​How AI Changes The Way Organizations Deal With Messy Data

Kunal Chopra is the CEO of Certivo, an AI-powered compliance system of record for global manufacturers.

A compliance manager at a manufacturer asks a supplier for evidence that a component meets a regulation. The reply arrives as a scanned PDF, a test report photographed on a phone, a bill of materials in a spreadsheet nobody else uses and an email that says the material hasn’t changed since last year.

The information is all there, but the manufacturer’s software can’t use any of it until a person reads each document, decides what matters and types values into the right fields.

I’ve spent the last several years working with manufacturers on product compliance, and this scene repeats at nearly every company I’ve visited. Capable people spend most of their time retyping information that already exists elsewhere.

AI-native software can alleviate these challenges, specifically through the growing trend of contextual intelligence, but this will also challenge the way work gets done.

Traditional software needs structure first.

Every system of record built over the last 50 years shares one assumption: Data must be structured before the software can make sense of it.

Businesses don’t produce information that way. MIT Sloan cites estimates that 80% to 90% of enterprise data is unstructured. Suppliers, labs and engineers send what their own systems produce, and much of what matters lives in email.

So, organizations pay people to translate and to build supplier portals and data-cleansing projects to push the structuring work upstream. However, this leads to repeated challenges: records lagging the documents they came from, two people reading the same certificate produce different entries and unclear answers when an auditor asks why a value was recorded.

AI-native software takes data as it comes.

This is referred to as contextual intelligence. A contextually intelligent system extracts what the task requires rather than everything on the page.