Latest / Technically Sound Strategy / Episode 42 - Why Humans Beat Bots with Ted DeBettencourt

Transcript

Episode 42 - Why Humans Beat Bots with Ted DeBettencourt · Technically Sound Strategy

• 0:01
  Welcome to Legal Marketing Radio, a production of LaFleur Marketing.
• 0:05
  Today's guest is Ted DeBettencourt, co-founder of Juvo Leads, and a
• 0:08
  veteran of the legal marketing space.
• 0:11
  Ted started out running a legal marketing agency in Boston and
• 0:14
  frustrated by the limitations of chatbot services took a hands-on approach
• 0:18
  that eventually grew into Juvo Leads.
• 0:20
  A nationwide live chat team now serving over 1100 law firms.
• 0:25
  We're diving into a topic that's increasingly relevant in the age of AI.
• 0:29
  Why humans still beat bots when it comes to converting digital leads.
• 0:34
  is great to have you here with me today, Ted.
• 0:36
  We've been working with Juva for quite a few years and we tend to
• 0:41
  recommend the platform because we've seen the results that you bring.
• 0:44
  But, some of the things that I think set you apart you and I have talked
• 0:47
  about, but they tie in really closely with our philosophy, which is that.
• 0:53
  You really need that human element, right?
• 0:55
  You really need to bring that human element in, especially when you're
• 0:57
  dealing with things that are maybe a little bit traumatic or people
• 1:00
  at different parts of their lives.
• 1:03
  We make a lot of pushes towards AI and we implement AI and a
• 1:06
  lot of the things that we do.
• 1:07
  But when it comes to our voice and when it comes to the voice of our
• 1:11
  clients and producing content, we rely on it to a point, and then
• 1:15
  we need that human interaction.
• 1:16
  So we really have this human AI hybrid that we work in, but mm-hmm.
• 1:20
  But give us a quick, give us a quick elevator pitch.
• 1:23
  What is juve leads?
• 1:24
  Where'd you come from?
• 1:25
  And what are you doing today?
• 1:26
  Sure.
• 1:26
  So Juva Leads is a human powered chat started by me and
• 1:30
  a partner back in 2015, 2016.
• 1:33
  I had a small marketing agency up in Boston, was doing the SEO, the PPC,
• 1:37
  the content, some of this stuff that you're doing now for your clients chip,
• 1:41
  but uh, you know, at a much smaller scale and much kind of different level.
• 1:45
  kind of the, everything, the SEO guy, the PPC guy the content guy.
• 1:48
  I was using different chat plot products for my clients.
• 1:52
  I was kind of the go-between, like write the chat scripts, tell it to 'em, and then
• 1:56
  the firms would tell me if they liked it.
• 1:58
  What happened was all the chat companies I was using kept screwing
• 2:01
  up the chat scripts, so I was getting pretty frustrated 'cause my firms were
• 2:04
  yelling at me and I was like, this is, I'm not really even doing it, but
• 2:07
  I was getting yelled at it for it.
• 2:08
  So what I realized is that no matter how many times they told
• 2:12
  the chat companies to make changes.
• 2:14
  They'd say they make 'em, but they didn't make 'em.
• 2:16
  So I got would get frustrated and I actually fired 'em and
• 2:19
  started answering shots myself.
• 2:21
  And then when I learned, after doing it for a few months, I'm like, oh wow, you
• 2:24
  know, I'm getting a hundred percent, 60%, 80% more leads from my firms.
• 2:29
  What am I doing differently?
• 2:30
  And I said, oh, I'm quicker.
• 2:32
  I'm more empathetic.
• 2:33
  I'm able to start conversations and answer 'em faster in ways
• 2:36
  that the firm would want us to.
• 2:38
  So I said, oh, well, maybe I can apply this knowledge and try on other firms.
• 2:42
  So first I tried it as like an add-on to get more SEO clients,
• 2:46
  and I said it doesn't make my SEO any better, to be honest with you.
• 2:49
  It's really just about the chat side.
• 2:50
  So I sAId, all right, why don't I just try it for some clients?
• 2:52
  So what started is me answering chats for about, 10 firms in Boston has now
• 2:56
  turned into us answering chats for about 1100 firms around the country.
• 3:00
  And I'm thankfully no longer a chat agent, so it's not me replying anymore.
• 3:04
  That's awesome.
• 3:05
  So you've been around now as, AI has kind of taken hold and, as it has
• 3:09
  kind of infiltrated more and more industries, more and more services.
• 3:13
  So give me just a broad perspective on that.
• 3:15
  How is that impacting you and what are you seeing in the industry?
• 3:18
  Sure.
• 3:19
  AI is a great tool.
• 3:20
  I love it.
• 3:21
  And we get this question all the time.
• 3:22
  Why do you use people?
• 3:23
  Why not use AI?
• 3:25
  If we went with a pure AI approach and we're always testing, we probably have
• 3:28
  about 30 AB tests running right now.
• 3:30
  If we went for a pure AI model, what would happen?
• 3:33
  Would the.
