Harvey AI Speaks!
Most Candid Interview - No Priors
February 14 dropped a new “No Priors” Podcast (Elad Gill and Sarah Guo) with Winston Weinberg CEO of Harvey.
This has to be the most candid press in the history of the nortoriously shy Harvey — so was keen to listen in.
For your enjoyment and information, here is the Podcast, Ai Counsel’s Summary following, and the Transcript.
Ai Counsel Summary
Discusses the company's journey in developing an AI platform for legal and professional services. Founded in 2022, inspired by successful experiments with GPT-3 on legal workflows. Raised over $500 million and more than 250 clients. Building trust is crucial - through partnerships with prestigious firms and data providers
Product strategy involves a platform combining broad productivity tools with streamlined workflows. Focus on building reusable "AI patterns" for complex operations, while simplifying the user interface to enhance accessibility. Allows Harvey to address general productivity needs and specialized tasks with high-quality outputs
Team building is key, with senior lawyers as design partners and evaluators. Emphasizes hiring for adaptability and agency, as the AI landscape evolves. Company fosters a culture of innovation and resilience, leaders who take ownership and commit to continuous improvement
Predicts AI will drive significant changes in legal and professional, shifting toward fixed-fee models while billable hours for advisory work. New models expected from collaborations law firms and technology. Expanding into tax / audit.
Transcript
# No Priors Ep. 101 | With Harvey CEO and Co-Founder Winston Weinberg
00:00:03.070 [Applause]
00:00:05.879 hi listeners and welcome back to no
00:00:07.439 priors today I'm here with Winston
00:00:09.400 Weinberg the co-founder and CEO of
00:00:11.679 Harvey which is building domain specific
00:00:14.280 AI for law Professional Services and the
00:00:17.279 Fortune 500 they've now raised more than
00:00:19.920 $500 million from investors such as open
00:00:23.080 aai seoa finer Perkins GV lill and me
00:00:27.720 we're going to talk about how to do end
00:00:30.400 to-end workflows how to serve
00:00:32.238 conservative users impostor syndrome
00:00:35.120 keeping Pace with the blitz of the AI
00:00:37.559 Eco system and also what lawyers will do
00:00:40.320 5 years from now Winston thanks for
00:00:42.399 doing this yeah of course it has been
00:00:44.440 like a wild two and a half years for you
00:00:46.559 and Gabe and Harvey um when you started
00:00:49.559 the company in August of 2022 or at
00:00:52.320 least when I met you guys for the seed
00:00:53.960 yeah we started a little bit earlier but
00:00:55.160 about then yeah uh what was the moment
00:00:57.239 of inspiration yeah so G and I actually
00:01:00.800 had met a couple years before and I
00:01:03.640 definitely didn't know anything about
00:01:04.680 the startup world and didn't have a plan
00:01:06.479 of of doing a startup and what had
00:01:09.000 happened was he showed me gbd3 which at
00:01:11.840 the time was you know public and and I
00:01:14.119 was first of all just incredibly
00:01:15.960 surprised that no one was talking about
00:01:17.400 gbd3 and no one was using it in any way
00:01:19.400 shape or form um and he showed me that
00:01:22.840 and I showed him kind of my legal
00:01:25.000 workflows and we started the the kind of
00:01:28.000 aha moment was we we went on uh r/ legal
00:01:31.880 advice which is basically you know a
00:01:34.079 subreddit where people ask a bunch of
00:01:36.360 legal questions and almost every single
00:01:38.439 answer is so who do I sue um almost
00:01:40.840 every single time and we took about a
00:01:43.119 hundred landlord tenant questions and we
00:01:46.320 came up with kind of some Chain of
00:01:47.840 Thought prompts and this is before you
00:01:49.640 know anyone was talking about Chain of
00:01:50.920 Thought or anything like that and we
00:01:53.119 applied it to those landlord tenant
00:01:54.840 questions and we gave it to three
00:01:56.399 landlord tenant attorneys and we just
00:01:58.360 said nothing about AI we just said here
00:02:00.799 is a question that a potential client
00:02:02.680 asked and here is an answer uh would you
00:02:05.520 send this answer without any edits to
00:02:07.680 that client would you be fine with that
00:02:09.080 you know is that ethical is it a a good
00:02:11.000 enough um answer to to send and 86 out
00:02:14.560 of 100 was yes um and actually we cold
00:02:18.040 emailed the general Council of open aai
00:02:20.480 and we sent him these results and his
00:02:22.840 response basically was oh I had no idea
00:02:24.879 the models were this good at legal um
00:02:27.200 and we we met with the the SE Suite of
00:02:28.879 open AI a couple weeks after and and the
00:02:31.640 view was just is going to be good enough
00:02:34.360 um we should build a company around it
00:02:36.360 yeah for what domains I mean I think
00:02:38.360 what happened or the reason we were so
00:02:41.000 confident about this was the models and
00:02:44.159 even with gbd3 you could get it to do a
00:02:47.159 lot of tasks you just had to really
00:02:49.360 brute force it right like you had to
00:02:51.080 Brute Force the amount of context
00:02:52.920 telling it which steps to take Etc and
00:02:55.560 the idea was over time this is just
00:02:57.560 going to get better right they are going
00:02:59.159 to be either the models themselves are
00:03:01.159 going to get better or we're going to be
00:03:02.800 we're going to get better over time at
00:03:04.799 figuring out how to provide them the
00:03:06.120 correct context how how to improve them
00:03:08.400 how to evaluate the results Etc um and
00:03:11.560 even just by playing with it for a
00:03:13.319 decent amount of time you could get that
00:03:15.159 sense you are obviously not focused just
00:03:18.040 on property law now like how do you
00:03:20.080 think about the mission or scope of
00:03:21.879 Harvey today yeah um so mostly we're
00:03:25.720 we're de developing it for legal overall
00:03:28.319 but I would say that what we're building
00:03:30.879 is the AI platform for legal and
00:03:32.640 Professional Services right and if that
00:03:35.040 sounds vague or it sounds like there
00:03:37.040 aren't incredibly defined use cases for
00:03:40.000 you know the small areas that we're
00:03:41.400 building that's on purpose like the
00:03:43.280 reality is if you are using these tools
00:03:46.560 and you don't think that you can take
00:03:48.640 basically Ai and apply to X industry and
00:03:51.519 transform the entire industry I don't
00:03:54.040 think you're thinking ambitiously enough
00:03:56.319 right and I think it's really hard
00:03:58.280 because the you know these models can't
00:03:59.519 just onot all of these really complex
00:04:01.879 legal tasks or or in these other domains
00:04:04.040 like tax and Prof other Professional
00:04:06.360 Services and so what you have to do is
00:04:08.599 you have to build a platform that is
00:04:10.680 kind of constantly expanding and
00:04:12.480 constantly collapsing and so what I mean
00:04:14.400 by that is you need to build specific
00:04:17.079 features and maybe agentic workflows Etc
00:04:20.000 that can do parts of a task and then you
00:04:22.160 need to combine them all together so the
00:04:24.040 UI is simple and you don't have this
00:04:25.800 like tentacle monster of a platform when
00:04:28.360 did you realize that that was the like I
00:04:30.080 think it's a really elegant framing of
00:04:32.199 sort of product strategy for the company
00:04:34.199 yeah um uh first implied in it is just
00:04:37.120 like this you know and this has been
00:04:38.639 true for you guys from the very
00:04:39.680 beginning I I did think you were a
00:04:40.759 little bit crazy when you described the
00:04:42.360 level of sophistication of tasks end to
00:04:45.120 end that Harvey would be able to I still
00:04:46.680 feel crazy uh but you appear to be more
00:04:49.919 you know like you don't look as crazy
00:04:51.160 when you're right fair right um but I
00:04:53.680 remember like a year and a half two
00:04:55.240 years ago um uh like it was very
00:04:59.240 distinct to me how much you guys
00:05:00.759 believed in capability Improvement yeah
00:05:02.960 like where does that come from and then
00:05:04.199 also how do you think about it
00:05:05.320 internally in the company yeah so where
00:05:07.800 does it come from using the tools I mean
00:05:10.560 I think I have been blown away by the
00:05:13.440 amount of kind of seite and Etc that
00:05:16.600 haven't actually like spent that much
00:05:18.240 time using Ai and the the big moment for
00:05:21.720 me was just the jump between gbd3 and
