaLab, a leading product design and development agency.

Angelo is also a Programming Board Member at StartOut, where he helps and inspires early-stage entrepreneurs and companies. Not to mention he also is the Founder of Sparks, a platform that helps experienced professionals in the 50+ age group connect and network for potential work, advising, and volunteer opportunities. 


Angelo studied Psychology and Computer Science at Claremont McKenna College for his Bachelor’s degree, which gave him a unique perspective when it comes to making products and understanding his customer base. He is very skilled when it comes to data analysis and research.

What is your advice to product companies that wanted to implement AI for certain product lines within the company?

I get this question every year when we doing AI conference at MSFT and I’d do a “PM in AI” talk. I think if you are in a company that is slowly getting started on AI or haven’t done AI at all, the starting point would be to try to reuse existing API or products that simplify AI. You’d be surprised at what API use AI underneath the hood or feature that is related to AI.

For example, a lot of search is related to AI. Existing APIs like face detection model or products like this make it easy to implement AI quickly. Then you can slowly start building a team. I bet there would be engineers in your company who is interested in AI and want to try it out.

man with lasers scanning his face against a black vackground

On top of the broader business strategy here, the thing to know about including AI in a product is that:

  1. You have to think through  edge case scenarios and perhaps better to start with a recommendation type of feature because AI innately is deterministic
  2. AI needs a culture of data so you can see how AI adds value. For example, when you are rolling out a new change, it may be hard to see how this directly impacts user experience, say you are switching from a dumb recommendation logic to an AI-backed smart recommendation logic. The impact may not be as visible from a UX standpoint. You will need data to see how AI improves the bottom line.

Read next: Cracking the AI/ML Product Manager Interview

What advice do you have for someone transitioning from B2B products to an AI driven product?

These days I think a lot of B2B products are AI-driven products. I am part of one of the world’s biggest angel investment networks and I just saw a company that does AI-based doc classification yesterday. Often I find having a good understanding of metrics and data is a good starting point for Product Managers to add value to the data scientist’s work. So that can be your way in.

Data and AI are so closely linked that one can often help the other. If you work in a B2B product, there is a good chance there is a lot of data and metrics work you can talk about in your interview. Plus as I mentioned earlier, you’d be surprised to see how many places in an existing B2B product you can introduce AI into.

What major tips would you give to people who would like to transition to a Product Management role?

I found that if you are actively looking to transition, having a clear project and milestone can help make you feel a sense of achievement. I know a number of my friends/coworkers who are engineers and transitioned to a PM-like role. Within your company, you can start by perhaps telling your managers and PM you are interested in PM and start owning some spec telling or participating in user research sessions. If you are looking outside, consider a TPM (technical program manager) role which usually needs people who understand engineering better than typical PMs.

For side projects, I found a hackathon project or writing your own thoughts down help. There is a growing trend of people writing their own newsletter that includes product analysis. It is okay if people don’t read it as much but it is a good way for prospective managers to know your thinking. Who knows, my CEO at my new medium-sized company read my writing the first two weeks in!

person writing at a laptop

What are your tips for virtually networking with Product Managers with the intention of joining their teams? I have tried using LinkedIn with little success.

It may be helpful to join an event or join a cohort of people to network with. There may be some board opportunities. For example, I am on the programming board of StartOut, which is an LGBTQ entrepreneur organization. My work there is just helping them put together panel events. It’s been a great way to meet people.

Lastly, I found when directly messaging someone on LinkedIn, it is best to state your intention that you are interested in the company than trying to network. Since the pandemic, people are getting way too many LinkedIn messages that being more direct may help.

For more networking tips, read…So, Tell Me About Yourself: How to Crush Your Elevator Pitch

What’s a good go to market strategy for Business Intelligence tools built with AI capabilities? What metrics do you use for AI products?

Yeah, that is always a bit tricky. It depends on the situation. First of all, do you have a product that you can put together a really cool sizzle reel video for? If so, showing a before/after comparison can often make people see how magical AI can be.

camera lens facing us with someone out of focus holding camera from behind

Second, if the experience is not as flashy, try connecting with your sales leader to see what they need to drive sales. Sometimes the businesses just want expediency so you can produce some metrics like “this feature makes you do things X times faster” etc. I think the core KPI won’t change. The core business metrics like MAU (monthly active user) for a SaaS product is going to remain the same. We tried see/tried/kept as a good metric for recommendation type of product, basically how many people have seen the recco, tried it, and then kept it going forward. For the data pipeline, it depends on whether it’s a pre-collected dataset or if you are training the model on the fly from data coming in from the business.

Any final advice?

I guess my final advice would be that don’t think of product as too complicated. Even very senior tech people get it wrong sometimes. There seems to be a black veil on the product but when you look inside it, it is really not that complicated. It is always good to have a companion on the road so I highly encourage you to talk to more people instead of keeping your ambition to yourself. You’d be surprised at how many people are willing to help! And as always, if you have any more questions, feel free to dm me.

Enjoyed the article? You may like this too: