Introducing AI Risk Agents from Inscribe

Not only do they need to tighten risk controls to address the higher likelihood of fraud and credit abuse in an economic slowdown, but they also need to address a slowdown in business (due to the increased cost of borrowing) and a hyper-competitive market, where immediate approvals are make-or-break.

And they have to do it all with less budget and operational headcount.

But at the same time, one thing continues to be true:

This was an insight my co-founder Conor and I had in 2018 when we first started working with one of our earliest customers, Bluevine:

Even for some of the most innovative fintechs, large manual review operations exist behind the scenes to manage risk.

There are many risk tools and data sources. There’s also been an explosion of new risk vendors and data sources (many attempting to solve these very problems) since 2018. But every time a new vendor is added, a human agent still needs to interpret the data or insights provided by a vendor. As more solutions have emerged, larger teams have been required to collect and analyze the data.

And though our customers have operations teams reviewing this information, they’ve consistently shared with us that it’s a source of this pain, pressure, and stress on those teams due to the volume and variety of data. The expectations on these teams keeps increasing.

From generalized to specialized LLMs

Along with these industry shifts, we’re also seeing a major technological shift with LLMs.

They’ve become quite ubiquitous, primarily due to the key unlock over the past few years of being able to compete with humans on performance of fundamental tasks.

Subsequently, the worlds of academia and business began asking the question, “What does this mean for industry?” Across various different industries such as customer support (Intercom), legal and companies (Harvey.ai), and even general work assistance (Glean.ai), they're building and bridging the gap between the advanced LLMs and what this means for the industry.

Harnessing the power of LLMs has huge potential for financial services — and specifically risk — in terms of increasing efficiency and achieving more with less.

This is the unlock we’ve been waiting for in the technology industry for decades.

86% of risk leaders plan to use LLMs (but half don’t know where to start)

In a webinar we hosted with About Fraud earlier this year, 86% of risk leaders reported that they plan to use LLMs. But half of them said they didn’t know where to start. How do you translate this technology into something your team can utilize and benefit from?

At Inscribe, we have always worked hand-in-hand with fintechs, banks, and lenders to deploy reliable, safe, and scalable AI. We want to help our customers not only achieve their objectives, but shape the future of risk; a future where human agents aren’t completing these workflows themselves, but are instead complemented by AI Agents.

This new operating model would enable Agents to automate the mundane tasks that AI is well suited for: looking over huge amounts of data over long time horizons and coming to quick and accurate conclusions. And then their human counterparts can focus their time on the more nuanced work, analyzing outputs from Agent workflows, modifying them, adjusting the objectives, and managing the various Agents on their team.

With Agents, risk teams can increase their output without increasing budget or headcount, ultimately reducing fraud and credit abuse while staying competitive with a great customer experience.

Anatomy of an Inscribe AI Risk Agent

What will make these “superhuman” teams possible is that LLMs have the ability to interpret and reason at a level we previously thought only humans could.

Over the last six years, we've worked with many risk teams to create machine-learning models for document-related tasks and more. We've developed advanced models for detecting document fraud and credit risks. These models have been widely used and tested, showing they effectively solve the specific issues that risk teams face.

By incorporating these models into our AI Agents, we can handle more complex workflows. This allows our agents to provide comprehensive solutions, addressing complex risk scenarios with greater accuracy and efficiency, thus improving the overall performance of risk management processes.

Behind the scenes of our AI Risk Agents, there are various components.

• Proprietary models: Our proprietary machine learning models for fraud detection, document parsing, transaction categorization, cashflow analysis, and more have been trained on the largest and most diverse database of real-world financial documents, consistently delivering data our customers can trust.
• Third-party tools: Our agents can access various third-party tools and data sources from an LOS or CRM solution to credit bureaus and government databases.
• Code interpreter: We give our agents access to analyze and process programming code, converting it into machine-readable instructions.
• Reasoning: We equip our agents with a set of objectives and instructions, as well as provide it with a context. Then it also has a built in reflection and feedback loop, so that it can improve over time.

Our pre-trained AI Risk Agents autonomously perform routine tasks for onboarding and underwriting — working alongside your team to 70x their outputs. There are limitless ways to use these AI teammates, and we’ll work with you to determine where to start.

If you’re not quite ready to get started with AI Risk Agents, you can use our state-of-the-art machine learning models to detect invisible document fraud and get cashflow insights for better, faster risk decisions.

Get started on your AI journey with Inscribe

Whether you’re just getting started with your AI journey or are looking to leverage the latest AI and ML advancements, Inscribe can help. We take a collaborative, relationship-driven approach to delivering AI Risk Agents and Models to our customers.

We have found that this strong relationship is what enables us to build a product that really solves the most pressing problems for risk teams, but also ensures that you're taking the right steps to benefit from the potential of AI with successful deployments and strong ROI.

If you’re a risk leader or financial services operator interested in learning more, we’d love to have a conversation. Simply reach out to schedule a meeting with our team.