00:05:24.319 gbd4 like when we got access to gbd4 I
00:05:27.880 went into my room I think for 24 hours
00:05:30.600 straight and try to do every single
00:05:32.800 thing that I couldn't do with gbd3 or I
00:05:35.400 had to very much brute force and like
00:05:37.360 wasn't getting any efficiency gains from
00:05:39.600 and try to do it with with gbd4 it
00:05:41.600 didn't do all of them but the
00:05:43.080 Improvement was crazy right and no one
00:05:46.039 was talking about it and I I found that
00:05:47.840 something that also was very weird is I
00:05:50.000 would show gbd3 and and gbd4 to a bunch
00:05:52.560 of my friends and they would try to get
00:05:55.199 it to do one thing it didn't do the
00:05:56.759 thing perfectly and then they just stop
00:05:59.240 right and so it seems like there was
00:06:01.360 this large gap between I log into chat
00:06:04.360 gbt or I just use gbd4 from an API or
00:06:07.000 whatever it is and I try something a
00:06:09.039 couple times versus I'm going to sit
00:06:11.160 there and just hammer on this until it
00:06:13.000 works right um and I think if you do
00:06:15.720 that enough you get the intuition for
00:06:17.639 where things are going and how much
00:06:18.919 better they can get um how do we do that
00:06:20.960 from a product standpoint you can kind
00:06:23.000 of think about if you go back to that
00:06:24.800 expand and collapse as like two general
00:06:27.000 themes you need to build productivity
00:06:29.280 tools and what I mean by this is things
00:06:31.800 that are useful for the highest amount
00:06:33.520 of seats right and then you also need to
00:06:36.000 build things that are streamlined
00:06:37.880 vertical workflow from start to finish
00:06:40.039 right and what you can do is you can
00:06:43.160 take those streamlined vertical
00:06:45.319 workflows and you can chain them
00:06:47.360 together to do more increasingly
00:06:49.520 powerful things right and so that's kind
00:06:52.039 of how we think about it internally at
00:06:53.560 the company is you know there's a bunch
00:06:54.960 of bells and whistles and things to add
00:06:56.560 in terms of like sharing collaboration
00:06:58.639 things like that but you also just can
00:07:01.319 basically build certain specific
00:07:03.199 features that will do something from
00:07:04.560 start to finish like I upload all my
00:07:06.879 documents uh and I Target company's
00:07:09.160 documents and it will tell me in 72
00:07:10.720 countries where I need to file antitrust
00:07:13.160 and what do I need to file and what
00:07:14.360 other information do I need and you can
00:07:16.120 take pieces from that and add it to
00:07:17.960 another part of a project how do you
00:07:21.039 organize that effort at Harvey because
00:07:23.879 it's you know like I think you have a
00:07:25.960 very special Instinct and um commitment
00:07:28.879 to going on the the AI bendor I suppose
00:07:31.160 to like figure out what to have going be
00:07:32.840 a long bendor yeah long you know
00:07:34.360 multi-year Bender but you know one one
00:07:35.879 day at a time um uh to develop intuition
00:07:38.639 for you know what the models are capable
00:07:40.280 of doing with this like expansion
00:07:42.560 workflows manipulation assembly process
00:07:45.680 I imagine that it's really hard to find
00:07:47.840 all of that in a single person that
00:07:50.080 understands the research the engineering
00:07:51.639 the domain the data available like
00:07:53.840 interacting with users so like you know
00:07:56.360 how do you do that across teams yeah so
00:07:58.080 I think we're starting to get better at
00:07:59.840 it um and the thing that we're starting
00:08:01.639 to land on is basically identify these
00:08:05.400 systems as what we're calling AI
00:08:07.280 patterns so these are the 30 50 things
00:08:11.759 that we need to build that will be
00:08:13.639 integrated into all of these different
00:08:15.319 pieces of the product right so I'll give
00:08:17.319 you an example of this if you build an
00:08:19.800 AI system that is really good at case
00:08:21.800 law research that can go into a motion
00:08:24.479 for summary judgment that can go into a
00:08:26.680 billion different types of litigation
00:08:28.479 use cases right it doesn't do that use
00:08:30.800 case from start to finish but you can
00:08:32.399 add that on top and so what we've been
00:08:34.279 doing is we basically have teams that
00:08:36.559 will work on building these are the
00:08:38.000 patterns and then teams that are kind of
00:08:40.120 implementing that across the entire
00:08:41.599 platform and that has been working
00:08:43.559 really well um it is definitely a work
00:08:46.080 in progress I think the other
00:08:47.800 interesting piece of this is we have a
00:08:49.320 lot of lawyers on our staff right um and
00:08:51.800 we have a lot of we're going to do this
00:08:53.080 also in the other verticals that we're
00:08:54.519 going into and the domain experts are
00:08:57.480 very helpful for basically like two main
00:08:59.720 reasons one is as design Partners so
00:09:02.399 they're very good at saying actually you
00:09:04.880 know this is what we need to teach the
00:09:06.519 models how to do this is the
00:09:07.800 step-by-step thinking we need to do this
00:09:09.720 is the output that the user wants Etc
00:09:12.440 and then the second piece which is
00:09:13.519 really hard is eval right and so most
00:09:16.240 benchmarks are completely useless for us
00:09:18.079 right and so we'll get a a model you
00:09:19.839 know someone will give us early access
00:09:20.959 to a model and they'll say it's way
00:09:22.480 better on all these benchmarks and we'll
00:09:24.440 respond what it actually isn't like it's
00:09:26.600 not use as useful for us as as a
00:09:28.440 different checkpoint or something like
00:09:29.800 that um and the reality is you have to
00:09:33.240 hire very good lawyers who can actually
00:09:35.360 evaluate these systems and the same with
00:09:36.959 intact and these other areas and they
00:09:39.200 can't be to Junior because if they were
00:09:40.760 to Junior and they were able to eval it
00:09:43.920 they would be senior right and so I
00:09:46.560 think the intertwining of the domain
00:09:49.480 experts into the actual like product
00:09:51.839 sequence and development and then
00:09:53.560 evaluation at the end state is very hard
00:09:56.560 um and it's something we're constantly
00:09:57.519 working on what is the ENT and task
00:10:00.000 you're most excited about that you think
00:10:01.399 Harvey will be able to do this year yeah
00:10:03.560 filing an S4 um I think is something
00:10:06.240 that is the reason I like that process
00:10:09.279 is it is a combination of external data
00:10:12.120 internal data and a million steps right
00:10:15.240 um and I think of workflows as basically
00:10:18.519 a bunch of agentic systems that need to
00:10:20.440 combine together right and if you think
00:10:22.200 about knowledge work Professional
00:10:23.839 Services legal honestly any swath of
00:10:26.680 that what you were doing is manipulating
00:10:29.440 Things based off of your internal
00:10:31.240 context external context whether that's
00:10:33.839 external data whether that's data from
00:10:36.120 your whoever your customer is your
00:10:37.760 client is ETC and then process of this
00:10:41.240 is how you know this is how you do an
00:10:43.160 lbo this is how you do Selter compliance
00:10:45.800 and then there's another step of that
00:10:46.959 which is this is how you do Selter
00:10:48.399 compliance for this particular private
00:10:50.440 Equity Firm right this is market for
00:10:52.160 this particular clause and the more
00:10:54.959 complicated the workflow is the more you
00:10:56.760 have to combine all of those different
00:10:58.279 elements together
00:11:00.360 when you think about like having these
00:11:02.560 very senior lawyers internally at Harvey
00:11:05.120 doing eval I imagine another piece of
00:11:08.399 what is important there is um like the
00:11:11.519 experience that you had with your
00:11:13.079 friends looking at these models which is
00:11:14.959 like you don't get that many shots on
00:11:16.240 goal yeah right it works or it doesn't
00:11:18.240 yep um the uh legal profession
00:11:21.079 especially like you know um uh more
00:11:23.600 senior more prestigious lawyers um you
00:11:27.120 know billable hour rate is very high
00:11:29.120 yeah and expectation of quality from
00:11:31.680 let's say like an associate or a model
00:11:33.920 is very high like how do you how did you
00:11:36.639 go about building that trust because the
00:11:38.800 like you didn't start at the
00:11:39.959 capabilities you have today yeah I'll
00:11:41.760 start from one kind of like highlevel
00:11:44.040 distinction between two pieces so if you
00:11:45.959 go back to the productivity versus like
00:11:48.360 specific specialized output right on the
00:11:51.600 productivity side the minimal viable
00:11:53.959 quality of that output can be lower
00:11:55.760 because you're selling seats and at the
00:11:57.279 end of the day there are multiple people
00:11:58.720 reviewing it right um and so what you
00:12:00.920 want to do in that state is have just
00:12:02.720 show your work right like that is the
00:12:04.320 most important thing you want to mimic
00:12:06.480 exactly how a senior associate reviews
00:12:09.519 the work of a junior associate so you
00:12:11.040 say this is why I did this this is the
00:12:12.680 information that I pulled and here's an
00:12:14.519 inline citation to the literal sentence
00:12:16.959 that I pulled it from is this correct or
00:12:18.600 not right make it cheap to check and um
00:12:21.040 partially correct is still useful yes
00:12:22.920 exactly and because you know a lot of
00:12:25.160 how these law firms work and and again
00:12:26.720 the big four as well is the junior
00:12:28.519 produces some sort of output the
00:12:30.160 mid-level associate reviews that and
00:12:31.720 then the senior level associate reviews
00:12:33.240 that then the partner reviews that and
00:12:34.360 then it goes to the client right and so
00:12:35.920 you do get a structured pyramid of like
00:12:38.040 a hierarchical review right now let's go
00:12:41.040 to the specialized system the
00:12:43.240 specialized system at the end of the day
00:12:45.399 you're actually trying to just from
00:12:46.880 start to finish produce that output
00:12:48.839 right so the minimal viable quality on
00:12:51.000 that is much higher but it's also easier
00:12:54.240 to get higher because you are trying to
00:12:56.240 build systems that just produce the same
00:12:58.519 output and do the same task from start
00:13:00.760 to finish kind of like a specialist
00:13:02.600 right it's just much more scoped it's
00:13:03.920 much more scoped and evaluation is much
00:13:05.760 easier right like how do you do
00:13:07.079 evaluation on co-pilot or Enterprise gbt
00:13:09.880 Etc that is very difficult it is much
00:13:12.279 easier to do evaluation if you can do
00:13:14.240 evaluation at each step and make sure
00:13:16.279 that recursively that step was done
00:13:18.120 correctly okay so there are different
00:13:19.920 expectations for different pieces of the
00:13:21.680 Harvey product yeah um they're still
00:13:23.959 like conservative customers right so you
00:13:27.120 know one of the things that was
00:13:28.639 surprising to me in like 2023 early 2024
00:13:32.279 was the set of people that you decideed
00:13:34.279 to work with first and that were willing
00:13:36.320 to work with you right ano PWC like high
00:13:39.480 quality brands that were generally
00:13:41.480 larger like there's been a um uh I think
00:13:44.680 for maybe a decade in technology
00:13:46.920 investing a uh expectation that if you
00:13:50.880 start with the midm market you start
00:13:52.120 with your friends and startup like it's
00:13:54.240 just easier the requirements are lower
00:13:56.320 um they're willing to take risk uh why
00:13:58.399 do the opposite here yeah so I think
00:14:01.120 that if you are I guess going back to
00:14:02.720 what I said in the beginning of the
00:14:04.199 ambition should be take X industry apply
00:14:07.199 AI to it um if you are doing that you
00:14:10.399 need to partner with that industry and I
00:14:12.480 think you also need to make sure that
00:14:14.560 your brand and what you're building
00:14:16.800 actually makes sense to what they care
00:14:19.000 about I'll give you an example of this
00:14:20.480 or or just kind of like a reference
00:14:21.680 point anos Sherman is almost hundred
00:14:24.079 years old and they became famous
00:14:26.639 basically advising on the abdication of
00:14:28.639 I think it was King Edward iith from the
00:14:31.120 throne okay good trivia my point is
00:14:34.079 these are incredibly old like have very
00:14:38.040 prestigious and a long line of History
00:14:40.519 right and so in order to work with them
00:14:43.880 you actually have to kind of partner
00:14:45.480 with them and say these are the
00:14:46.440 different workflows that we're going to
00:14:47.519 start building with you like use them as
00:14:49.000 design partners and you also need to
00:14:51.040 make sure that you are kind of like
00:14:52.759 building a brand around that so not only
00:14:55.600 is it just actually working with those
00:14:57.759 High tier firms but you also want to
00:14:59.600 work with the data providers that are
00:15:01.320 really important right so Lexus IS in is
00:15:03.880 involved in this round and we're really
00:15:06.399 partnering with them to also bring a lot
00:15:08.560 of the kind of Goodwill and trust and
00:15:10.399 things that they have brought to the
00:15:11.560 industry and and really good products
00:15:13.399 and combine that with REI as well um and
00:15:16.560 so if you are able to do that if you are
00:15:19.040 able to show that you are working with
00:15:21.320 the industry and get the trust from the
00:15:23.480 folks that are basically seen as the
00:15:25.240 highest you know degree in those
00:15:27.240 Industries it is much easier to work
00:15:30.079 with the rest of the industry and we
00:15:32.199 thought that that was very important you
00:15:33.720 know we we thought for a while that
00:15:36.079 maybe we would do you know a self- serve
00:15:38.199 model um and we would kind of try to do
00:15:40.319 plg and and things like that but we
00:15:42.880 found that no matter what you eventually
00:15:45.120 have to get the trust of the industry
00:15:47.319 and the best way to do that is to
00:15:49.639 actually go after the hardest people
00:15:50.959 first yeah I I think there's um perhaps
00:15:54.079 an opportunity in AI we we'll see if
00:15:56.279 this is true um you're already like
00:15:59.440 continually collapsing and trying to
00:16:01.440 simplify the experience of using Harvey
00:16:04.120 right and this has been like impossible
00:16:06.279 in software to date um one of the
00:16:08.680 reasons people would start like you know
00:16:10.800 SNB or mid-market is they'd be like well
00:16:13.319 like once it becomes an Enterprise
00:16:15.000 product we will never have the
00:16:16.920 Simplicity required to serve other the
00:16:19.000 training just the training cost of like
00:16:20.920 implementing that that technology is
00:16:22.759 going to be impossible no one will buy
00:16:23.839 it yeah yeah exactly um but I I think
00:16:26.920 like because you like you know a bunch
00:16:30.040 of different factors like I'd like to
00:16:31.480 see if you agree but um maybe building
00:16:34.880 ux ends up being cheaper in the future
00:16:37.160 than past and like you're also building
00:16:39.720 a muscle as a company that other um
00:16:41.839 companies are not encouraged to have in
00:16:43.639 terms of like the constant
00:16:45.279 simplification so I mean the thing that
00:16:47.440 the models are incredible at is
00:16:48.560 orchestration right and so if you are
00:16:51.759 thinking about like how to constantly
00:16:53.560 collapse the UI I'll give you an example
00:16:56.079 you can build a bunch of different
00:16:57.399 workflows that deal with Spa's share
00:17:00.079 purchase agreements right it extracts
00:17:02.199 the Reps and warranties from the share
00:17:03.639 purchase agreement it takes them and
00:17:05.000 turns them into a sum a complex summary
00:17:07.160 it does the closing conditions all these
00:17:08.799 different things right so you can build
00:17:10.559 those separately because you know maybe
00:17:12.679 your system can't handle that plus every
00:17:15.160 other horizontal use case at once which
00:17:16.839 they can't um so you build those
00:17:18.839 separately and then you just make it so
00:17:20.400 that whenever a user uploads an Spa
00:17:22.720 harvy says hey would you like to do one
00:17:24.480 of seven things to it right and so you
00:17:26.280 can collapse that back down into a UI
00:17:28.600 that is is actually very simple and if
00:17:30.559 you actually think about how the UI for
00:17:32.880 Professional Services is just email like
00:17:34.799 that's it right and so what you can do
00:17:37.280 is again you can build all these
00:17:38.600 different features in the system but
00:17:40.400 then you have to have a really clean and
00:17:42.160 elegant way to combine it all so the
00:17:44.559 user can actually find those things
00:17:46.480 right you don't want a system that has
