Armilla offers verification and warranties for enterprises using AI models  

As enterprises across sectors look to adopt new AI tools and technologies, one issue standing in their way is concerns among leaders and employees about the trustworthiness and risks of these systems. 

In response to this challenge, Toronto-based startup Armilla AI’s insurance spinoff, Armilla Assurance, this week announced the launch of Armilla Guaranteed, a new warranty service aimed at providing certainty around the performance of AI products.

Lawsuits over AI models increasing

Determining liability of artificial intelligence and machine learning-based systems is no longer hypothetical or rare. Last month a judge said that Facebook could be sued over discriminatory algorithms, insurance provider State Farm is facing a racial discrimination class action concerning its Black agents, as well Wells Fargo is accused of discriminating against black individuals in offering home loans. 

“The lawsuits have already started,” said Karthik Ramakrishnan, co-founder and CEO of Armilla Assurance in an interview with TechForgePulse. “We just see them in these discrete events. But if you look at them as a whole, there’s a whole host of these liabilities that are appearing today.”

The root of the complaints

At the root of all these complaints are the deployment of automated algorithms which are used to help those companies make decisions, and that’s why it’s important for enterprise decision makers to invest in trust and safety products like Armilla Guaranteed.

Armilla’s new offerings consist of two key pillars — an independent assessment of AI models through Armilla’s evaluation platform which will confer a “verified” status to algorithm developers upon successful review, and a warranty, Armilla Guaranteed, to protect buyers if a “verified” model does not meet defined key performance indicators (KPIs). 

If a guaranteed model falls short of its promised KPIs, the buyer will be reimbursed the license fee through Armilla’s reinsurance provider partners Swiss Re, Greenlight Re and Chaucer.

“While AI creates a myriad of opportunities, it also carries material risk which is where the (re)insurance market steps in to provide cover where it’s needed, helping to close an ever-increasing AI protection gap,” said Hayley Maynard, Head of Innovation at Chaucer in the release.

The goal is to give both vendors and customers greater confidence in AI solutions by directly addressing issues of quality, accuracy and risk.

“What we’re doing is unlocking the adoption curve,” said Ramakrishnan. “If you do these right things, you’ve tested your model, you’ve put the right guardrails in. If things go wrong, despite all that, then we’re here to protect you. Even for the initial adoption of AI, we need to have some of these early solutions and insurance even though it’s the last mile. We do need them now.”

Peace of mind for AI pioneers

Importantly, Armilla Guaranteed isn’t liability insurance — rather a warranty product to allow improved evaluation of models. Armilla’s evaluation process involves both qualitative and quantitative analysis of models tailored to their specific use case. 

Assessments through the Armilla platform can be completed in just a matter of hours, providing a deep understanding of any risks. Models that pass will receive the “Armilla Verified” badge.

“What we bring to the table is the underwriting expertise. The risk assessment of these models, quantifying that risk and pricing that risk appropriately,” said Ramakrishnan. 

Is the risk contained enough and predictable enough so it can be insured?

Promising early customer testimonials

Automated medical reimbursement platform BUDDI AI believes that working with Armilla has been a “defining moment,” said Ram Swaminathan, CEO and Co-Founder of BUDDI AI in a news release. “AI-driven automation in the healthcare industry is riddled with risks stemming from bad clinical data, data bias, and misinterpretation of clinical context and reimbursement guideline variances.”

To make it easier to convert patient records into medical codes used by insurance companies, reducing denials by 60%, BUDDI AI uses deep learning algorithms combined with sophisticated systems built by experts — offering contractual guarantees of over 95% accuracy on medical codes and insurance claims for more than 70% of the monthly volumes. 

Now BUDDI AI’s model benchmarks are backed by Armilla product verification service and the Armilla Guaranteed warranty.

By directly addressing concerns around AI quality and risks, Armilla Guaranteed aims to help boost adoption of responsible, trustworthy AI solutions. The new offering provides a model for the industry as concerns continue to grow over how to ensure AI system performance and outcomes. With insurance backing, it intends to give both vendors and customers greater confidence and protection.

“We’ve been missing a shield to protect us from this inherent risk and give our customers comfort for those ‘what if’ scenarios. This warranty offering is what the AI industry has been waiting for and has helped us greatly in cutting the sales cycles with hospitals,” said Swaminathan.