Biglaw is dying Forum

[Desert Fox]

I) transactional isn't legal work

II) doc review isn't legal work

III) seems like most biglawyers aren't doing repetitive tasks that are easily automated. Sure they doing god awful boring and needlessly detailed. But computers still have zero ability to comprehend anything. It's all just correlating facts in a database.

[RaceJudicata]

This is great news for everyone that isn't a lawyer. Anything that drives the cost of legal services down is a win for the economy.

Pretty sure legal costs will stay high because it's a necessary good, it's just there will be fewer lawyers.

I also don't see how litigation can be automated (without sci-fi AI) because it's all based on subjective arguments anyway. You could probably automate settlement negotiation (wouldn't be surprised if that's already done), but for cases where the law isn't clear or it requires some shit like "what would an individual having ordinary skill in the art find obvious" you'd still need lawyers.

I mean, if you can automate legal brief writing and reasoning, we've reached the point where robots are our overlords.

Yeah, transactional makes some sense. I can't even comprehend how litigation could become fully automated.

[zot1]

Plot twist: the robots get bored doing biglaw work too and they revolt by going in-house.

[star fox]

This is great news for everyone that isn't a lawyer. Anything that drives the cost of legal services down is a win for the economy.

Pretty sure legal costs will stay high because it's a necessary good, it's just there will be fewer lawyers.

I also don't see how litigation can be automated (without sci-fi AI) because it's all based on subjective arguments anyway. You could probably automate settlement negotiation (wouldn't be surprised if that's already done), but for cases where the law isn't clear or it requires some shit like "what would an individual having ordinary skill in the art find obvious" you'd still need lawyers.

I mean, if you can automate legal brief writing and reasoning, we've reached the point where robots are our overlords.

Yeah, transactional makes some sense. I can't even comprehend how litigation could become fully automated.

Dunno how you automate a special indemnity for a particular contingent liability, but maybe.

[jrf12886]

I'm not sure how automation will affect litigation outside the doc review context. I don't see computers writing briefs, writing letters to opposing counsel/the court, taking depositions, or attending conferences/hearings. They may become more involved in legal research, however.

And even with respect to doc review, there will always have to be a human involved, at least to do (1) QA of responsiveness calls (no client will just let a computer decide what to produce and blindly press the send key),(2) privilege determinations, and (3) reviewing documents produced by the other side to build your case/defense and prepare for depos.

[Pokemon]

When i was a first year, I thought it was a matter of time to automate transactional work. As a third year, there just seem to be too many pieces that are based on imperfect, not easily categorized information to fully automate stuff.

[Glasseyes]

I'm not sure how automation will affect litigation outside the doc review context. I don't see computers writing briefs, writing letters to opposing counsel/the court, taking depositions, or attending conferences/hearings. They may become more involved in legal research, however.

And even with respect to doc review, there will always have to be a human involved, at least to do (1) QA of responsiveness calls (no client will just let a computer decide what to produce and blindly press the send key),(2) privilege determinations, and (3) reviewing documents produced by the other side to build your case/defense and prepare for depos.

Academics actually suggest we will see computers drafting briefs and memos in the next decade, and the professions are at almost as much risk as blue collar jobs (which are already gone and never coming back, but no one yet told the workers [or the president]). AI is going to reach a point where it can do virtually anything better than humans, at which point you won't need any of us around. When that happens is anyone's guess, but judging by the rate at which machines can learn, I'd peg it around 10-15 years. Bottoms up.

For further depressing reading, this was published just today:
https://www.wired.com/2017/02/ai-threat ... dle-class/

And this is probably the most thorough scholarly work to date:
https://www.amazon.com/Future-Professio ... rofessions

[Desert Fox]

We could automate e discovery even more by just doing discovery. It's unconstimatutional anyway.

[TLSModBot]

Things that can be automated in whole or part, or at least made drastically simpler so as to net reduce billable hours:
1. Doc Review
2. Most aspects of diligence related to reading and parsing contracts
3. A ton of finance document updating/drafting/organization/sig page bullshit
4. Legal research

There is so much more to law than this, but note that the above is largely junior associate work. We might see more streamlined firms with smaller starting classes, which might as well be a fucking recession as far as law students are concerned.

[TLSModBot]

I'm not sure how automation will affect litigation outside the doc review context. I don't see computers writing briefs, writing letters to opposing counsel/the court, taking depositions, or attending conferences/hearings. They may become more involved in legal research, however.

And even with respect to doc review, there will always have to be a human involved, at least to do (1) QA of responsiveness calls (no client will just let a computer decide what to produce and blindly press the send key),(2) privilege determinations, and (3) reviewing documents produced by the other side to build your case/defense and prepare for depos.