00:17:48.400 10,000 workflows and you have to somehow
00:17:50.480 filter through them and figure out what
00:17:51.799 you need right what we have in business
00:17:53.600 software that's exactly what we have
00:17:55.640 yeah and it's awful and it takes forever
00:17:57.400 to learn how to use it I know how to use
00:17:59.840 email yeah it's still hard but um I
00:18:03.440 think it takes a like a particular
00:18:05.400 attitude to play with gpt3 and then as a
00:18:09.640 you know young lawyer be like we should
00:18:11.919 do this versus like be threatened by it
00:18:14.919 um uh you guys now serve you know more
00:18:17.159 than 250 clients more than 50 million in
00:18:19.280 AR people are obviously adopting it I
00:18:21.000 look at the like weekly user charts um
00:18:24.360 what is the reaction you get from users
00:18:26.559 and how do you talk to them about the
00:18:27.840 impact of automation for them it changed
00:18:30.480 massively over time um I think that
00:18:33.799 there was a lot of fear of automation
00:18:36.559 when there was a lot of press but they
00:18:38.039 haven't used it basically um and I think
00:18:40.720 like I I'm more bullish on model
00:18:43.400 capabilities and just capabilities in
00:18:45.720 General on the application side and and
00:18:48.400 everything um so you still think you're
00:18:50.240 crazy but you're going to go in that
00:18:51.320 direction I still think I'm crazy um but
00:18:53.240 I do think that one thing that not just
00:18:56.559 our customers are realizing but I think
00:18:58.200 a lot of people that haven't worked in
00:19:00.000 legal or Professional Services are
00:19:01.440 realizing is these industries are very
00:19:03.799 messy in the sense that all of this is
00:19:06.559 not kind of a simple oh I know how to
00:19:09.000 draft an Spa I'm a lawyer like I'm done
00:19:11.360 or I'm going to grab all of the legal
00:19:13.000 data on Earth and just train on top of
00:19:15.000 it and the model Works somehow that's
00:19:16.960 not how any of these industries work
00:19:19.159 right um and so I think that from the
00:19:21.440 lawyer side once you have used the
00:19:23.960 product you start saying oh wow this is
00:19:26.320 really good and I can see that this is
00:19:28.200 going to get better especially if you're
00:19:30.159 someone who used Harvey a year ago and
00:19:31.919 use it now right but it does seem like
00:19:35.320 no matter what we're going to need a
00:19:36.840 bunch of lawyer collaboration involved
00:19:38.760 in the process right and so they're very
00:19:41.120 happy about that and the junior folks
00:19:42.919 are incredibly happy about this because
00:19:45.320 what ends up happening in the legal
00:19:46.720 industry and other Professional Services
00:19:48.799 is you know you go to a really good
00:19:50.240 school you study a bunch of things that
00:19:51.640 have nothing to do with by the way like
00:19:53.200 the practicing law okay um and you come
00:19:56.280 out and you're in a new kind of vertical
00:19:58.280 and you're figure out how to do private
00:19:59.600 Equity or an lbo or something like that
00:20:01.760 and it's you spend a very long time kind
00:20:04.720 of doing repetitive tasks right so
00:20:06.559 whether that's in you know reviewing
00:20:08.159 documents in Discovery or it's reviewing
00:20:09.799 documents in a data room Etc and you end
00:20:12.080 up not being able to do the Strategic
00:20:13.799 level things until like 10 years into
00:20:15.520 your career you know if you're lucky
00:20:17.240 five um and so I think that you did
00:20:20.240 complain about that grind work you're
00:20:21.760 like I should not be doing this there
00:20:23.080 was a a joke that um I I was presenting
00:20:26.000 to the firm I used to work at that wasel
00:20:27.919 and Myers and this is I think maybe a
00:20:29.559 year ago and one of the partners said
00:20:31.679 you know Winston there are way easier
00:20:33.200 ways to get a dock review than trying to
00:20:35.080 automate dock review which I don't know
00:20:37.360 if that's true um but my point is there
00:20:40.520 are all these lower-end tasks that you
00:20:42.760 end up having to do just because the
00:20:44.880 industry is really complex and legal is
00:20:46.720 getting more and more complex right and
00:20:49.880 these tools allow you to do that stuff
00:20:51.520 faster and so what I think will end up
00:20:54.039 happening is the timeline will compress
00:20:56.200 so you will start being able to actually
00:20:57.600 do the high level strategic work and
00:20:59.320 interact with clients which is what
00:21:00.840 people really want to do earlier on your
00:21:02.919 career and so they're happy about that
00:21:04.640 and so I don't think there is as much
00:21:06.840 displacement fear it is not job
00:21:09.360 displacement is Task displacement and I
00:21:12.080 think that's a super important
00:21:14.000 distinction because getting rid of those
00:21:16.640 tasks does not mean the legal industry
00:21:18.080 falls apart it will evolve my uh younger
00:21:21.400 sister Camila um is an engineer by
00:21:24.480 background and works at an AI and
00:21:26.039 Robotics company she still wants to go
00:21:27.480 to law school about it what advice would
00:21:30.360 you have for somebody who wants to be
00:21:32.200 like a star lawyer five years from now
00:21:34.600 yeah I would actually have given the 10
00:21:36.279 years from now I would have given the
00:21:37.279 same advice before doing the startup
00:21:39.640 actually okay um I would say that now
00:21:42.720 spend even more time on it but the most
00:21:45.440 important thing in legal and
00:21:47.120 Professional Services is how well you
00:21:49.679 can deal with client requests like how
00:21:52.000 much you can navigate and figure out
00:21:53.840 what the client what is actually best
00:21:55.240 for the client and so I would try to get
00:21:58.039 as much hands-on experience even if
00:22:00.279 that's at something that isn't as
00:22:01.840 prestigious it isn't the you know bet
00:22:03.919 the company litigation because you're
00:22:05.480 probably not going to get hands-on
00:22:06.640 experience or the massive hundred
00:22:08.640 billion dollar merger Etc like get as
00:22:10.960 much handon experience because that is
00:22:12.679 the skill that is most important is
00:22:14.679 figuring that out and now I would just
00:22:16.760 say that is the main skill that is going
00:22:18.799 to matter over time but even you know
00:22:21.400 three four years ago I think I would
00:22:22.840 have given the same advice what do you
00:22:24.960 think happens to the structure of law
00:22:26.640 firms that have been built on the back
00:22:28.559 of you know Junior Associates doing
00:22:30.120 these repetitive tests yeah um so I
00:22:33.240 think there's a a bunch of ways this can
00:22:34.840 evolve I think the most likely one is I
00:22:37.279 don't think the billable hour is going
00:22:38.520 to just completely disappear okay I
00:22:40.679 think what is going to happen is a lot
00:22:42.799 of these tasks that are going to get you
00:22:45.480 know AI can automate a lot of that task
00:22:47.600 with a lawyer in the loop those tasks
00:22:50.000 will end up being kind of a fixed fee
00:22:51.720 model and I think the high level
00:22:54.039 advisory work on top will still be
00:22:56.520 billable hour and will be maybe more
00:22:59.320 expensive like there's an argument that
00:23:01.520 the specialist at a law firm who has
00:23:03.440 seen all of these different mergers in
00:23:04.880 the pharmaceutical industry their rate
00:23:06.919 should our should not be actually 3x the
00:23:09.559 junior associate in the data room maybe
00:23:11.840 10x I I don't know my point is there is
00:23:15.000 a specialization in Professional
00:23:16.919 Services that is incredibly valuable and
00:23:18.760 is going to be more valuable over time
00:23:21.279 and so I think it'll be a mix of those
00:23:23.120 two and we're also already seeing this
00:23:25.520 we're seeing law firms willing to do
00:23:28.080 things like you know take a bunch of
00:23:30.480 their domain expertise and turn it into
00:23:32.240 software and we are starting to do that
00:23:34.039 we are starting to work with firms and
00:23:35.840 basically take you know the special
00:23:37.840 things that they do and the way that
00:23:39.360 they practice law and turn it into their
00:23:41.440 specialized system in Harvey and then
00:23:43.360 they go and sell that to their clients
00:23:44.960 right and so that's a completely new
00:23:46.320 business model uh how does that happen
00:23:48.520 who drives that at one of your partners
00:23:50.320 and yeah what's the incentive um I mean
