What an AI Startup PM Does

What on earth does an AI startup PM do?

When I say I work as a PM at an AI startup in response to 'what do you do?',

nine times out of ten I get a reaction like this.

'Oh, AI is so hot these days~ but what does an AI startup PM actually do?'

I always found it tricky how to answer this question.

So 'I plan the AI that clients need, together with them, and manage the project.'

is roughly how I'd answer, but it never felt like a crisp answer,

and seeing the other person nod along out of politeness left me feeling uneasy.

AI is supposedly so hot, yet there isn't much content that kindly explains

'what exactly' a project manager at an AI startup does.

Of course the specifics differ a bit from one AI company to another,

but since the broad strokes are probably similar, I'll write a quick piece for 'those curious about how AI startups run projects,'

or 'those interested in the work of an AI startup PM.'

The peculiarity of AI projects… even the client doesn't quite know what they want.

AI projects fall broadly into two types.

1. New business (AI for a new service)

2. Work efficiency (AI to streamline existing internal work)

An example of #1 is a new AI-powered service like the AI Profile that Naver's SNOW unveiled, and

an example of #2 would be using AI to automate the per-product category tagging that shopping malls used to do by hand.

The biggest characteristic of these AI projects is that the requirements aren't clear.

The reason the requirements aren't clear is that

1. even the clients themselves don't really know what they want

2. and lack knowledge about AI. (e.g., can AI do this part too?)

For AI projects aimed at streamlining existing work,

an existing process they want to automate already exists,

so it's easier to grasp what output they want for a given input

(of course, I mean it's relatively easier ^^ ),

but for AI projects aimed at new business,

when you ask the client's contact 'so what is it you'd like to do?',

'Honestly I'm not sure either, and since you're the AI experts, I'd love a lot of help...'

is the kind of answer you get more often than not,

so it's hard to define the project with a passive attitude.

If the client doesn't know what they want, draw the whole sketch for them!

In the end, for an AI startup PM to carry a project forward well,

you need a 'kind, leading attitude' that guides the client through everything one by one!

Once you reach out, you need to clearly grasp what problem the client feels and what service they need,

then play the role of concretely guiding them on how AI can solve it

.

To do that, the areas an AI project PM must work hard at are as follows.

1. Understanding of the latest AI trends and AI technology

An AI company's PM must, quite obviously, have a basic knowledge of AI.

Things like what machine learning is; what supervised/unsupervised/reinforcement learning are; what models exist, and so on.

Also, because the AI domain changes so fast day by day,

it's important to keep track of how far which technologies have advanced and what new AI models are being used.

Based on this knowledge of AI technology,

you need to understand 'what you can do with AI.'

2. Grasping a variety of cases

As you follow AI technology, you naturally come across cases that use it.

If you keep a good grasp of what services other AI companies plan and offer,

and what AI companies in various fields have adopted and use,it becomes

rich material to propose to clients

, so it's good to reference well.

Also, as you look through various cases, it becomes easier to grasp what problems each domain wants to solve with AI.

3. Proactive problem definition and planning

As mentioned above, because of AI's peculiarity as a technology and the ambiguity of project definition,

if you only wait for the client's plan, the project definition is likely to drag on endlessly.

So, based on #1 and #2, you should propose various directions of what's possible with AI technology,and plan the service together with the client, concretizing input/output and so on

.

When a client throws out a naive idea, the PM grasps the essential need and structures it

, then

1) what data is needed / whether the data is sufficient,

2) what the current level of AI technology is,

3) how the service is done in similar cases,

4) whether it's at a technical difficulty the internal dev team can handle,

and after reviewing all of this comprehensivelymust propose a concrete, feasible AI project grounded in viability

.

(In this process, you may sometimes feel like you've become the client's internal PM 😅)

So far I've given a brief intro, among an AI startup PM's tasks,

to the project-definition process.

In the case of Dalpha, where I work,

we're not a company that advances/develops and sells a specific AI,

but rather one that builds custom AI for each client based on their pain points,

so there seem to be even more areas the PM has to think about proactively!

I can't say there are no difficulties in the planning and pondering process,

but as I talk with and plan alongside clients who had a lot on their minds — wanting to adopt AI but not knowing how —

I find myself wanting to make sure they really feel the benefit of AI.

Seeing a service built with the AI we developed launch,

and the pride and wonder I feel using it myself, also seem to be a driving force.

Anyway, I'll wrap up here!

Thanks for reading this long piece, and if you have any questions, feel free to ask 🙂

Go to the AI Store