• 3:34
  Started the conversation's easy, so AI can handle that.
• 3:38
  It's, you know, get the name and phone number.
• 3:40
  What happens is when you get down further into the conversation,
• 3:43
  AI will kind of lose track of it.
• 3:45
  So then what we're seeing is more of kind of a hybrid model, but
• 3:48
  the hybrid model to date has been more very systemic, systematic.
• 3:52
  So they ask a question and then they get you the name, get the phone
• 3:54
  number, when did the accident happen?
• 3:56
  Where did it happen?
• 3:57
  When did it happen?
• 3:58
  If it's a, PI firm, and what we find is that it's too robotic.
• 4:03
  So it has to be, right now the AI models out there aren't good enough for people
• 4:08
  because it can't take over all the way.
• 4:11
  Like with content, if a client says, Hey, I just wrote this letter, they just wrote
• 4:15
  this article on chat, GPT, can you post it like we can, but 60 other firms have
• 4:20
  this exact same one and it's kind of junk.
• 4:22
  So that's kind of where we are with on the content side.
• 4:25
  If it's not good enough to write a blog article on itself.
• 4:27
  It isn't good enough to have a conversation with itself, and
• 4:31
  what we see is that 62% of our chats start with a question.
• 4:35
  What we're finding is that people aren't raising their hand and saying
• 4:38
  like, Hey, I want to be a lead.
• 4:39
  Hey I want a free consultation.
• 4:41
  They're saying, Hey, here's my situation.
• 4:43
  Do I, is this something you can help me with?
• 4:46
  So by using a person, we're able to take their question, reassure them
• 4:50
  that they're in the right place.
• 4:51
  So they've been through an injury.
• 4:52
  Empathize with them.
• 4:53
  Oh my goodness.
• 4:54
  I'm sorry.
• 4:54
  You're going through that.
• 4:55
  Sounds horrible.
• 4:56
  And by having that empathy, by having that real conversation, by reassuring
• 4:59
  that they're in the right place, we're able to get a lot more leads than
• 5:02
  any of the AI, any of the foe AI.
• 5:04
  There's a lot of foe AI out there as well.
• 5:06
  Models that are more like a form that you fill out that says it's AI or
• 5:12
  tries to give a, just an AI answer.
• 5:14
  So we're able to get a lot more leads than any other model in the market.
• 5:18
  That's awesome.
• 5:18
  So , a couple times you've cited some specific statistics.
• 5:21
  How are you keeping track of that?
• 5:22
  You said you had you running 30 plus AB tests, you've got six, 2%.
• 5:28
  Chad, start with the question.
• 5:29
  What are you measuring?
• 5:30
  I feel like what gets measured gets managed.
• 5:32
  Mm-hmm.
• 5:32
  What are you using to measure and how does that inform how you run the team?
• 5:37
  So everything we do, everything we've had to develop internally.
• 5:41
  'cause when we first came out, we had people that were saying, Hey, I
• 5:44
  don't know if chat gets more leads.
• 5:46
  So at the time we created an AB testing module for all our clients.
• 5:49
  Chat would show up on half the traffic chat would show
• 5:51
  up on not half the traffic.
• 5:52
  And then after 30 days, we'd look at the night data and
• 5:55
  say, who got you more leads?
• 5:56
  So we set up the call tracking, we set up the form tracking so
• 5:59
  we could track that as well.
• 6:01
  Now, most of the time the things we track against are other chat companies as well
• 6:04
  as scripting changes and different popups.
• 6:07
  So we have the ability, no matter who they come from, to give us
• 6:10
  either another chat code or have two different variations of a script.
• 6:13
  Or maybe a client won't like a particular popup and they'll
• 6:16
  say, Hey, start chat suppressed.
• 6:18
  So if they start chat suppressed, we generally see about 50% more less leads.
• 6:23
  So instead of just saying to, to a client, we'll say, Hey, let's do a
• 6:27
  split test to see what does better.
• 6:28
  Because at the end of the day, 90% of the clients, they want more leads.
• 6:32
  So we can say, well, let's try that high hypothesis and test
• 6:36
  it and see what does better.
• 6:37
  The biggest one we've been CHI testing lately is some clients love video.
• 6:41
  Some clients hate video, so we do a lot of AB split tests against
• 6:44
  video chat, non-video chat, and just measuring the data.
• 6:47
  And honestly, the AB split test is kind of what drive our company so.
• 6:51
  If we see another chat company doing something cool, we can
• 6:54
  AB split test against it.
• 6:55
  And maybe we don't have a feature that they do, we'll just build it.
• 6:59
  But we don't know how effective some stuff like that is until we AB split test
• 7:03
  against it and can really see the results.
• 7:04
  So what drives us is the AB split test and the tagging our team does.
• 7:08
  So a lot of tagging on all the different ways we're answering chats to make sure
• 7:12
  we categorize things and make sure we can look at the data post split test.
• 7:17
  That's cool.
• 7:17
  So do you think that there's do you think are, do you think that
• 7:21
  there are some ways to enhance the live operator model with AI tools?
• 7:26
  There can be, I haven't seen a way that does it perfectly yet.