Academics actually suggest we will see computers drafting briefs and memos in the next decade, and the professions are at almost as much risk as blue collar jobs (which are already gone and never coming back, but no one yet told the workers [or the president]). AI is going to reach a point where it can do virtually anything better than humans, at which point you won't need any of us around. When that happens is anyone's guess, but judging by the rate at which machines can learn, I'd peg it around 10-15 years. Bottoms up.

For further depressing reading, this was published just today:
https://www.wired.com/2017/02/ai-threat ... dle-class/

And this is probably the most thorough scholarly work to date:
https://www.amazon.com/Future-Professio ... rofessions

Susskind is a fun read but dude never practiced a day in his life. He's been literally saying "AUTOMATION WILL RULE EVERYTHING IN LAW" for 30 years. I think he makes some good points, but like most academics in this field, not in touch with client realities that will delay this kind of automation significantly.

[foregetaboutdre]

Applying machine learning and expert systems seems to be able to automate tasks such as due diligence and easy statutes.

Current machine learning struggles immensely to predict came outcomes right now (with a decent confidence ratio).

One thing that is helpful is Ravel Laws judicial analytics which gives what cases and arguments particular judges often cite.

It's just too hard right now to be able to codify a machine learning alogorithm or expert system with a particular set of facts in a legal dispute with many claims and have it spit out an answer. In the past, computing power and storage may have been an issue. Now it seems like it's too hard to assign weights or determine which variables are useful to build something with a decent confident ratio.

[los blancos]

I'm still fighting with partners and clients to use 6 year old predictive coding tech.

Yeah, this.

Also, doc review is probably already like 50% of the way there.

Y'all are saying that legal research will be automated but it's also already a good chunk of the way there. I'm sure there'll be better search functions but what else can computers really do?

Might be wishful thinking but I don't think computers can replaced not-junior litigators.

[Desert Fox]

What percentage of lit billables are legal research or mass scale doc review? 15%?

I also don't really believe legal research will be totally automated anytime soon. It'll get a lot better but you are still going to have to read cases. that start up has been shilling legal Watson for years now and it's gotten nowhere. I tried signing up for a demo and they never got back to me. There haven't been any public demonstrations. It's probably just a shittier version of westnext

Siri can't even answer 30% of my questions. Lol at law Siri knowing whether the claims of a prior art patent are prior art.

[favabeansoup]

No way is ai or machine learning good enough to replace actual legal work.

Legal work will be some of the last work to disappear.

My firm started using an AI program for some basic due diligence stuff this year. It sucks. Like, really sucks. The few times people have used it it has spit out data in terrible formats that people needed to spend hours to make presentable and readable.

I get technology will get better. I get we may be using unoptimized programs.

I just don't see anything substantially changing for at least 5-10 years. And even then only for low level due diligence/doc review stuff.

[dixiecupdrinking]

Yeah I honestly don't see how you could even trust a computer to do doc review on anything remotely important. I barely trust first years and they are infinitely more capable than some algorithm of making subjective judgment calls on what matters. Fact of the matter is outside some incredibly narrow areas, programing a computer to do a lawyer's job is way harder than just doing the job.

[jbagelboy]

Plot twist: the robots get bored doing biglaw work too and they revolt by going in-house.

I lol'd

[jbagelboy]

Automation isn't the primary threat to american biglaw associates; it's the chennai and jaipur mass contract staff attorneys.

[TLSModBot]

Automation isn't the primary threat to american biglaw associates; it's the chennai and jaipur mass contract staff attorneys.

Nah. Industry studies are showing there's a very limited cost savings by exporting review work abroad. I doubt it's a model anyone is going to expand on into other areas.

Plus the attraction of automation is verifiable statistical verification of accuracy that is infinitely better than human reviewers, much less the cheapest contract attorneys abroad.

[Pizzaburger]

Pretty interesting read: http://www.mckinsey.com/business-functi ... y-cant-yet

[Anonymous User]

Things that can be automated in whole or part, or at least made drastically simpler so as to net reduce billable hours:
1. Doc Review
2. Most aspects of diligence related to reading and parsing contracts
3. A ton of finance document updating/drafting/organization/sig page bullshit
4. Legal research

There is so much more to law than this, but note that the above is largely junior associate work. We might see more streamlined firms with smaller starting classes, which might as well be a fucking recession as far as law students are concerned.

Legal research is one of the most complex aspects of law,
and no it will not be easy to automate. You can teach a computer to find a case that states the elements of unjust enrichment, but a lot of research is more complicated than that (and often it is by doing the research that you develop new arguments).

[los blancos]

^And like DF said, at the end of the day there's still going to need to be someone to read the cases themselves.

What percentage of lit billables are legal research or mass scale doc review? 15%?


Siri can't even answer 30% of my questions. Lol at law Siri knowing whether the claims of a prior art patent are prior art.

