Eight Practical Skills needed for B2B AI Product Management

AI requires patience. That’s one of my major lessons as I spent the last 2+ years building AI enabled applications. Also, the level of uncertainties in AI algo development is just too high. So that being said, what exactly are the practical things a Product manager leading projects with AI components should know?

Skill 1: Manage accuracy requirements from stakeholders

Ya, it’s always 100% accuracy or 99.98% accuracy. Bruh, even 100 humans checking and re-checking the same item will not give you 0% defects. Most practical and successful projects having AI based elements start off with 40-60% accuracy and move to 80-95% accuracy.

The best way to make the stakeholders understand this is not by going into the accuracy game at all. Rather pitch prodcutviity imporvements from current situation and move on from there.

Skill 2: Move stakeholders away from PoCs as fast as possible

PoCs are were AI solutions goes to die. If you want your application with AI charateristics to actually thrive you need to move away from the PoC stage as fast as possible. In AI PoC is a very loose term. It can define anything from accuracy with x variables to trained models with y dataset.

Once you have an idea. Validate it with the engineering team on feasability and do a quick PoC with a barebones success crieteria and move it into full development. Trust me, AI is still evolving and we dont have enough data to say what kind of PoCs will work and what won’t.

Skill 3: Beg, borrow, steal for data

Unless you are working for an established firm, or a startup with influential investors odds are you will spend most of your time answering your engineers request for more data. In the beginning it becomes a chicken or egg problem. The second most difficult activity in these kind of scenarios is getting the first customer. The first difficulty is getting the 10 customers and making them use the solution. But to do that you need data first.

How to get data? That’s the skill. Beg, borrow and steal (legally :P)

Skill 4: Keep focus on a single ICP per major release

It’s easy for everyone to say it’s AI. Let’s make it work for everyone. No, No and a million times No. AI is not that strong enough to work for anyone. Pretty sure, it wont be in our lifetime. Models and Algos are best used when it’s built for specific ICPs. If you have four different ICPs depending on the persona have different models catering to it. I know it’s difficult. But over time it will be much more easier than trying to build an one size fits all solution.

Skill 5: Keep stakeholders from pivoting into an platform business

Just because we have a couple of Algorithms which have high accuracy dosent mean we have a platform business. The obession with platforms within AI is bordering on the insane. Unless you are Amazon or Microsoft don’t attempt that. Don’t be that person and make sure your stakeholders are not people as well.

Keeping your team inline is going to be crazy difficult because the clariyon call for plaforms is ever present.

Skill 6: Make stakeholders understand “data scientists” need data to be effective

We hired 5 data scientists. I think we can get inisghts now. No, we need data first. For data we need deployments. For deployments we need sales. Focus on getting deployments before getting data scientists. That is one of the biggest skills a B2B Product manager tinkering with AI needs

Skill 7: Make eveyone understand AI is not a single skillset

Just because we have a computer vision engineer dosnet mean she can build recommendation engines. Those are two different skillsets. Managing up and making the executive branch understand that NLP is different from computer vision is a skill we need learn. Of course, that being said this is more of a problem in established firms than in startups.

Skill 8: Make everyone understand the Algorithm alone is not the whole product/ application

Although AI is exciting, in most cases it’s less than 10% of the application. Just like an engine has a 100 moving parts, AI is one of the moving parts or rather it’s the oil which greases the parts. I think that’s the better Analogy. It’s easy to get carried away with AI Focusing on the business problem and making sure everyone understands we need to make sure the entire application/product is seemless is a major practical skill.