Rainbird Technologies: Handling Complex Decision-Making at Scale

Ben Taylor, CEO Almost every organization in the world relies on their employees to make high-quality, nuanced decisions on a daily basis. Examples include assessing potential fraud, determining the value of insurance claims, and recommending products. These sorts of operational decision-making tasks must scale cost-effectively if today’s enterprises are to remain competitive tomorrow.

While most automation technologies are limited to automating simple tasks efficiently, it’s rare to find a technology that can automate complex decision-making. Robotic Process Automation (RPA) may enable simple processes to be automated, but RPA cannot reason in the way a human can, or cope with the complexity of multi-faceted decision-making. This has remained the domain of skilled humans.

A trained and experienced worker is typically better at handling complex decision-making than an algorithm, that is if you give the human enough time to be trained, to learn from their mistakes, and given access to all of the relevant data. Only then can a human excel at complex decision-making, and of course this takes years.

Businesses rely on the human brain to solves complex problems, at least problems that cannot be encoded in a decision tree—the ‘if-this, then-that’ approach that RPA follows. How do we do this? By subconsciously using probabilistic reasoning, evaluating various factors interloping in a vast web of logic. The end result: a seemingly sensible decision.

I say seemingly sensible because it transpires that there is more error or noise in routine decision-making that most organizations are aware of. Recent research at Harvard reveals that, while company executives estimate skilled decision-making to have a variance across the workforce of just 5-10 percent, the typical discrepancy in judgements is a staggering 40-60 percent, which is clearly not acceptable. This statistic does not vary much across sectors, and remains at this level even amongst workers who have been doing the same job for more than five years.



Automating this sort of complex decision-making is Rainbird’s purpose. Rainbird’s AI-powered software platform automates large-scale complex decision-making, by making decisions the same way that humans do. What’s more, Rainbird can justify its decisions in plain English. The platform has been shown to deliver far superior results when compared to human workers, in part because it is impervious to psychological pressures, human bias and noise.

While people are an organisation’s most valuable resource, the reality is that they are also expensive and fallible. No matter how experienced the individual, when it comes to decision-making, human error is inevitable.

Unlike people, Rainbird does not have any bad days. “The Rainbird Platform allows our clients to replicate knowledge from their best people and scale it, so that decision-making across their business is more consistent and of a better quality. We are automating types of decision-making that nobody thought possible, and well beyond the abilities of other technologies.” explains Ben Taylor, CEO at Rainbird Technologies.

The idea for Rainbird struck Taylor and his co-founder James Duez after they designed software for a claims management company, to automate the process of verifying third-party motor claims.

“There was no historic data to work from, so we took the knowledge from two subject matter experts and built it into an AI-powered platform that automatically and systematically evaluated the value of each claim, saving insurers millions and reducing operation costs,” recalls Taylor.

The solution was a huge hit amongst its clients who were previously dependent on large numbers of workers. Previous attempts to encode knowledge in the industry was limited.


The Rainbird Platform allows our clients to replicate knowledge from their best people and scale it

Programmers would try and encode simple rules into software, which was not scalable and frequently resulted in essential logic being lost in translation between the programmer and the subject matter expert.

Taylor and Duez realized that there was a universal requirement for an easy-to-use platform that could apply these principles to any type of decision-making. They knew that organizations in every sector would benefit from technology that could quickly replicate human reasoning within their operations. This led to the genesis of the Rainbird platform.

Today, Rainbird has a strong foothold in the world of automation and has executed successful projects in finance, banking, automotive, retail, and law.

A Business-friendly Build Process

The process of creating a system capable of automatically synthesizing human decision-making, all starts with the Rainbird Authoring Platform. This is an exceptionally visual web-based interface that enables business experts to visually model their knowledge to create what Taylor calls “knowledge maps.” Knowledge maps can be connected to each other, as well as to numerous external data sources and APIs, enabling Rainbird to reason over real-time data.

Even a single knowledge map is a holistic representation of business ‘know how,’ capable of answering hundreds of different questions. This is in stark contrast with a decision tree which is difficult to build, hard to maintain, and can only answer a single question.

Rainbird navigates through the logic in these knowledge maps, pulling in any relevant connected data as required. It uses the logic and data to deduce the best possible answer to each query, delivering every judgement with a confidence rating and a full rationale to explain its conclusions.