00:23:52.720 it is massively based off the law firm
00:23:55.919 um I think that you will find a couple
00:23:57.679 partners that think that this is the
00:23:59.320 future and they will rally the entire
00:24:01.559 firm behind them and that's what we'
00:24:03.120 have seen and it's really started
00:24:05.080 snowballing in the past kind of 3 to six
00:24:07.080 months and the way that it works really
00:24:09.520 is it's a combination of you know they
00:24:11.640 provide the domain expertise and we
00:24:13.400 provide the translation of the domain
00:24:15.520 expertise and then the tech right um the
00:24:18.039 incentive for them is there is so much
00:24:20.640 legal work that law firms actually do at
00:24:23.720 a loss and they do it at a loss it
00:24:26.200 doesn't feel like that I know I know but
00:24:27.799 there actually is is um there's a lot of
00:24:30.080 legal work that they do at a loss in
00:24:31.760 order to get the big deal so you know a
00:24:34.320 lot of law firms will do work for
00:24:35.720 private equity and they will do that
00:24:37.720 work at a massive discount or even again
00:24:39.960 even at a loss in order to get the lbo
00:24:42.399 or the m&a right and so I think these
00:24:44.640 systems are a way for law firms to
00:24:46.679 compete in these spaces and it also
00:24:49.320 another way for law firms that maybe
00:24:50.720 aren't as large and don't have as much
00:24:52.240 of those resources and they can compete
00:24:54.200 in these others because they have
00:24:55.480 software margins for this work and then
00:24:57.679 they can get the really intense
00:24:59.159 strategic deal afterwards MH yeah that
00:25:02.120 uh feels analogous to like the client
00:25:04.559 service work yeah um in an investment
00:25:06.960 Bank to work on you know large
00:25:09.159 transactions that really matter yeah
00:25:10.720 exactly I mean you see this already and
00:25:12.760 it grows over time the legal industry is
00:25:15.760 one of reasoning how much has the growth
00:25:18.880 of models and interest in developing
00:25:21.200 models that um scale test time inference
00:25:23.919 impacted you guys massively in a good
00:25:26.120 way um the best way to think about about
00:25:28.520 this is if you are building a system
00:25:32.000 where you are trying to basically break
00:25:35.279 down every single problem into a sub
00:25:37.039 problem because the models can't quite
00:25:39.399 do it that just unlocks different pieces
00:25:41.880 of it so let me give you an example of
00:25:43.320 this if we go back to that antitrust
00:25:44.960 example right where the first step of it
00:25:47.279 is we're trying to take all of the you
00:25:49.240 know Target financials your acquires
00:25:50.919 financials and just say in all these
00:25:52.799 different countries you know this is
00:25:54.120 what you need to file well the next step
00:25:56.360 to that would be can you help it
00:25:57.960 actually do all of the filings right and
00:26:01.240 we were having trouble figuring out how
00:26:03.279 to do that now with reasoning models you
00:26:05.520 can start unlocking those steps right
00:26:07.159 and so the best way to think about this
00:26:09.240 is we are constantly building all of the
00:26:12.000 steps that we can and being on The
00:26:13.399 Cutting Edge and then when a model
00:26:15.360 improves that just pushes our ability to
00:26:17.880 go out the next Cutting Edge more um and
00:26:20.840 then the other thing too is just cost
00:26:22.080 going down is incredible for us right we
00:26:24.679 try you know we're not optimizing for
00:26:27.120 cost at all times right now we're
00:26:28.640 optimizing for quality and if the prices
00:26:31.720 go down that makes it so that we can
00:26:33.600 increase our quality across every single
00:26:35.600 user base much
00:26:37.240 faster how do you think about the
00:26:39.240 domains that are not legal right you
00:26:41.799 have um announced like research and work
00:26:44.360 and tax and audit now yeah um what makes
00:26:47.120 this the right time to do that yeah so
00:26:50.000 the the way that we think about it is
00:26:51.600 there's so many areas where legal is
00:26:53.799 kind of the tip of the spear and then
00:26:56.200 you can actually just kind of paralyze
00:26:58.600 the same things that you're building and
00:27:00.520 tweak them and it works really well in
00:27:02.080 other Industries um so an example of
00:27:04.559 this is you know you do tax diligence
00:27:07.200 and then you do commercial diligence and
00:27:08.559 then you do Financial diligence right
00:27:09.919 over all of these documents and tax and
00:27:12.399 legal diligence there are a lot of
00:27:13.760 similarities in terms of what is the
00:27:15.880 high level problem that you are trying
00:27:17.279 to solve right it is basically apply all
00:27:19.880 of these rules to these documents right
00:27:23.080 um and so we have found that there are a
00:27:25.720 bunch of ways to take what we have
00:27:27.360 learned and things that we've built in
00:27:29.440 legal and then just improve on that and
00:27:32.159 apply it to tax or deals or these other
00:27:35.240 areas so I'm thinking about like how
00:27:38.320 unique it feels um in terms of just the
00:27:41.240 product effort at Harvey I think is like
00:27:43.279 very first of its kind and feel very
00:27:45.120 lucky to be involved when you think
00:27:46.880 about like the team that goes and works
00:27:49.039 on this like does your like you um
00:27:53.600 engineering hire understand like tax
00:27:56.760 diligence after a while how do you think
00:27:58.399 about like what makes somebody like the
00:27:59.840 right next person to add to Harvey we're
00:28:01.600 working on it um you are hiring yes yeah
00:28:04.159 tons um we're working on the context
00:28:06.600 sharing I think that is something that's
00:28:08.440 really important one thing that was
00:28:10.480 really nice that I actually think we did
00:28:11.960 get right is the respect is incredibly
00:28:14.640 high and so what I mean by that is a lot
00:28:17.240 of Engineers don't have experience in
00:28:18.840 legal and Professional Services and we
00:28:21.240 had a couple talks that were basically
00:28:23.440 talking about these the structure of
00:28:24.919 these really complex like take private
00:28:26.880 deals and you could just see the
00:28:28.799 engineers basically saying oh wow this
00:28:31.320 is the work that all of these folks are
00:28:32.840 doing is really impressive right um and
00:28:35.399 so that I think we've done well I think
00:28:37.360 on the context sharing it is a difficult
00:28:40.120 problem where the a lot of the folks
00:28:42.240 building the product they haven't
00:28:44.320 actually used it in the sense of in
00:28:47.159 their past career sure you know they
00:28:49.240 they haven't done tax diligence or or
00:28:51.080 whatever it is um you guys do like a
00:28:53.559 weekly suits viewing something is that
00:28:55.799 how this and then and then you're done
00:28:57.720 um yeah if you watch all of Suits we
00:28:59.320 actually just make everyone watch every
00:29:00.840 season every season of suits and then
00:29:02.760 they're good um no it's a very difficult
00:29:06.720 problem I would say that the way that we
00:29:08.840 look for hiring folks is basically
00:29:12.279 agency right like this is one of the
00:29:13.840 biggest things
00:29:15.840 is there are so many things that we are
00:29:18.399 doing that is new and I have found that
00:29:22.880 over time hiring folks that might not
00:29:25.760 have a bunch of experience in this
00:29:27.320 particular thing
00:29:28.720 doesn't matter like it is so important
00:29:31.159 that you hire people that are really
00:29:32.600 smart hungry care about what you're
00:29:34.679 doing and are willing to you know be
00:29:37.279 very decisive in what they decide to
00:29:40.240 ship and if it doesn't quite work
00:29:41.840 iterate on that and and uh go again that
00:29:44.519 is better than necessarily someone who
00:29:46.720 has tons of experience and I am very
00:29:49.120 much doubling down on that as a company
00:29:51.000 and I I think this is probably always
00:29:53.200 the case but I think it's more the case
00:29:55.159 now because everything changes every
00:29:58.240 months and you have to be able to adapt
00:30:01.279 to that change and if you can't you're
00:30:04.559 not going to make it regardless of how
00:30:05.919 much experience you have yeah I think
00:30:08.320 that's true even in the experience of
00:30:10.120 like the founders that we are working
00:30:12.039 with yeah right in terms of like well
00:30:14.640 like your ability to predict or follow
00:30:16.720 somebody else's pattern for more than
