Curvestone AI Bags $4m Seed For Workflow Automation

Curvestone AI, a workflow automation platform with a strong consultative aspect, has raised $4m in Seed funding led by MTech Capital. It’s known for working with firms such as Walker Morris, Stephenson Harwood and Browne Jacobson.

Before raising external funding ‘Curvestone reached profitability and grew revenue 7x in 12 months’, they said. The platform now processes ‘billions of tokens quarterly’ across legal and financial services workflows, they noted.

One key point they made is about how accuracy falls away when AI performs a string of tasks – and that’s one area they want to address.

The company stated: ‘While individual AI tasks might achieve 98%+ accuracy, in complex multi-step workflows this small error rate compounds — and by step 12, overall accuracy can fall to just 30–40%.’

And that’s a key point, especially as we move more and more into using agents, i.e. we need to avoid compounding errors, especially in legal work.

The business was founded by brothers Dawid and Sebastian Kotur after running AI automation programmes at PwC, Metro Bank, and GKN.

Some of the achievements they wanted to share include helping Walker Morris to cut service agreement reviews by over 90% (from 4 hours to 15 minutes), and helping Stephenson Harwood build a multistep compliance workflow addressing compliance with DORA EU rules. 

Dawid Kotur commented: ‘You can review everything and go broke, or cut corners and hope for the best. AI that actually works changes that equation by handling routine validation at scale while humans focus on the complex cases that need expert judgement.’

AL asked him some more:

Why do the funding now? And what is the cash for?

We’ve proven the model works – this funding lets us scale faster to capture the opportunity in front of us. We’ll be using the funding mostly for marketing and sales to our target markets.

Will your work help make agents more accurate?

It’s not about making agents work in isolation, it’s about knowing when to deploy them and when to layer in other approaches to hit the accuracy and consistency thresholds regulated industries require.

—

Congrats to them.

You can see more about the company here.