Eliot Benzecrit, Co-founder & Director

Most trainees and junior lawyers will know that, right at the end of a transaction, one of the most nerve-wracking parts of their job is about to start. The pro-formas at the back of the main transaction documents, which were meticulously negotiated by both sides, will have entered real-life legal service. Those board minutes, application letters or notices will need to be “split out”, populated with data found in a spreadsheet, then e-signed. Sounds easy, banal even – that’s until you suddenly realise that this needs to be done for each of the 250 individuals involved and each needs slightly different wording. Oh, and the client’s directors who need to countersign – they just want to sign their document pack once. Cue the DocuSign counterparts and a bucket-load of anxiety...

The irony is that the process outlined above will happen in most corporate law departments around the world, meaning it’s formulaic – so in theory it can be automated, banishing the anxiety with it. In 2019, this exact flow was outlined to us by (then) Allen & Overy when Avvoka was part of their Fuse incubator – mass document generations were an ordinary part of the rhythm of a transaction and improving efficiency (and most importantly, accuracy) would be a major ROI victory for automation tools like ours.

The biggest “win”, we were told, would come where most automation tools struggled: quickly ingesting data in Excel format – the amount of coding to take the pro-formas to “Excel-ready” meant they were never suited to end-of-transaction deadlines. Our solution: every template, once automated, could be extracted as an Excel file of data points on a single download click. The resulting spreadsheet acted like a hopper – the data points that needed to go into each split-out pro-forma would appear as column headings – “all” the trainee needed to do was populate each row with the data they received from the bank, accountant, etc. for each participant/entity and hit upload.

By all estimations it was a massive time-saver – confining the manual amendment of square-bracketed pro-formas to history. All-nighters turned into a few hours – get the data ready, upload and then send to DocuSign.

Though it turned out the “send to DocuSign” aspect wasn’t easy at all – in fact, it was an envelope minefield that required hours of manual preparation – only to then find out DocuSign lacks a “send to all” feature, or the ability to properly streamline the counterparts aspect. Solving this particular problem spurred us to build an entirely new product: Massdraft. The application is entirely focused on mirroring how a law firm carries out signings – recognising the staging gates and the need for real-time execution updates that mean a closing can occur smoothly.

So, you’d assume that this means we solved the mass-generation workflow. The honest answer would be – no, we seriously improved it. The problem lies in the assumption that “all” the trainee needed to do was populate each row with the data they received from the bank – it turns out this is a massive oversimplification. While what Avvoka produces each time is a consistent “hopper” Excel for the data, what the junior lawyer receives from the client varies massively transaction by transaction. There’s simply no quick copy and paste – the data terms are often very inconsistent, so the task of transposing what’s received into what Avvoka expects takes time, care and effort.

This is why we’re so excited to launch headless Massdraft – the ability to drive the process from your generalist AI tool of choice – Harvey, Legora, Claude or Copilot – and still combine the precision and UX wins that Massdraft shines at (perfectly created documents, live signature tracking, etc.). The LLMs truly excel at transposing unstructured data rapidly – so turning the bank-received spreadsheet into a format that Massdraft can work with is finally a job that takes seconds, feeding the document creation pipeline. What’s more, headless Massdraft combines AI-templating, too – so the task of building the template from the pro-forma is a job that’s rapid, too.

It’s been a long road since 2019, when this problem was posed to us – we finally think we’ve cracked it.