AI Project Management
If you have all the project management skills and the know-how to manage your current initiative, then you may be able to deliver the expectations today successfully. The question is: are you ready for the next initiative, which will most likely be murkier and ambiguous?
This article was originally published on Medium: Emerging Careers in AI Project Management, by Rajesh Verma
Change is hard
Project Management, in the traditional sense, relates to managing Scope, Schedule, and Budget. These are the constraints that Project Managers are trained to work with. Folks who have spent some years managing a few projects will eventually end up managing a portfolio or program. Leadership, delivery, failure management, direct & indirect responsibility for resources, cross-functional relationships, communication, risk mitigation, etc., comes with the territory.
This whole mesh of activities and constraints builds a mindset of thinking within the boundaries, whereas AI is all about thinking outside the boundaries.
The challenge is to break free of what you are good at and embrace the unknown, to add value beyond imagination.
To shift from a traditional project management mindset to managing an AI project, you need a different way of thinking and a different skillset.
If you are used to the familiar waterfall or agile methodologies of managing a project, let's put that aside for now, none of that will work, at least at the onset of an AI initiative.
The two other things you should not do are: update your role in your resume and assume that the shift is going to be easy. Change is hard; changing mindset is even harder.
Three major challenges
Lack of clarity: Project Managers are trained to think within bounds. AI, by nature, is exploratory and needs an out-of-the-box thinking mindset. The ability to weave and communicate the benefits of AI into the day-to-day activities of an organization is at a very superficial level at best, that is — if it exists.
Weak conceptualization: Most AI implementations are copycats. Someone did something. It looks cool, so let's do it too. Well, guess what, so are all your competitors. And yes, they all have the same belief that their solution will be the best. While learning from others and keeping an eye on what's happening in the industry is a good thing; doing what everyone else is doing might not be the only route you want to take. After all, every organization is different. Your abilities, problems, vision, mission, and goals should be the main drivers of your strategy and concept.
Poor Integration: An AI solution does not deliver in a silo. To generate value, AI models have to be designed to better integrate with the operational production systems. And, no, building everything from scratch with the hope that you will have a perfect system is a myth. Well, to be fair, it might be a source of some additional revenue, but to your surprise, you will have one more set of production systems to manage, which will have similar issues as other systems. In short, you have further fragmented your application set.
To successfully manage and deliver an AI initiative:
We will need the ability to understand and narrate a story for your audience — Content Writer.
We will need a better understanding of Data Science, the value it adds, how the value will be consumed, and obviously implementation — Data Science Product Manager.
And we will have to learn to integrate AI solutions to the mother-ship for it to generate exponential growth, organizational interest, and motivation for future AI initiatives — AI Operations.
Content Writer
One practice at Amazon is that they write the press release first. This helps build clarity in thinking, developing an idea, and puts the customer-first mindset at the core of the strategy. Your customer is interested in the value your solution will add to their life. How you get there is for you to solve.
Read: Jeff Bezos Requires Amazon's Leaders to Perform This Powerful Ritual Before Launching Anything
This approach forces you to think outside-in and narrate the story for your internal audience, for them to engage and be motivated as they see the full picture beforehand.
The entire journey has to be narrated well for the uninitiated, and the experts both for them to appreciate the value the AI initiative will bring to the organization as a whole.
Lack of clarity in storytelling can lead to the immature death of an idea. Storytelling is not about building a castle in the air. It’s about narrating how you plan to build the castle and the enjoyment it will bring to the owner.
Data Science Product Manager
The idea of product management is old, so is project management, and to a certain extent, data science too. The main issue is that we still continue to think of them in silos. Product Managers think only about the product design and consumer aspect of it while leaving the execution to the Project Manager.
Project Managers manage the effort from start to finish, scope, schedule, and budget, without a thorough understanding of the customer experience and value addition, depending heavily on the technical team for correct implementation.
The Data Scientist develops a model with little or no expertise in deploying it to a production environment. The technical team responsible for production systems, on the other hand, has no understanding of the AI model but is being asked to make it usable.
When all these roles work in isolation, all experts in their own stream of work deliver a half-baked solution that adds little value, is high maintenance and is the start of the blame game and finger-pointing eventually leading to failure.
The Data Science Product Manager is a little bit of everything. On a T-shaped skills factor, it is the horizontal bar. In this role, you need to understand the six Ws: Who, What, When, Where, Why and How about your product, the utility or value your product adds to the consumer's life, the design aspect of it, the solution to the problem, the data science concepts that will help build the solution and most importantly the end game: going Live.
In short, you are the owner. You need to pay the bills, change the battery in the fire alarm, change the bulb, call the plumber, the electrician, the AC guy, the cleaning crew, and mow the lawn, you get the point.
Read: Interview: What does a Data Science Product Manager do, and how do you get that job?
AI Operations
Developing an AI solution as a proof-of-concept is easy. The hard part is to integrate that solution into the main production system that brings in the bread and butter both.
It is a fallacy to believe that the existing system will continue to keep the lights on while the AI solution will work in isolation and eventually grow big enough to dwarf the current system.
For the successful implementation of an AI solution, it must integrate with the mother-ship. This is the most critical step in adding longevity to the value addition. Also, this is the most undermined stage in product development and is seen as a step for the not-so-bright to do. (Yes, the feeling of eliteness in the AI team kills the solution.)
Integration is only the first step in making an AI solution real. The ongoing maintenance, upkeep, change management, flexibility to meet changing business demands needs a team with AI Operations expertise. This team must have the talent to build a comprehensive data analytics platform and AI strategy, which includes the AI/ML solution and IT systems while supporting Continuous Integration & Deployment (CI/CD).
Read: 12 Steps to Excellence in Artificial Intelligence for IT Operations
Summary
The AI/Data Science industry is too focused on the technical aspect of problem-solving. But, with the level of platform maturity, it's time to ask if we are solving the right problem?
For the willing Content Writer, Data Science Product Manager, and AI Operations are the potential career paths with broader skills in play.
