Cleansing Augmented Intelligence: How we solved a problem that has been burning for decades

In the first of a series of blogs, Adi Hazan, Founder of AnalyCat, Software and AI Thought Leader, explores the background to the ground-breaking CAI - Cleansing Augmented Intelligence - platform developed in partnership with Pro Global.

Insights
News

January 19, 2022

Share this page

Share on linkedin
Share on twitter
Share on email

There is AI hype everywhere and more solutions than a person can shake a stick at, but somehow a targeted solution solving an actual problem in insurance has remained elusive – until now.

Analycat and Pro Global have created CAI – Cleansing Augmented Intelligence. CAI is an augmented intelligence platform that solves a painful and time-consuming problem for insurers in the world of computation, namely converting spreadsheets such as property exposure schedules into fixed formats for databases.

This is called cleansing, and traditionally to get the job done, decisions are often made manually covering thousands of data points on every line of every sheet, as well as columns, again and again, year after year. The process is repetitive, prone to errors, and thanks to CAI now completely unnecessary.

Setting the Scene.

Since the advent of spreadsheets and computer models for pricing in insurance, marrying the two has been a significant challenge. The solution has been, and remained until late, throwing manpower at it. A human would look at each spreadsheet, geocode risks where needed, upgrade or downgrade them and then map columns from the spreadsheet into the fixed data format for a pricing model.

This process, cleansing, is an arterial wound in the body of insurance. It slows down quotes, introduces errors and never improves as each sheet is done from scratch for each client, each year. It is time consuming and ultimately a poor use of skilled human resources.

What was needed was a system that could remember what was input last year, automate the geocoding, provide an audit trail, spot new lines and offer them up for a human opinion as needed and then store that with the others in the memory bank for next year.

What made this more difficult was that although there is theoretically ‘big data’ here, no two changes are the same. Each line of each sheet – each property – is a single decision that needs to be stored and enacted.

It is ALL about the strategy.

A great deal has been written about ‘strategy’ throughout history, but not so much in recent years. In fact there has been a quiet revolution in human understanding of how to orchestrate and plan solutions that address specific challenges, but it’s not very well publicised.

But strategically, the path to success was spelled out 2,500 years ago in classics such as Sung Tsu’s the ‘Art of War’  – highlighting the need to form strategic alliances with those that could bring new advantages to a situation.

Far be it for me to liken the strategic partnership between AnalyCat and Pro Global with the art of warfare – but suffice to say that together we are greater than the sum of our parts. Indeed, by aligning a practitioner-led world leading insurance outsourcing specialist and consultant with a world leader in AI and developing custom approaches and handling small data and human decisions, the potential scope of applications in insurance is formidable.

In addition, execution was done by Wintellect in the U.S. who are highly regarded technical leaders in development and training. Each party brought their skills and unique expertise and perspective into the framework – a framework laser focused on helping insurers overcome the friction of manual data cleansing.

Agile Development.

John Boyd’s OODA loop for effective decision-making is a framework for handling rapidly changing situations on the ground. It is very much like the famous ‘agile’ programming method used today.

Before Agile, a piece of software was planned from start to finish meticulously on paper and then sent for execution with a specification or statement of work. Very often on completion, situations on the ground had changed such that the software was obsolete or irrelevant on completion.

Agile produces software in two-week ‘Sprints.’ This means that there are fortnightly meetings to decide what to do next and to show work to people at the coal face. Often the company starts to use the software incomplete as it becomes functional to find bugs ahead of time.

This keeps the software up to date, freshly integrated into the current needs and above all stops that end-of-project disaster loosely summed up as: ‘You build what?’ and it works very well in software companies as they are designed around these needs.

Agile, however, integrates poorly with current corporate management cycles. Companies work in quarters and years. The software is built in fortnights and sometimes can take several years. For a regular corporation, these timings need to be in a superstructure that synthesizes all these elements and timings. Boyd’s OODA loop provides guidance.

The OODA loop.

OODA is an Acronym that explains the four steps of decisions making:

  • Observe
  • Orient
  • Decide
  • Act.

This means your approach must evolve as you go. Sprints must be rolled up into quarterly meetings. Testing must be done by the people who have  the problem. The senior management must acknowledge in writing that this will be a two-year project. Milestones and versions must be in place, scaled, checked and measured throughout the process.

Time Horizon

  • There is more than one cycle
  • Keep small movements in line with larger actions
  • No-one ever failed after a long string of agreed successes

One of the key success factors behind CAI has been the similar mindsets of the strategic partners in the project. With CAI, all three parties were experts – not just in Agile and programming – but in the management of such projects.

Experts from Analycat designed what they thought  the system should be, but at the end of the testng and constant improvement process, the system had no resemblance to the initial design. Management was made aware on day one that there would be versions and setbacks but in the end, Wintellect were within 1% of budget and time on two versions! I highlight this because it is a rare tale indeed when it comes to technology development.

Some do’s and don’ts

So what did we learn along the way that we would recommend to others embarking on strategic partnerships? Some simple do’s and don’ts to bear in mind:

  Don’t   Do
  Assume you know what’s best for the organisation.   Observe: Go to the key decision-makers who you think might be interested in your idea and check to make sure.
  Stick to the tools you know and skills you have.   Orient: Master the correct tools and set up the important metrics. Checks and validation throughout the process are critical.
  Design the whole thing because you know what the business needs.   Decide: Based on your learnings, revisit your designs and make improvements. The best improvements are small ones that you can test individually.
  Build the whole thing in private and unveil it in a big reveal.   Act: Release your improved version to your users. They’re getting real value quickly rather than waiting for a future ‘silver bullet’ release and you’re able to learn again.

This is the heritage that Analycat and Pro Global have brought to create CAI, a solution that directly addresses and fixes challenges experienced every day by insurers  based on tried and trusted strategies borne out of other challenging arenas and hard-earned experience.

We are very proud of this ongoing strategic partnership and the potential that CAI now offers to help insurers turbo charge their underwriting performance. Together, we’ve solved a problem that has existed since the first spreadsheets were invented, but it’s not one that the industry has to manually tackle any more.

Get in touch to find out more

Data: Right is Might

Louisa Pavis, Head of Consulting at Pro Global, shares some insights on how insurers can get more from their data, and why it isn’t quantity […]

Insights

COVID-19 Claims Momentum will Continue

Causation remains king when it comes to understanding the anticipated wave of COVID-related liability claims. Mike Mackenzie, Head of Speciality Claims at Pro Global, assesses […]

Insights