The document automation questionnaire, sometimes still called an interview (a term HotDocs coined decades ago), is dying. That’s the claim doing the rounds, and it’s not entirely wrong, it’s just aimed at the wrong target. It isn’t the questionnaire that’s dying, it’s the hours lawyers and paralegals lose retyping data that already exists somewhere else. The questionnaire itself isn’t going anywhere; it will still be doing the heavy lifting of data capture for years to come. What’s changed is who, or what, is doing the typing, and that shift is the biggest productivity gain this technology has delivered in years.
So what am I talking about?
Using any of the LLMs, it is now possible to drag and drop unstructured documents, such as PDFs, emails, Word documents, data files, or even hand-written notes, into a questionnaire, and AI will automatically insert the data (answers) into the appropriate fields.
For complex practice areas that require a lot of information to be captured, this functionality is a great time saver. And it is remarkably accurate too. But having said that, it is AI, after all, so it needs to be checked. And yes, of course, that’s not ideal, right?
That’s why each answer in the questionnaire is supported by its proof of source. By clicking on the field, the source document is displayed, highlighting exactly where the answer text was found. That makes it much easier to verify answers in the questionnaire. Each answer also has a confidence score, so where there is a conflict in data between uploaded documents, or where AI feels an answer isn’t accurate, the field will display an alert icon.
So this is unlike loading a known dataset or answer file, where variables in the data match variable names in the questionnaire. AI does a “smart match”, so input data can be completely unstructured. That presents a number of new opportunities.
Autofill saves the most time wherever data already exists somewhere else and would otherwise have to be retyped. Migrating from a legacy document automation system is one example, where existing client and matter data can be carried straight across into a new one. Client instructions received by email or PDF are another, where the relevant details can be uploaded directly into the questionnaire rather than transcribed by hand. Taking over a matter from another firm works the same way, drawing on whatever file material is available rather than starting the questionnaire from scratch.
The same principle extends to renewals, where a lease, policy, or services agreement can be updated from last year’s document instead of rebuilt from nothing. In transactions involving a bundle of related documents, such as a conveyance or loan settlement, autofill can populate the whole set consistently from one pool of matter data. Due diligence materials, such as company extracts or share registers, can be read directly into corporate documents, and related documents for the same client, such as a will, power of attorney, and guardianship appointment, can all draw from a single set of client data. Even client onboarding benefits, with details lifted straight from identification documents or engagement letters into the questionnaire.
There is also a quieter benefit here, beyond speed. Every time data is retyped by hand, there is a chance it gets retyped wrong: a transposed figure, a misspelt name, an address copied from the wrong line. Autofill removes most of that risk, because the data moves straight from the source document into the field, with the source kept on hand for anyone who wants to check it.
It also changes what the person completing the questionnaire spends their time on. Instead of hunting through a PDF for a policy number, or retyping a client’s details for the third time that week, they spend their time checking and confirming. That’s a better use of a skilled person’s time, and it’s a change firms should welcome rather than resist.
One concern is that this would be a complex process, which would take some learning before being usable. But that’s the beauty of this functionality. Any source document, large or small, in any format, can simply be dragged and dropped onto the questionnaire. No learning required at all.
So is this the greatest improvement in return on investment that document automation technology has ever had? That’s hard to say. Integrations, where structured data is fed straight into documents, would probably be the other great improvement in ROI. And with the new autofill functionality, users can have the benefit of both. For example, a user might key in the matter number, and the available data from a practice management system would be inserted into the appropriate fields in the questionnaire. Next, the user could drag and drop any other supporting documents and instruct autofill to only fill empty fields. The autofilled fields show a document indicator, whereas the structured data fields have none.
So, is autofill the end of the interview? The answer to that is “no”. But will autofill save firms time and reduce the cost of data capture? Without any doubt.
Chris Pearson
XpressDox
