As a lender, you may have noticed that AI is a particularly hot topic right now, and for good reason. Research (and practical application) is showing that AI powered underwriting models are vastly outperforming traditional underwriting models.
So now you face a decision: Do I hire a team of engineers and analysts to build a customized AI-powered underwriting model for my business, or do I buy a service like Trust Science?
Let’s take a look at a few important considerations for you to take into account before you make a decision:
AI Development will require a significant investment in time to build a solution – and the process is not over once you complete your initial build. The purpose of AI is to build a solution that will grow and learn based on your data – to have a loop that incorporates ongoing feedback to improve your solution and outcome as you move forward.
When you build an AI powered underwriting model, it is an ongoing investment. You will need a team to research, develop and then train your data, with ongoing testing, redesign and implementation of change based on what real-world data shows you. And don’t forget that the amount of data available to you is a big factor. If you are looking to access data beyond your own systems, it will take time for your team to identify tools and sources that can add to your data.
At Trust Science, our team of Data Scientists and Developers have built an out of the box solution for credit scoring models specific to each loan type – Short Term, Installment, Auto and Title. These “out-of-the-box” models are frequently monitored, tuned and tweaked.
We then combine the data you capture with our alternative data (financial & publicly available) and train our models to maximize the metrics you care most about. Oh, and we frequently monitor, tune and tweak these models too, to ensure we are returning the results we promise!
Access to Data
An AI powered underwriting model requires a significant amount of (quality) data to train, build and continue iterating. A company like Trust Science trains its models on an unfathomable amount of borrower applications, while a lender like yourself is constrained to a significantly smaller dataset.
Why is this important? We see a wider variety of borrower, economic and lender situations, which gives us much more certainty in scoring a borrower that you may only see once or twice a month.
Well-versed talent in AI is a lot easier to come by than it was two years ago, mainly due to the increasing number of options offered by online and traditional educational institutions. That said, the demand for AI researchers and developers is growing around the world leading to high wages and churn. Keeping data scientists in house is a costly endeavour that requires a strong vision, research funding and ongoing growth.
At Trust Science, we have a team of Data Scientists focused on research, development and ongoing improvement with our systems – it’s what we do. Plus, we are lucky to have roots in Edmonton, AB, where our team has access to a thriving AI community, thanks to a strong program at the University of Alberta, and the arrival of Google’s Deep Mind and AMII.
Building an in-house AI Solution requires ongoing investment and change. With past solutions, you could build an Access system in-house and bring in someone to support/change on an as needed basis. AI requires ongoing changes and updates to your model, so this is not a one and done process. You will need an AI team on staff to support the tool moving forward – both in terms of keeping your model up-to-date as well as ensuring that your integrations continue to work.
And don’t forget In-house documentation & knowledge sharing. If a team member were to leave, if they haven’t sufficiently documented their component or shared knowledge with your team, you can have a big gap to fill or re-design around.
Trust Science has you covered in terms of ongoing changes, improvements, and making sure that if we have a staff change, we have the knowledge to continue to support you.
Finally, lets touch on cost. Let’s be honest, hiring an in-house team is expensive. There’s no way to get around it. You are committed to that team until they get it right. On the plus side though, it is a fixed cost, so if your team (eventually) gets a model up and running and it has a positive impact, the incremental cost for each applicant you score is close to $0.
On the flip side, when purchasing an out of the box solution, you will be looking at upfront costs for services but significantly less than building out and maintaining a team. Once you are up and running, your team will have ongoing operational costs for usage and updates for the system.
Hopefully this has helped you make a decision of whether to build or buy. The most important thing is that you fully commit either way. Half-heartedly committing to either option will almost certainly result in a lower than anticipated impact and a lot of dollars (and time) wasted.