Yeah we don't spend that much time on legal research. Doc review probably higher than that but probably has been trending down for a while already.

I mean if you go by what the Fed Cir does, District Judge + her clerk often don't even "know" that so yeah.

[TLSModBot]

Things that can be automated in whole or part, or at least made drastically simpler so as to net reduce billable hours:
1. Doc Review
2. Most aspects of diligence related to reading and parsing contracts
3. A ton of finance document updating/drafting/organization/sig page bullshit
4. Legal research

There is so much more to law than this, but note that the above is largely junior associate work. We might see more streamlined firms with smaller starting classes, which might as well be a fucking recession as far as law students are concerned.

Legal research is one of the most complex aspects of law,
and no it will not be easy to automate. You can teach a computer to find a case that states the elements of unjust enrichment, but a lot of research is more complicated than that (and often it is by doing the research that you develop new arguments).

Think bigger picture

Once upon a time research was done in physical books and took ages. As Westlaw and Lexi's came along we got keyword searches, shephardizing, and other categorization features. Now we are seeing some really audacious conceptual linking that can A. help you find cases faster and B. start incorporating compiled analytics like certain argument success rates.

Sure associates will "do the research" but when something that took 10 hours now takes 2, you've massively slashed the market. And if you think clients will pay for endless research in most cases then you're either still a law student or insulated at a higher end firm doing bet-the-company Lit (which, surprise, is not representative of the market)

[smaug]

Even outside bet-the-company lit, you're sometimes (more often than I expected when I started at least) going to run into truly novel issues and issues where there are only a handful of cases on point. At that point, you need to get creative and start analogizing to existing but not identical or clearly applicable bodies of law (e.g. stealing a solution from bankruptcy courts even if your case isn't in bankruptcy; recognizing odd deeper-level symmetries between cases).

I seriously doubt that machine learning will be able to get "creative" as is necessary when you move past basic legal research and start developing actual persuasive arguments for a filing.

[obx]

Things that can be automated in whole or part, or at least made drastically simpler so as to net reduce billable hours:
1. Doc Review
2. Most aspects of diligence related to reading and parsing contracts
3. A ton of finance document updating/drafting/organization/sig page bullshit
4. Legal research

There is so much more to law than this, but note that the above is largely junior associate work. We might see more streamlined firms with smaller starting classes, which might as well be a fucking recession as far as law students are concerned.

Legal research is one of the most complex aspects of law,
and no it will not be easy to automate. You can teach a computer to find a case that states the elements of unjust enrichment, but a lot of research is more complicated than that (and often it is by doing the research that you develop new arguments).

Think bigger picture

Once upon a time research was done in physical books and took ages. As Westlaw and Lexi's came along we got keyword searches, shephardizing, and other categorization features. Now we are seeing some really audacious conceptual linking that can A. help you find cases faster and B. start incorporating compiled analytics like certain argument success rates.

Sure associates will "do the research" but when something that took 10 hours now takes 2, you've massively slashed the market. And if you think clients will pay for endless research in most cases then you're either still a law student or insulated at a higher end firm doing bet-the-company Lit (which, surprise, is not representative of the market)

I agree with the general argument (what took 10 hours will eventually take 2) but I don't think I understand the conclusion. Computerized westlaw also slashed the amount of time it took to do legal research, but firms still hire 100s of attorneys and make money hand over fist.

[smaug]

Things that can be automated in whole or part, or at least made drastically simpler so as to net reduce billable hours:
1. Doc Review
2. Most aspects of diligence related to reading and parsing contracts
3. A ton of finance document updating/drafting/organization/sig page bullshit
4. Legal research

There is so much more to law than this, but note that the above is largely junior associate work. We might see more streamlined firms with smaller starting classes, which might as well be a fucking recession as far as law students are concerned.

Legal research is one of the most complex aspects of law,
and no it will not be easy to automate. You can teach a computer to find a case that states the elements of unjust enrichment, but a lot of research is more complicated than that (and often it is by doing the research that you develop new arguments).

Think bigger picture

Once upon a time research was done in physical books and took ages. As Westlaw and Lexi's came along we got keyword searches, shephardizing, and other categorization features. Now we are seeing some really audacious conceptual linking that can A. help you find cases faster and B. start incorporating compiled analytics like certain argument success rates.

Sure associates will "do the research" but when something that took 10 hours now takes 2, you've massively slashed the market. And if you think clients will pay for endless research in most cases then you're either still a law student or insulated at a higher end firm doing bet-the-company Lit (which, surprise, is not representative of the market)

I agree with the general argument (what took 10 hours will eventually take 2) but I don't think I understand the conclusion. Computerized westlaw also slashed the amount of time it took to do legal research, but firms still hire 100s of attorneys and make money hand over fist.

electronic research (time saving) came at the same time as electronic discovery (gigantic time suck)