You are currently viewing Document Automation #1: What is (and is not) Document Automation?
Document Automation Specialists

Document Automation #1: What is (and is not) Document Automation?

Implementing Document Automation Series #1: What is (and is not) Document Automation?

What is Document Automation?

Document automation is a process which enables the end-user to create first draft documents by answering a set of questions. The exact process for automating a document will vary with different software packages, but is generally speaking as follows:

A precedent (usually a word document template) has a set of questions and guidance notes added into it. Rules-based logic is applied to the relevant parts of the document, which is published for users as an online questionnaire.  The end-user answers a series of dynamic questions and the software generates a first draft document, or set of documents, which are populated based on how the questions are answered. The questions asked can alter depending on the answers to previous questions, which can also vary the number and type of documents produced.

Since writing this definition, I have implemented systems which produce data-driven contracts at the push of a button i.e. there is no need for a questionnaire. It is also possible to produce finalised documents in various formats and to push these into other systems, but let’s take this one step at a time – many legal documents are negotiated and what we are talking about here is enabling lawyers/legal teams as opposed to replacing them, although this is often a contentious topic of debate (comments below welcome!).

What is the difference between Artificial Intelligence (AI) and Document Automation?

When I first wrote this article back in 2018, I boldly stated: “Document automation is not (and does not as of writing this contain any) Artificial Intelligence (AI)”. The truth is that “AI” as a term is very vague and its definition over time seems to change… it is usually banded around to explain things that people don’t yet understand. Once something is understood, it then becomes #boring and is boiled down to being “just” a simple formula, algorithm, or piece of software. Some will argue that expert systems (like document automation) are AI but I think when the majority of people hear the term AI, they immediately think of machine learning. This is my key issue with AI – such a broad term that has received so much hype over the last few years – it ultimately leads to disappointment when the system in question doesn’t pull a rabbit out of a hat and do your ironing at the same time.  So to be clear – a document automation system will only do exactly what you tell it to do. Thankfully it has no machine learning element i.e. the system is not able to teach itself, and will, therefore, make no mistakes. A Machine Learning system’s accuracy could vary from say 50-95% depending on how many lawyers and example documents are used to train it, whereas a document automation system will always be 100% accurate (or 100% inaccurate!). This lack of Machine Learning and therefore 100% certainty of output addresses some of the ‘black box’ fears that users often have in regards to trusting the system (preview functions address the others). It highlights the importance of lawyer input and testing during the automation process. If a mistake is coded into the system (and not spotted), that mistake will be repeated every time it is used. As a general rule, AI/ML is being used in the legal profession mainly for the review of existing documents, whereas document automation is concerned with the initial drafting of new documents.

This was the first in a series of posts based on a pre-publication version of an article I wrote for Legal Information Management, (18 (2018, pp 93-97)) the Journal of the British & Irish Association of Law Libriarians (BIALL). The original article was titled “Implementing Document Automation: Benefits and Considerations for the Knowledge Professional”. My next post will focus on a discussion of the benefits of document automation for KM. I hope you enjoy reading this series – please feel free to provide any comments below and to sign up to LIM for the full original article: https://www.cambridge.org/core/journals/legal-information-management/latest-issue/.