Example method for AI/ML use case discovery: Run a Possibilities Session

How do we come up with new machine learning use cases? It’s a hard question to answer for business teams and machine learning teams. A lot of times, teams are simply not used to thinking about new opportunities and new possibilities. And let’s be honest, you can’t just go up to colleagues or clients and ask: What can machine learning do for you??

I’ve been a design executive on machine learning and data science teams, bringing together the disciplines of data science and design strategy to ensure that machine learning initiatives are solving people’s needs. One team I had the pleasure of being part of was called data+design in a large financial services corporation; and we were tasked with developing new machine learning capabilities to augment our partner business teams. How did we go about finding opportunities to do that?

The data+design team began collaborations with our internal business partners with a Possibilities Session. The objective of the Possibilities Session was to build a list of potential ML initiatives to address current business or client issues… and do it collaboratively. Here’s a brief run down of some activities in a Possibilities Session to integrate the knowledge of the business and machine learning teams and articulate high-value machine learning initiatives.

1) Share out each team’s capabilities

If this is the first collaboration, it would be wrong to assume that the business team knows very much about data science or machine learning. Likewise, the machine learning team might not know exactly how the business team functions, be it client service, marketing, customer acquisition, corporate finance, fraud, or any business team for that matter. A primer for both sides can go a long way in setting context about how they might bring their knowledge together.?

2) Create a shared vision of partnership

Both teams probably have different motives for working together, not to mention all of the different stakeholders they will need to answer to. Time to be explicit about the benefits for each team individually, and what both teams will get out of this by working together. It starts to evoke the emerging culture of the teams and how they might interact moving forward. This can reveal a lot about early alignment, expectations, and intentions.

3) Develop an initial POV of customer/client needs

A good discussion to start with is the team’s perspective on what they think their customer/clients needs are. What are their clients trying to accomplish? (Secret: this quickly turns into issues affecting the business team.) Then build out the context for most pressing needs: What is going on with the client? Why is this important? What would it mean to get this “fixed” with ML? What’s the benefit for the client and the business? Why haven't we fixed this yet? Any ideas on how we would go about taking care of this issue? Think how helpful this is for the business team to get these issues out in the open. Think how much eye-opening and helpful context the machine learning team is getting by being part of this discussion.

4) Develop a list of potential projects

Turn the top prioritized needs into initial project briefs that describe what the teams might work on together. Go ahead… write up mini nascent project briefs with as much information as possible. This is why you need both the business context and the data science thinking together. The teams won’t be able to think of everything and that’s okay. The goal is not perfection, but teams should attempt for clarity so that when they come back to these mini briefs in a few days, it still makes sense.

5) Plan how to move forward together

The Possibilities Work session is not a “deciding” meeting. The whole point is to come up with a list of potential use cases for machine learning (and get some momentum for the teams). But nascent baby machine learning ideas are just potential use cases. They will need to be fleshed out in terms of the size of the problem, data and technical assets, potential gain, key showstoppers, resourcing and timeline estimates. Once the team gets all of the nascent ideas into more robust articulations of ML initiatives, you can have a coherent and strategic discussion with stakeholders about which initiatives matters and which are worth pursuing.?

Did I mention a magic ingredient in the Possibilities Session? A design strategist, who is a the human-centric expert and knows how to excavate deep and technical thinking from people's brains and enable them to collaborate with the right activities and discussion. Typically, design strategists have the ability to design a work session that pushes teams through ambiguity, allows for context and creativity, and then can facilitate people’s thinking toward concrete outcomes. Pairing a design strategist with the machine learning team can augment how machine learning teams and business teams work together, by making space for strategic context and helping them both articulate what’s possible.?

Certainly there are many methods for teams to come up with machine learning use cases and new machine learning product initiatives. I thought I’d share just one way that the data+design team found success articulating new machine learning use cases with our business partners (and by the way, we did this inside of a more culturally conservative financial services corporation). So try it! A Possibilities Session will mix business insight, with data science know-how, all while setting the stage for the right partnership and collaboration.

This is article number 2 in a series where I’m sharing stories and experiences about combining design strategy and data science in order to achieve human-centric machine learning. Topics will stem from working with business partners, partnering with data scientists, understanding people's needs, visual storytelling, ML product development, and all of the invisible work it takes to get new ML initiatives off the ground. Stay tuned for more!