00:30:18.880 two months at a time is is very bad yes
00:30:21.480 right um I mean some things are
00:30:22.799 monotonic like you know capability
00:30:24.640 Improvement but still figuring out like
00:30:26.519 I'm going to pick this point in the
00:30:27.600 future and like put engineering
00:30:29.760 resources on solving a particular task
00:30:31.640 because I think it'll be possible I
00:30:33.240 think it's just very very different from
00:30:35.919 building uh deterministic SAS workflows
00:30:39.880 yeah from a few years ago I think you
00:30:41.200 have to pay more attention so I think
00:30:42.840 like that's one of the biggest things is
00:30:45.240 I have found that the people that have
00:30:47.159 been very successful in this space so
00:30:48.679 far are like very obsessed um and it's
00:30:52.159 not just obsessed with your problem but
00:30:53.720 you have to pay attention to what's
00:30:54.919 happening with all the model providers
00:30:56.320 as well like you have to pay attention
00:30:58.000 to all of these different facets or I
00:31:00.799 think you're going to miss out on
00:31:02.240 something what you have you know now
00:31:04.960 been a Founder for two and a half years
00:31:06.760 yeah uh I'm sure that's a learning
00:31:08.559 experience you feel like you got a focus
00:31:11.080 on agency right in you know hiring in
00:31:14.639 the business like what did you get wrong
00:31:17.720 yeah I mean I don't think we have like
00:31:19.240 12 hours to do this um a lot a lot of
00:31:22.120 things I think if I had to say there's
00:31:24.919 one thing that I got wrong over
00:31:27.000 everything else
00:31:28.440 it is figuring out when to scale
00:31:30.600 yourself um I have a certain kind of
00:31:33.880 like tendency for how I work and there
00:31:35.880 are some things I I want to keep so one
00:31:37.399 of the lessons that I definitely want to
00:31:38.720 keep is I do think you should do every
00:31:42.159 single role for a certain amount of time
00:31:44.799 before you hire for it almost all of my
00:31:47.080 Mish hires were because I did not
00:31:49.039 understand what that role was and maybe
00:31:50.159 it's a lack of my experience I don't
00:31:51.399 know but there's a Hands-On piece having
00:31:53.960 said that you can't scale a company by
00:31:56.320 wanting to be Hands-On in everything all
00:31:58.159 times right um and I think that I didn't
00:32:01.440 spend enough time transitioning from I
00:32:04.480 mean the beginning of last year we were
00:32:05.639 40 people transitioning from everyone
00:32:08.159 knew what was going on everyone had all
00:32:09.919 the context because they were working
00:32:12.840 with me directly right um you touched
00:32:15.519 every single customer all the time yeah
00:32:17.039 basically right and like all pieces of
00:32:18.519 the product Etc and I took a long time
00:32:22.159 to Fig me sorry I haven't solved it at
00:32:24.240 all I have took a long time to even
00:32:26.559 think about solving it of you just need
00:32:29.639 to learn how to scale yourself and you
00:32:31.840 can't be in every single thing having
00:32:34.279 said that I do think that you want to be
00:32:36.760 in as many things as you can because you
00:32:39.120 are it's an environment where you have
00:32:40.480 to be very decisive and make very quick
00:32:42.360 decisions and the easiest way to make
00:32:44.279 quick decisions is you were paying
00:32:46.399 attention to everything right you kind
00:32:48.639 of monitoring everything at all times
00:32:50.840 and then your gut is normally actually
00:32:53.240 pretty right because it's not your gut
00:32:55.200 it's you have taken all of these
00:32:56.880 different stimuli and and you're just
00:32:58.080 absorbing them at all times right and so
00:33:00.200 I guess balancing how important that
00:33:02.840 piece is with actually learning how to
00:33:04.679 share context with the team how to make
00:33:06.840 sure you know upleveling everyone how to
00:33:09.519 transition yourself as a Founder has
00:33:11.679 been hard for me and I think it was
00:33:13.399 continuously hard and I'm sure if you
00:33:14.880 ask anybody they say that it's you know
00:33:16.679 it's going okay but it still needs to
00:33:18.919 work so when you agreed to do this
00:33:22.639 podcast you're like okay okay sure Sarah
00:33:24.960 but um we're going to talk about the
00:33:25.840 company not me yeah and I lied um so I'm
00:33:29.320 going to ask you just a couple things um
00:33:32.039 one is like you you work continuously
00:33:34.840 you obviously really like think that
00:33:36.519 there's um something important to be
00:33:38.240 built here uh as far as I can tell you
00:33:40.519 don't do a lot else like maybe work out
00:33:43.039 yeah that's like basically it yeah um
00:33:45.200 where does where does the like Drive
00:33:47.360 come from I mean as simple as possible
00:33:50.080 like this is the most fun thing ever um
00:33:52.320 I don't think you know I started this
00:33:53.639 company when I was
00:33:54.880 27 um there was nothing in the 27 years
00:33:59.240 leading up to there that was even close
00:34:00.760 to as much fun as this is I I think that
00:34:03.919 it is just so much you get so much
00:34:07.639 energy from things moving so quickly and
00:34:11.440 you being able to actually have the
00:34:13.440 agency to come up with an idea and then
00:34:16.199 see it built that is a crazy experience
00:34:19.599 and I think that you can do that faster
00:34:21.760 than you used to be able to and it is
00:34:23.359 addicting like it is incredibly
00:34:25.000 addicting um and so I actually think
00:34:28.000 that it's it's less like where do you
00:34:30.040 get the drive or where do you get the
00:34:31.520 inspiration I think that most people
00:34:33.199 that I've met that have been successful
00:34:35.040 in this space they just have it like
00:34:37.480 it's just naturally they love the moment
00:34:41.760 or the like very compressed timeline of
00:34:43.800 you can have a very large
00:34:45.879 impact it's the most fun you've ever had
00:34:49.159 uh uh some people like they just have
00:34:52.119 that like the the uh desire to like know
00:34:55.480 what's happening have the leverage see
00:34:57.480 that like really fast impact cycle um is
00:35:00.960 that the expectation you have of
00:35:02.040 everybody at Harvey culturally yeah good
00:35:03.960 good question um I think the expectation
00:35:07.000 that I have for everyone at Harvey um we
00:35:10.280 have a line that we have kind of like
00:35:12.520 talked about a lot and it actually comes
00:35:13.800 from Kobe Bryant um and it's j not
00:35:16.440 finished basically he was in an
00:35:18.880 interview and I I forget which game it
00:35:20.800 was but you know they're up like I think
00:35:22.599 it was like three or 2 or something like
00:35:24.280 that and a reporter asked and he like
00:35:26.480 you know how how do you feel right now
00:35:28.119 like how how are you feeling he's like
00:35:29.800 you know it feel good it's it's not done
00:35:31.440 and he's like well shouldn't you be
00:35:32.400 excited you're up like 3 0 and he's like
00:35:34.320 well the job's not finished like we're
00:35:36.280 not done like we haven't won right the
00:35:38.200 thing's not over um and I think that
00:35:41.200 what I'm trying to convey when I say
00:35:43.200 that or and hopefully what Kobe was
00:35:44.640 trying to convey when when he said that
00:35:47.400 was there are certain moments when you
00:35:50.920 need to give it your all and if you give
00:35:53.200 it your all in this massively compressed
00:35:56.440 timeline it will serve you for a very
00:35:59.359 long time afterwards and so that's what
00:36:02.079 I I think I expect of everyone that we
00:36:04.319 hire at Harvey is look you need to
00:36:08.000 recognize and trust us so it's maybe a
00:36:10.040 combination of either recognize or trust
00:36:12.640 that we are on that timeline that we are
00:36:14.880 on an incredibly compressed timeline for
00:36:18.200 honestly just all of humans right I'm
00:36:21.240 serious yeah it's a very special moment
00:36:22.800 it's a very special moment and there are
00:36:25.599 going to be insane expectations and the
00:36:28.359 company isn't always going to be there
00:36:30.240 I'm not always going to be there at all
00:36:31.880 times to tap you on the shoulder and be
00:36:33.200 like hey remember those expectations
00:36:34.720 remember this is true and so you have to
00:36:36.640 be able to do that to yourself and you
00:36:38.720 have to have that mentality right um and
00:36:41.760 so I think what that looks like is some
00:36:43.720 combination of
00:36:45.680 recognition and willingness to keep up
00:36:48.560 the intensity and keep raising the bar
