Does your experience with Atrium suggest that this kind of functionality is feasible? Even developing a 'law as code' platform where NDAs can be written by computer experts in a DSL, formally analysed and compiled to English seems like a major engineering challenge, let alone building a system that can accept an unstructured PDF and parse out its legal implications. That's a far cry from the 'busy work' (filling out templates, making sure the right people have signed them, and keeping copies) discussed in this article.
At this point in our relationship, Atrium's basically been a really well-run law firm who is economically incentivized to make legal more efficient. They also have significant funding, an experienced CEO, and silicon valley thinking, so yeah I think the foundation for something significant like "law as code" is there, even if it's years away.
That being said, I agree the first releases are underwhelming but "If you're not embarrassed by the first version of your product, you've launched too late." [1]
Legal English documents are already a "DSL" to non-experts.
With recent advances on Neural Network models such as GPT-2, we (engineers, researchers, etc) are closer to be able to analyse any documents written in correct English. It won't be fully autonomous at first, but we can imagine it will help to answer questions such as "given my policies database, do I break the rules defined in that contract?". The tool will give hints, which will be reviewed by an actual lawyer to say the final word.
