Klarity is the world’s first AI-driven software that helps humans save time by reviewing repetitive documents. Founded in 2017 by an MIT engineer and a Harvard Law graduate, the company has raised USD 18 Million in a Series A funding round. Klarity has companies such as Okta, PagerDuty, Zendesk, Coupa, UiPath, etc., on its clientele.

Analytics India Magazine caught up with Aditya Thiruvengadam, founding engineer at Klarity, to know more about its product offerings and expansion plans.

Excerpts:

AIM: Why did you decide to build a startup around document review automation?

Aditya Thiruvengadam: The process of “document review” is the most soul-sucking job for many, but it is equally important for any organisation.

Every organisation relies on a painful, manual review process to gather mission-critical information from documents that directly influence different business departments such as order processing, billing, and revenue recognition – all these activities are extremely important and help businesses make better decisions.

We, at Klarity, imagine a world where AI can automate this manual, painful part of the jobs for humans and allow them to focus on high-value work. The problem had existed since time immemorial, but no one had come up with an end-to-end solution, which led us to build a startup focused on document automation.

AIM: How does Klartity leverage AI to solve the problem statement of ‘automating document review automation’?

Aditya: As humans, we look at documents to understand the content contained inside. At the same time, we are also observing the structure, layout and styling of the document. All this helps us to piece together the context we need to comprehend the information presented in the document.

The AI system designed by Klarity works in a similar, multi-modal fashion. We use computer vision to understand the visual structure of the document and natural language processing, which uses this information to identify the semantic boundaries within the document.

We collaborate with expert accountants and lawyers – to constantly innovate with the current state-of-the-art architectures like BERT, GPT3, YOLO – and build custom AI that can understand complex legal documents like the human brain.

AIM: What challenges did you encounter while building your product?

Aditya: Finance and accounting teams have to handle extremely complex documents – both from linguistic and structural standpoints. NLP, at the moment, has evolved to a point where it works perfectly well with plain and short blocks of text. The same holds true for computer vision, which delivers great results when the documents are fairly well-structured.

Real-word documents are complex, are not structured well enough, and most of the time, are long enough that defeat most out-of-box NLP solutions.

We have spent multiple years researching these fundamental challenges and perfected our AI system capable of handling complex real-world documents.

AIM: What kind of products/ services do you offer, and how do they integrate with existing environments deployed at enterprises?

Aditya:: Presently, accountants spend more than 30-40% of their work hours in organising, reading, and managing documents and spreadsheets with document summaries. They have to manually read Order Flows, Purchase Orders, and Master Agreements to extract information and validate transactions.

The software and AI developed by us completely eliminates the need for human review for more than 85% of documents, which, in turn, saves thousands of man-hours every year. Accountants need not spend their valuable time reviewing documents and can instead focus on complex, high-value accounting tasks. 

Our solution connects with existing systems like Salesforce, Ironclad – where documents are stored and automatically categorised into deals. The AI system automatically reviews each document, populates an accounting checklist and alerts human users about any documents that may have a non-standard term or an exception.

Our solution can also generate automatic summaries of the documents and deals reviewed and can push the data into our customer’s ERP system, such as Netsuite and SAP.

AIM: What makes your product stand out compared with existing products in the market?

Aditya:: All the existing products in the document review automation space focus on generic ‘partial automation’ or ‘workflow-based automation’ – both of which need a human user to review and lack scalability due to the limited value of the product.

We focused on ‘full automation’, one vertical at a time, solving real challenges associated with document review faced by our customers.

Although ‘full automation’ is still in its infancy, our solution is still able to achieve over 85% accuracy, much higher when compared to existing solutions. This makes us unique and better than other products in the document review automation market.

AIM: What are your plans in terms of new product offerings, entering newer geographies, and increasing headcount?

Aditya:: We are currently focusing on our first vertical – automating document review for finance and accounting teams to assist them with revenue accounting for SaaS/software companies in the US. However, every industry across the globe is struggling with automating document reviews.

The technology we have built is extremely generalisable and can be applied to future use cases like Vendor and Lease Agreements. We plan on expanding our scope, one step at a time, and hope to solve the problem revolving around document review automation across industries.

We plan on growing our team of data scientists and machine learning engineers in both San Francisco and Bangalore.