In May 2020, McKinsey issued a report about the future of work in the wake of coronavirus, making key observations about firms effectively carving out a competitive edge in this market by changing the way they look at their data. Opportunities abound to mine the mountain of data on which most firms sit; it is to their advantage to introspect this data for ways to inform key business decisions. McKinsey urges:
“Companies need to incorporate new data and create new models to enable real-time decision making. In the same way that many risk and financial models had to be rebuilt after the 2008 financial crisis, the use of data and analytics will need to be recalibrated to reflect the post-COVID-19 reality. This will involve rapidly validating models, creating new data sets, and enhancing modeling techniques. Getting this right will enable companies to successfully navigate demand forecasting, asset management, and coping with massive new volumes.”
But for those who haven’t yet started mining their data, or are only at the beginning of their journey, choosing the right use cases to get started with can feel overwhelming. Since Salesforce launched its initial set of Einstein capabilities in late 2016, many of us at Silverline have been working closely with clients to identify and formulate use cases for the industries in which we specialize. We’ve been building a library of common use cases in hopes that this helps demystify the process for our clients. One of the key insights we’ve had is that AI experimentation is not linear; it requires an effective feedback loop and a strong data governance program.
In a recent interview, Giles Whiting, partner at AI investment giant SoftBank, conveys why their investment strategy has been so focused on companies that internalize this feedback loop to bring operational innovation to their teams. He says:
“I think we’ll look back on the COVID situation and say it was something of a watershed moment for AI. If there’s a group of companies that will end up being losers or casualties coming out of this situation, it’s the ones that aren’t thinking about how they bring in more automation and autonomy into their business. There are probably rare industries where things haven’t changed much and the old business models are going to work.”
AI, data, and finance — Oh my!
Our Financial Services team has been grappling with many of the technology trends that have rapidly emerged in the post-COVID reality via our Webinar from Home series. This week, I am excited to host a session with my friend, long-term client, and CTO of Stonebriar Commercial Finance, Steve Siler. We’ll be covering “Design Thinking Secrets for Your First Einstein Experiment.”
I recently sat down with Steve to talk about his AI journey and insights he has learned along the way. We are excited to share our process with other Salesforce trailblazers looking to get started and will be sharing some great resources, including our AI Use Case Template. But first: Let’s dig in and hear from Steve firsthand.
How did you get interested in AI? How did you get started?
Several years ago, it all started when I was working at an investment bank in Chicago. My team had been using Salesforce for many years and had several environments for our different lines of business. Salesforce had just come out with Einstein Data Discovery, we had a ton of data at our fingertips, and I was excited to get my hands on it to try and test things out.
I know you’ve never been one to shy away from trying out new features on Salesforce, so can you help us understand how you thought about getting started? I know lots of people may feel like they have data to work with but are intimidated by the process of beginning.
Right. To be transparent, our first attempt was not that successful because it was hard to understand how to use the tools when they first came out – and what use cases the tools should ideally be used for. Initially, we thought Einstein could ingest an entire database structure and derive insights. It turns out Einstein couldn’t quite read my mind like I’d expected, but it certainly could read a summarized flat file of straightforward data and start to confirm some of the suspicions we’d already had.
What were some of the conditions you felt positioned you well to get started?
Since I was working at an investment bank, the statistics and reporting we had access to were deep and rich. I had a ton of data on my hands — nearly a decade of deals. Mostly I was curious about what insights I could pull from the statistics, plus I knew AI provided a value proposition to the dealmakers at the firm who were trying to source comps. They were excited by the fact we were doing this work. Plus, it sounded fun! I like to tinker and knew I’d learn a lot in the process.
What were you trying to prove out? What problems did you try to tackle first?
We looked at operational insights first – how to predict which deals might engage or close. Ultimately, use cases like we first toyed with made it into Einstein Sales Cloud in later versions of the platform. But back then, we wanted to understand how to inform team performance and optimization. We suspected where the bright spots were and we knew which teams were highest performing by the numbers. What could the data tell us about why they performed best?
Fascinating use case. What were the cultural reasons that engendered this experimentation?
Both at my previous firm and currently at Stonebriar, it really began with a willingness to try new things and a mindset for experimentation. When I first started with Einstein, I was looking for confirmation bias because we wanted the data to tell us what we already knew. Financial Services firms are always looking for the edge, the insights into how to better perform and optimize.
So fast forward a few years, and you’re currently leading the technology strategy for Stonebriar Commercial Finance. What learnings did you pull forward into your new role? How did you approach this the second time around?
The second part of my AI journey began when I moved my family to Dallas to join Stonebriar. I went back to my roots in equipment finance and leasing with a whole new technology stack at my fingertips, and wanted to apply the same mindset of experimentation to a new fast-growing firm with colleagues I’d worked with previously. I knew I needed to create a mountain of data to get started, which meant a big digital transformation undertaking.
What types of data did you consider critical to this process, knowing your ultimate goal was not just operational performance, but building AI predictions?
Stonebriar had lots of data in disparate systems, and my job was to bring it all together in a thoughtful way. We first started with our existing book of business — loans and leasing contracts. Ultimately we layered in other domains of accounting data, operations data, pipeline and prospecting, market rate data, and ultimately data about customer engagement in the form of payment.
What are tips you have for other firms looking to make the jump into AI experimentation?
First of all, we had to get a lot of buy-in from stakeholders across the team to make sure we did it the right way. It took a bit of time to prepare a digital transformation roadmap, and what surprised me the most was that within three months of starting, we had enough information to begin the experimentation. That’s fast! A lot of companies don’t think they have enough data to start mining for AI, but that’s just not true.
One of the major lessons I’ve learned is that keeping it simple was the best approach. It helps you keep your data super focused and clearly aligned to the problem you’re trying to solve for. Then getting creative and thinking about other data to bring in helps refine your hypothesis and continue iterating on your experiment. This process has helped me to continue sketching out our digital transformation roadmap.
Further Einstein insights and experiments
Silverline understands the unique challenges faced by companies and firms in the capital market space and beyond. For more financial services industry use cases, as well as the AI Use Case Template we’ve used to approach these experiments together — be sure to watch the webinar, now available on-demand.