Part 0: Start with desire

This is a 5-part mini-series (with a bonus Part 0, this post) on my experience building AI/ML and data products over the past decade. I have worked in different companies large and small during this time and shipped several products in domains like healthcare, sustainability, open access publishing, and others. I have also founded two AI start-ups of my own and built 0–1 products within each. This series is a collection of insights I gained through this diversity of exposure to various AI/ML products. Some may find what I say here to be agreeable, others heretical. Nonetheless, I hope some of you will find what I say here to be useful in your own product development journey. Whatever your take, I’d love to hear your feedback and comments.

There are many pundits and frameworks out there — each of them have something useful to give. I have taken two of my favourites — the Design Thinking process for product development, and the Cross-Industry Standard Process for Data Mining (CRISP-DM) and bridged them in my AI/ML product development process, which I have used across all the products I have shipped. This mini-series will do the same in each part. If you would like a checklist of how to ensure you have done the proper due-diligence necessary to move on to the next stage of AI product development, please reshare and comment “checklist”. Although many of these steps will likely be done in parallel, and all of them in iterations, there is a rough logical order to these things. At every opportunity I will highlight the critical differences between “regular” software product development and AI/ML product development.

In Part 1, I take a step back and describe the “business understanding” part of the development process. This is strictly discovery — a critical step which a lot of product builders have a tendency to skip and go right into the ‘building’. In Part 2, I start to dig into the meat of AI/ML product development by having a clear understanding of the data.

Part 0: Start with desire

Part 1: Viability + Feasibility (Define problem / empathize)

Part 2: Focus on data / ideate

Part 3: Model / prototype

Part 4: Evaluate / user testing

Part 5: Deploy / iterative refinement

But first, let’s get back to the basics.

Part 0: Start with desire

As anyone who has built any product knows, great products sit at the intersection of desirability, viability, and feasibility. The thought processes outlined below help you get to the root of desirability.

First, what is a product, really? Clayton Christensen equates products or innovations to “help for hire” in his Jobs-To-Be-Done (JTBD) framework. At face value, the JTBD framework helps define the physical task-based aspects of a job that a user wants to accomplish. But before we do this, it is important to go deeper and listen to something more fundamental, more primal. Let me explain.

Desire sits at the core of our wants and needs. It shapes what we buy, where we work, who we love, it influences our values, norms, and the decisions we make. Each decision we make is at the service of changing our current state to some future state. Our desires heavily influence what that (end) future state is. The decisions we make, and therefore the actions or jobs we want done, are a means to get us to that end state. So it makes sense that these jobs to be done are not merely functional but inherently emotional.

For example, I am bored when driving long distances. I desire to be engaged (not bored). Therefore I go get a McDonald’s milkshake for the drive (to engage my senses). In spite of my preparation, I don’t feel confident about my presentation this afternoon. I desire to feel confident so I dress to impress for the occasion (to boost my esteem). I don’t like being excluded in social settings. I desire to feel part of (belong into) my progressive group of friends therefore I donate to Greenpeace. You get the picture. In essence, products are facilitators (means to an end) of these desires — therefore it is vital to consider not just the functional but also the emotional (and social) aspects of the jobs to be done.

So how exactly does one take the emotional and social aspects in addition to the functional when considering jobs to be done? Well, there is a method to this madness. Around every innovation, there exist the immutables — things that haven’t changed (much) throughout human evolution. These are intangible aspects of Maslow’s hierarchy of needs — our need for human bonding, self protection, sense of self-worth and status, disease avoidance, etc. Products that align with these immutable human “values” acquire stickiness and virality. Products that go against them are quickly forgotten.

Another rule of thumb is to look for the struggles. Humans are lazy animals, we always have been. By observing real world behaviours of how we make decisions or get things done today (and therefore transition from current state to a desired future state) and honing in on the kinks that exist we can look for clues to improve the process. Christensen’s “Disruptive Innovation” is a series of these incremental improvements piled heavily enough to overcome the threshold of indifference or non-action. These innovations help smooth out the kinks in our current processes and are better (cheaper, faster, more efficient, easier, etc.) than the hacks we use to circumvent these kinks today.

Once it is loudly known that there is a clear desire for the kind of product you want to bring into the world, we can start thinking about how feasible that is and how viable it would be for the business.

The next post (Part 1a) will focus on assessing the viability (business or market demand) of your AI/ML product.

♻️ If you like what I write, please reshare with your network.