Machine Aided Patent Drafting: A Second Look

Back in June, I wrote about a service that augments patent drafting by allowing a practitioner to submit claims, and receive back the rest of a draft specification.  That post is here.  I pointed out then that the terms of service revealed that the business, Specifio, had seemingly thought through a lot of the ethical issues that I could see.

A month later, I posted here about how faster speed-to-market, and shorter product life cycles, mean that for a greater number of products, patents make less sense: having a right to exclude spring into existence after the product has come and gone isn’t very valuable.  Of course, there are other reasons to have patents, and a patent may still be useful even if technology has marched on, to prevent others from following that lead.  But, using Track 1 and other means to speed up issuance may be needed more often. (I have an article coming out soon with a billion footnotes, and I’ll let you know about it.)

Since then, I’ve been researching the ethical issues arising from artificial intelligence, machine learning, and related developments.  These things are here, now.  Early adopters obviously have to be careful, but late adopters may be left in the dust.

In light of all of the above, if you take into account the increasing need for speedier issuance (and the need to file first under the AIA), the need for speedier drafting is obvious.  The capacity of AI to satisfy that need is here, and its role will increase. (I’ve read about memo drafting services that area already in operation, for example. That’s coming on fast, too.)

In light of all of this, I had a conversation with one of Specifio’s founders and was given the opportunity to see how it worked. (They offer a free trial, too.)  So, I took a claim from a published application and submitted it by email to their system. The actual patent application is here, and the first draft back from Specifio is here.

Take a look.  And I missed the fact that the draft spec did include drawings, here.

As for me, I was surprised and impressed.  Now remember, the draft spec was drafted by automation entirely off of the single claim I submitted — nothing else — and that process took about 5 minutes of my time (less if I had formatted the claim the way it told me to.)

5 minutes of lawyer time for a first draft.

Second, obviously the specs are different.  One key reason is that I only submitted Claim 1, so there’s obviously more in the actual spec than the one from Specifio.    I would hope that the initial draft of the actual patent was also significantly different from what you see in the filed version. I also know that no one would take a draft spec from a young lawyer and file it.  But take a look.  (The fact that a computer would draft differently is an interesting thing, and I’ll return to that below.)

Second, on top of the speed discussed above, automation like this has the potential to free practitioners from the time of preparing at least the initial draft of the spec.  (The boring part.)  This has obvious benefits to lawyers and clients: fees can be reduced, time can be better leveraged, and lawyers can do more interesting work.  In that regard, Specifio’s lawyers use the system to draft its own specs.  They report that rather than taking 15 to 20 hours to write a case, so two full days of attorney time, a lawyer can draft five cases in a day.

Think about that for a minute.

There are some other things.  The system will obviously write specs the way it “likes” to.  It will have its own approach.  That is good, and bad. Things will need to be “fixed,” no doubt.  Substantive errors will need to be spotted.  And, idiosyncratic things — suppose you don’t like “wherein” but like “in which” — will need to be changed, too.  But the consistency will allow lawyers to create macros to quickly deal with such things.

I’m curious what you all think.

About David

Professor of Law, Mercer University School of Law. Formerly Of Counsel, Taylor English Duma, LLP and in 2012-13, judicial clerk to Chief Judge Rader.

8 thoughts on “Machine Aided Patent Drafting: A Second Look

  1. 2

    This example made me think about what useful information should be included in the summary or the detailed description that is not already in the claims. Hard questions…

    1. 2.1

      Too “hard” for me PiKa, your last sentence. Seriously, I don’t know whether you are being sarcastic, or whether you really do find it hard to decide what exactly to include in the written description, additional to the claim text.

  2. 1

    Interesting. I draft medical device cases. I start with sketches, then start thinking about the claims. Then I draft claims. That is what takes the time.

    Once I am satisfied with a full set of claims, I dictate the specification. It doesn’t take long. In so doing, I sometimes identify marginal improvements to the claims. David have you ever heard the saying “Claim the novelty but describe the non-obviousness” Can the machine do that yet?

    I doubt that the machine can replace me yet, as the claim drafter. Handing over to the machine to write the specification saves me little time, deprives me of my opportunity to review my work and results in a specification that is more or less ineffective in every jurisdiction except for the USA.

    Incidentally, David, what drawings did the computer generate. I don’t see any in your write-up.

    1. 1.1

      I work with a lot of medical devices and automotive inventions as well. This can’t write specs for mechanical devices, only simple software inventions. It starts with a method claim and creates a “system” claim and drawing that is boilerplate-ish with each method step being converted into a basic “module” in memory.

      The hardest patent to write is a simple mechanical invention. Maybe in 10-20 years something will come al

      1. 1.1.1

        I don’t think it will be 10-20 years. Machines learn fast. Everything is going to change faster than we can imagine. (Recall that in 2007 there was no iPhone…. 2 years from now, things will be just as different as they are today from 10 years ago is the expectation I have.)


          I hope so. The most time consuming, and least rewarding part of drafting an app is the specification and drawings. Using AI to draft the basic information about an invention and they allowing the attorney to tailor the draft to highlight the advantages and improvements would be unbelievably useful.

Comments are closed.