“The author of a knowledge map doesn’t need to build unwieldy decision trees and doesn’t need to chain all the logic together. They just encode the individual pieces of logic, and let Rainbird do all the mental ‘heavy lifting.’ This makes it both easy and rapid for the business to work with,” explains Taylor. Furthermore, if the author then makes changes to anything in the model, the implications of those changes are handled automatically without the author having to propagate changes downstream, as would be required with a decision tree.

Key to Rainbird’s success is that they have designed the Rainbird Authoring Platform for business people to use, not software developers. The subject matter experts are therefore central to the knowledge encoding process. Rainbird authors can toggle between a graphical view of their knowledge map, or can see and edit their creation as XML-based code which the platform automatically generates when the author works visually. This scripting language, is known as RBLang, and is Rainbird’s proprietary knowledge representation language. It may be code, but it is straightforward to learn and designed for non-developers.

It is the combination of this graphical authoring interface plus the powerful RBLang scripting option that makes Rainbird so accessible and quick to work with. Choosing XML as a language also makes it easy to import existing structured knowledge from other formats.

Once the author has finished creating their knowledge-map, they can publish their creation immediately, effectively turning their model into an automated decision-engine that is accessible via an API or a chat interface.

End-users can literally have a conversation with the knowledge map via an existing chatbot interface or a web assistant. Alternatively, they can connect to the knowledge map’s API from another application, so it can refer to Rainbird for an automated complex decision—all in real-time.

Rainbird can reach conclusions even if the data is uncertain or missing. What’s more, it can explain that uncertainty and provide a full rationale. That’s an important differentiator because it enables the technology to automate decisions in regulated industries like banking and lending.
Rainbird and Robotic Process Automation (RPA)

“We have now overseen projects where the Rainbird platform has been used seamlessly alongside RPA systems, enabling truly complex workloads to be completed without any human intervention. This has been shown to dramatically reducing operation costs while concurrently improving outcomes for customers,” adds Taylor.



If RPA is about automating tasks, by building a platform that can automate decision-making, Rainbird has undoubtedly pioneered the next level of operationally-ready technology. This has been evidenced by the company moving away from working with innovation teams, and instead working with CIOs, helping to solve some of enterprises most significant scaling and quality challenges.

Slashing Costs in Credit Card Fraud

Rainbird was the ideal solution for a major credit card provider who processes over half a million transactions every minute. The client had already implemented RPA to filter potentially fraudulent transaction data to offshore teams for further analysis.

These teams continuously monitor data feeds on multiple screens, in order to identify whether credit card transactions are actually fraudulent or false positives.

“The manual process was very time-consuming and only 74 percent accurate. In the worst cases, errors in human judgement resulted in unnecessarily blocked credit cards. This resulted in high levels of customer frustration and ultimately, lost customers,” reveals Taylor.

Almost all of this friction was eliminated with the implementation of the Rainbird platform, which integrated seamlessly with their existing RPA technology. The client was able to improve fraud detection and replicate best-practice analysis methodologies reducing the need to rely on offshore workers.

This transformation is significant given that up to 40 percent of cardholders are understood to abandon their cards after being inconvenienced by a fraud notification, even one that turns out to be a false positives. The benefits of faster and more accurate detection are significant for both the credit card company and their customers, reducing costs and improving customer experience, and retention.

The numbers speak volumes: Rainbird reduced the time spent on each case from 15 minutes to just 8 seconds and could automate 85 percent of all cases with 94 percent accuracy, way above the performance of the human workforce. The technology also looks set to reduce back-office processing costs by a massive 60 percent.

Those savings are impressive, especially when you consider they are being delivered alongside better outcome for customers.

What’s Next for Rainbird?

Rainbird is focused on both their research and development program and international expansion.

They aim to better serve their growing international customer base by growing their network of partners. These include Heron (www.heron.ai), a wholly owned London-based AI consulting and solutions company, Hong Kong-based JOS Group, covering Asia plus a string of partners soon to be announced in North America and Europe.

Meanwhile, the company’s research and development team are working on materializing the next generation of innovation and expect to announce a string of patent applications later this year.

As automation continues to sit at the top of every CIO’s agenda, and as RPA continues to show its limitations, it is clear that Rainbird’s automated decision-making platform is becoming a smart strategic choice. Rainbird undoubtedly heralds a step change for any enterprise choosing to take advantage of its capabilities.

After all, what CIO doesn’t want to both reduce cost and improve outcomes for their business and customers?

Company
Rainbird Technologies

Headquarters
London & Norwich, UK

Management
Ben Taylor, CEO

Description
Provides an AI-powered automated decision-making platform that can model human-like cognitive reasoning

Rainbird Technologies