00:36:50.960 um and that is an expectation of anyone
00:36:53.240 who joins yeah that that resonates very
00:36:55.960 strongly for me because I think like a
00:36:58.119 very parallel Journey yeah like this
00:37:00.800 opportunity is going to happen once and
00:37:03.000 so for me and my team and even you know
00:37:05.319 Mike who joined us a little while ago
00:37:06.760 he's like how could I not do it right
00:37:08.240 now for fun right like you know me land
00:37:10.640 of many different options but I can't
00:37:12.520 sit this one out and then if we're going
00:37:13.880 to play we're gonna fully play yeah and
00:37:15.520 and that that is maybe a really good way
00:37:17.280 to look at it I can't sit this one out
00:37:19.560 that is like a amazing signal for
00:37:22.480 somebody one other thing that you said
00:37:24.880 to me that I think is like Universal but
00:37:27.119 also a little specific to this moment in
00:37:29.520 time is um like you know you're the
00:37:31.960 founder of this company that's scaling
00:37:33.560 really fast and delivering value to
00:37:35.119 users you're not a research scientist
00:37:37.240 and it's a very technical field and so
00:37:38.880 there's some imposter syndrome around
00:37:40.440 that how do you square it for yourself
00:37:42.680 yeah that's another work in progress um
00:37:44.920 I mean I'm lucky I am also not a
00:37:46.599 research yeah there we go um so I'm an
00:37:49.520 important question actually my
00:37:50.520 co-founder is um and I think that the
00:37:53.440 thing the most effective way to do this
00:37:55.800 is just figure out a couple people that
00:37:57.800 you really trust and spend as much time
00:38:00.240 with them as possible I think maybe this
00:38:02.040 is also another piece of I think that it
00:38:04.520 is very
00:38:06.240 overvalued how much value you get from
00:38:08.680 asking people that you really respect
00:38:10.760 very specific questions I think is
00:38:12.760 overvalued I think spending time with
00:38:15.640 people that are incredible at whatever
00:38:17.400 they do is undervalued and I think it's
00:38:19.599 because the latter has like a nebulous
00:38:21.359 effect like you're like I don't really
00:38:22.960 know what I'm getting out of this right
00:38:24.920 and versus the former is I have a
00:38:26.960 question they gave me an answer but the
00:38:28.960 reality is they don't have they have
00:38:31.240 like a basis point of the context that
00:38:33.319 you have right if that and so what has
00:38:36.440 been really helpful for me is just
00:38:38.200 spending time with a lot of people that
00:38:39.920 I really respect in this industry and
00:38:42.440 you start to just learn the intuition
00:38:45.000 that they have for things you start to
00:38:46.760 absorb it and all of a sudden it starts
00:38:48.680 becoming your intuition right um I think
00:38:52.160 that's the only way that I've gotten to
00:38:55.079 this point uh and I am going to keep
00:38:57.240 doing doing as much as possible yeah I
00:39:00.119 get comfortable with it mostly just from
00:39:01.880 my I mean actually what you describe
00:39:03.400 makes sense to me like I'm like I'm glad
00:39:05.480 Andre is hanging out in the office exact
00:39:07.440 me stuff but but also like I I think
00:39:09.960 like the outcomes make me feel better
00:39:11.760 correct right where I'm like well
00:39:13.440 companies look like they're working yeah
00:39:15.200 right um and users seem to be doing
00:39:17.119 something in the product or you know
00:39:19.040 consuming more infrastructure whatever
00:39:20.359 it is and um I I think as you said if
00:39:22.880 you get the repeated pattern back that
00:39:24.800 your intuition is right then I'm a prag
00:39:28.079 yeah and I I do think so one thing we
00:39:30.319 had was in the beginning of the of last
00:39:33.000 year we had a lot of kind of like usage
00:39:36.880 problems in the sense that a lot of
00:39:39.240 customers were actually getting blocked
00:39:41.240 from using our product or they got to
00:39:43.640 our product and they didn't know how to
00:39:45.119 use it right and so we had massive power
00:39:47.520 users and then we had some other
00:39:49.160 problems and actually about six seven
00:39:52.440 months into the year it had done a
00:39:54.599 complete 180 and it was working really
00:39:56.280 well and we took a bunch of bets on
00:40:00.200 product things that not product features
00:40:03.240 that not only were kind of bets we made
00:40:05.200 but also bets on the models getting
00:40:06.560 better things like follow-up questions
00:40:09.119 things like you enter a user profile and
00:40:11.720 it tells you kind of like here are all
00:40:13.760 the different things that you should do
00:40:14.880 based off of your practice area and
00:40:16.280 things like that and it worked right and
00:40:18.640 so speaking of kind of the intuition you
00:40:21.640 start to absorb that intuition you start
00:40:23.440 making bets on your own and then you see
00:40:25.280 it work and all of a sudden you get a
00:40:26.839 little bit more confident hopefully not
00:40:28.000 too confident um but a little bit more
00:40:30.520 so you're saying you and the people at
00:40:31.839 Harvey feel the AGI just enough just
00:40:34.400 just enough not too much not too much
00:40:36.760 yeah yeah it's not clear what too much
00:40:38.119 is actually yeah I don't think there is
00:40:39.520 a limit at this point so like the you
00:40:42.960 guys have been on like the right side of
00:40:45.599 a uh raging debate in the investing and
00:40:49.119 Tech Community for a long time which is
00:40:50.880 just like is there value in the
00:40:52.240 application ler like it's pretty clear
00:40:54.000 there is now um what advice would you
00:40:56.680 have for the many like Founders Tech
00:40:58.520 business people that um are going to
00:41:00.400 listen to this about uh like a good
00:41:03.440 application to build yeah um so one
00:41:07.079 thing that I look at is you can look at
00:41:09.280 this by industry and you can look at it
00:41:10.680 by task which is how expensive as the
00:41:13.760 token um and I I've come up with like
00:41:16.560 different ways to call this but
00:41:17.800 basically if you look at something like
00:41:19.720 that share purchase agreement or a
00:41:20.960 merger agreement or just legal in
00:41:23.079 general the cost or the price of
00:41:26.040 producing each part of a word in a
00:41:28.680 50-page document can be incredibly
00:41:31.680 expensive right and so I think those
00:41:34.520 areas like that is one of the tests that
00:41:36.400 I would do for these things the thing I
00:41:38.200 would stay I I would make sure you don't
00:41:40.920 discourage yourself from is this
00:41:42.960 industry has never been you know touched
00:41:45.440 by Tech or or this is not something that
00:41:47.640 I know that t tons about
00:41:49.800 Etc you can really figure out a lot
00:41:52.599 about what people do by just talking to
00:41:55.359 them and asking a bunch of questions
00:41:57.400 right and it it might not you know you
00:41:59.480 might have to talk to a lot of people
00:42:01.480 and that might not be exactly what you
00:42:03.280 thought it was going to be as a Founder
00:42:05.000 you know you might not just be sitting
00:42:06.240 in uh I don't know like a construction
00:42:08.079 firm for I don't know 20 hours a day but
00:42:10.680 it's insanely useful right and so
00:42:13.280 whatever the area is that you're
00:42:15.040 interested in I would look at does this
00:42:17.319 have long-term value in the sense of you
00:42:19.359 know what is the price per token things
00:42:20.920 like that and then actually spend time
00:42:23.800 with it and you'll get a much better
00:42:25.000 sense for it yeah I have a lot of you
00:42:27.559 know friendss who are talented product
00:42:29.280 and Engineering folks um and um and
00:42:32.280 business folks who want to start a
00:42:34.160 company and I I think people generally
00:42:36.599 start by like ideating in their
00:42:38.680 underwear in their house yeah and like
00:42:40.839 almost like almost
00:42:43.119 universally um some some ideas happen
00:42:45.720 that way right but almost universally
00:42:47.640 like it helps to see the problem in real
00:42:49.520 life and I'm like just go hang out with
00:42:52.680 one of your friends or at any industry
00:42:55.160 that you like have some fundamental
00:42:57.359 interest in and just like watch what
00:42:59.119 people do um because if you have a
00:43:01.680 really General Hammer that is getting
00:43:03.440 better all the time like now is the time
00:43:05.319 to do that and you're going to be much
00:43:07.040 more inspired by um ideas of what is
00:43:09.839 possible by like being faced with the
00:43:11.800 problem I think that's right and I might
00:43:13.160 say something that might upset some
00:43:15.200 people but spend some time out of
00:43:17.000 Silicon Valley too I think it's
00:43:18.680 important like I I think that you know
00:43:21.800 there and I don't necessarily mean by
00:43:23.400 the way if you're building a tech
00:43:24.480 company like I think it needs to be in
00:43:25.800 San Francisco yes to be clear
00:43:27.800 um Harvey's in SF in New York and London
00:43:30.680 and London but I do think that there are
00:43:33.680 so many Industries and so many different
00:43:36.480 areas of work that a lot of people that
00:43:39.240 have worked in Tech just don't know
00:43:40.440 anything about and I would recommend
00:43:42.960 that you spend time with those folks and
00:43:45.720 a lot of the times you know just your
00:43:47.880 kind of small network of friends they
00:43:49.960 might not actually be a great
00:43:51.440 representation of that industry right it
00:43:53.839 might be a you know who knows what what
00:43:55.920 version of it but it might not be the
00:43:57.480 best representation and so make sure you
00:43:59.680 are willing to explore things outside of
00:44:01.760 your comfort zone like that I I push so
00:44:05.720 strongly is there anything that has like
00:44:08.440 really surprised you in the company in
00:44:10.599 the last couple months yeah
00:44:13.800 um there are a considerable amount of
00:44:17.359 people um at the company that do not
00:44:21.400 necessarily have management experience
00:44:24.000 and I put them in a management role and
00:44:26.119 the Improvement has been insane um and I
00:44:28.880 think it has not just surprised me it
00:44:31.319 has surprised a bunch of our leadership
00:44:33.000 team it has surprised our investors it
00:44:34.680 has kind of surprised everyone I'm
00:44:36.760 really proud of that like incredibly
00:44:38.599 proud of that um and I think you know
00:44:40.680 maybe part of it is my background as
00:44:42.079 well of I don't have experience in this
00:44:44.200 and so I'm willing to take bets on
00:44:46.400 people that do not have maybe like the
00:44:48.680 perfect resume or the perfect experience
00:44:51.359 um and we've seen it work out really
00:44:53.359 well um and I I guess maybe it's been
00:44:56.839 it's surprising because everyone said
00:44:59.280 that's not going to happen um and so
00:45:01.520 that's maybe why it's surprising like if
00:45:03.240 you could pick like top two three
00:45:04.520 signals of like why this person should
00:45:06.800 be a leader for you yeah um they I mean
00:45:10.000 care is one of the most important things
00:45:11.880 like there is a sense of ownership
00:45:14.160 whether that ownership shows itself in
00:45:15.720 Obsession whether that ownership shows
00:45:17.880 itself in doesn't make excuses and like
00:45:21.200 says I got this wrong these are the
00:45:24.280 things that I got wrong I'm going to fix
00:45:26.160 them one of the best ways to tell this
00:45:28.240 is there are a few folks that will come
00:45:30.359 up to me and basically just not ask for
00:45:32.599 positive feedback and just say what did
00:45:34.319 I get like what do I need to improve
00:45:36.400 that is a massive signal because I think
00:45:38.920 that something that happens as you get
00:45:40.680 more senior in your career is you don't
00:45:42.960 want to look like you made a mistake
00:45:44.880 like a lot of what you're doing is kind
00:45:46.160 of like covering and protecting your
00:45:48.079 position and it's really hard to learn
00:45:50.200 that way and get better and the reality
00:45:52.000 is like you learn from success you also
00:45:54.319 learn from mistakes I think a lot or you
00:45:56.359 learn from like taking mindfulness with
00:45:58.319 your mistakes um and so that's maybe one
00:46:00.559 of the best signals ever is how do they
00:46:02.680 take ownership is it an obset whether
00:46:04.480 that's a combination of obsessiveness
00:46:06.000 plus like
00:46:06.960 self-reflection plus I just want to get
00:46:09.559 better okay uh last question for you if
00:46:13.160 you look outside of your own ecosystem
00:46:16.400 of legal and Professional Services or
00:46:18.359 maybe just you know one concentric
00:46:20.800 Professional Services is fa game biggest
00:46:22.839 change you think happens this year um I
00:46:26.200 think we will end up
00:46:27.920 seeing work completion done in very
00:46:31.480 sophisticated manners like in medicine
00:46:34.040 and encoding like a lot encoding um and
00:46:36.960 a lot also in you know other areas too
00:46:39.760 that will kind of bring us back to the
00:46:41.920 chat gbt moment but it will be
00:46:44.319 specialized and so there will be a
00:46:46.240 doctor that sees something or a
00:46:47.960 researcher or something and it will be
00:46:50.599 like it was the first time they saw chat
00:46:52.280 tobt but they'll show it to everyone
00:46:53.440 else and it'll just be like I don't get
00:46:54.520 what's going on here right my point is I
00:46:57.240 think these systems are going to get to
00:46:58.760 the point where they're doing things
00:47:00.839 that are incredibly impressive even
00:47:03.480 though we're used to them now and it
00:47:05.160 will be kind of a new version of
00:47:06.720 impressive but it will be impressive to
00:47:08.960 very verticalized and specialized people
00:47:11.760 I agree with that if I think about the
00:47:13.440 like implication for um us or for
00:47:16.520 Founders like one of the things that has
00:47:18.599 just felt really um impenetrable about
00:47:22.119 industries that Silicon Valley folks
00:47:23.559 don't spend a lot of time in yeah is
00:47:25.559 they're like oh well you know know the
00:47:27.640 the customer won't buy or there's a
00:47:29.559 bunch of regulation for example
00:47:32.079 Healthcare um or uh uh or like we can't
00:47:37.200 it's not obvious how to cut the task in
00:47:39.440 a way where the minimum viable quality
00:47:41.160 is
00:47:42.000 like enough of a value ad yeah where
00:47:44.680 like the like the the ROI is there and I
00:47:47.280 think like I think we're going to get at
00:47:49.000 accelerating pace of people saying like
00:47:50.880 I can figure out how to shape the
00:47:52.720 capability to it like task completion
00:47:55.520 that people want and actually like these
00:47:57.559 industries are far more open to that
00:47:59.240 change than than we thought right at
00:48:01.200 least that's been that's been my
00:48:02.240 experience where I'm like oh like as
00:48:04.160 maybe as we were saying earlier like the
00:48:05.839 ROI just weren't wasn't there yeah
00:48:07.960 lawyers or doctors or whatever else they
00:48:10.240 um they want it like you just have to
00:48:12.640 figure out the shape that I think is the
00:48:14.520 biggest piece is I think people are
00:48:17.800 confused and they don't think that they
00:48:19.599 want it and the reality is a lot of
00:48:21.880 these folks do like they don't want when
00:48:23.440 you talk about Ai and it's like a first
00:48:25.480 blast or whatever it's like sky and all
00:48:27.200 these things yeah no one wanty replacing
00:48:29.280 your course want that but when folks can
00:48:32.880 see and they actually use these tools
00:48:34.960 they want this and it's because and
00:48:36.760 especially the folks that care tons
00:48:38.520 about their profession because a lot of
00:48:41.000 these conservative you know professions
00:48:43.640 haven't changed in a really long time
00:48:45.720 and so there's you know the spirit of
00:48:47.520 them is still there but there's all this
00:48:49.799 kind of junk on top of that Spirit right
00:48:52.559 and and this is true in medicine this is
00:48:54.319 true in law this is true in all of these
00:48:55.960 areas and I think that you will find a
00:48:58.799 bunch of just like you said people who
00:49:01.040 are really willing to fight for you as a
00:49:02.880 company because they do want that change
00:49:05.040 there just hasn't been enough of a
00:49:06.599 capability for them to really actually
00:49:08.920 put their you know their neck on the
00:49:10.280 line to push
00:49:11.599 it um awesome this has been great thanks
00:49:14.040 Winston yeah thank
00:49:15.920 you find us on Twitter at no prior pod
00:49:19.480 subscribe to our YouTube channel if you
00:49:21.119 want to see our faces follow the show on
00:49:23.359 Apple podcast Spotify or wherever you
00:49:25.720 listen that way you get a new episode
00:49:27.440 every week and sign up for emails or
00:49:29.200 find transcripts for every episode at
00:49:31.319 no- pri.com