• 7:30
  The ones I'm seeing on the, in the future is we're paying a lot
• 7:34
  of attention to phone answering.
• 7:36
  That does AI, that's AI phone answering that's coming right now.
• 7:39
  It's not quite there yet.
• 7:40
  When you call, you can tell you're talking to a person, but as that goes, we thought
• 7:45
  that chat would be ahead of that curve.
• 7:47
  What we're seeing is that chat's probably gonna be behind that curve.
• 7:50
  Um, so we're watching that space closely because we know as
• 7:53
  soon as the phone answering is caught up, chat will be as well.
• 7:57
  Yeah.
• 7:58
  But right now we're, we are getting more and more chats, people asking like, give
• 8:02
  me the recipe to your favorite key line pie, or something random, because they're
• 8:06
  trying to see if we're gonna trigger like a chat GPT response, which we don't.
• 8:10
  Um, yeah.
• 8:11
  But we've been seeing more of those chat prompts come through on our end.
• 8:14
  I mean, it's become more and more prevalent.
• 8:16
  I know just from personal experience talking to clients, things like that.
• 8:19
  It's funny because the expectation is that I. We can use a service that's
• 8:25
  fully AI and we'll get the same results.
• 8:27
  But then in the same conversation they'll tell me about an experience
• 8:30
  they have working with a bot and how much they hate it.
• 8:33
  They'll go through a drive through somewhere and they'll
• 8:34
  go, this is so stupid.
• 8:35
  This is so frustrating.
• 8:36
  Just gimme a human being.
• 8:37
  And it's the same thing.
• 8:40
  And you touched on something that really rings true to me too, is.
• 8:43
  That AB split test and actually making database decisions, right?
• 8:47
  We're terrible at it, I think as a species, like humans are
• 8:50
  not very good at doing it.
• 8:51
  Um, but then once we commit to doing that, and once we say we wanna maximize this
• 8:57
  metric, we've gotta listen to the data and we've gotta make decisions accordingly.
• 9:01
  And that can be very hard.
• 9:02
  You made the example of some clients like video, some don't, some like
• 9:06
  that, more that pop up, that pops up.
• 9:08
  Some don't like it.
• 9:10
  Ultimately, do you like more clients?
• 9:12
  And if the answer is yes, then that probably supersedes
• 9:16
  that preference, you know?
• 9:18
  So I'm on board with that.
• 9:20
  We've built out a product that we call Clearboard.
• 9:22
  Mm-hmm.
• 9:22
  And this may be something to not the product necessarily, but possibly.
• 9:26
  But we're pulling a bunch of different data sets from all of the different
• 9:30
  marketing channels that we monitor.
• 9:33
  So Google Analytics, GF four, Google my business.
• 9:37
  Chat actually, we pull some of that in.
• 9:39
  We actually do pull your data in what we have access to and, a
• 9:43
  number of other different systems.
• 9:44
  And we pull it into a data warehouse and then we layer on another product.
• 9:48
  So that we have basically GPT driven.
• 9:52
  Queries of our data sets.
• 9:54
  And I'd love to see as time goes on, I think the AB split tests are great
• 9:59
  and I think that those should continue to be designed, but I think there's an
• 10:02
  opportunity to pull in a data warehouse and start asking GPT, So our, the
• 10:07
  GPT built into our system actually writes the sequel to query the data.
• 10:12
  And how it does it is it gets the data writes the sequel restates.
• 10:16
  The query to say, this is what we wrote it to do.
• 10:18
  Is that right?
• 10:19
  And then it will run it, and then it will deliver these visualizations along
• 10:23
  with a verbal, like a textual response.
• 10:26
  And so I guess to frame this as a question are there any things that have really
• 10:30
  surprised you as you've done this deeper dive into the data that you're doing?
• 10:34
  What we're able to do is we're able to find, we're able to find correlations
• 10:37
  that a human wouldn't otherwise find.
• 10:40
  Those correlations are informing decisions that we make.
• 10:43
  And sometimes those things are unexpected, right?
• 10:45
  A, a split test, AB test needs to be designed and it
• 10:48
  needs to make an assumption.
• 10:51
  And the correlative events when we use a GPT, don't require the design.
• 10:56
  I. So there's advantages and there's a time and a place for both, and neither
• 11:01
  one of them is better than the other.
• 11:02
  You need both to find answers, but I'm curious, what, I'm curious kinda what you
• 11:06
  have in the works on going on that road, how do some of these insights you have to
• 11:08
  arrive at outside of just split testing?
• 11:11
  Right.
• 11:12
  Absolutely Chip.
• 11:13
  We've been actually using that a fair amount in our data analysis.
• 11:16
  The problem we're having is that every client's a little bit different.
• 11:20
  So what we're seeing on our end, if we ask for a data set,
• 11:23
  it's hard for us to compare.
• 11:26
  Like to unlike rules.
• 11:27
  So we have to kind of, when we're feeding it data, 'cause that's what we learned
• 11:30
  hey, let's look at the data from all our clients and see what we can learn.
• 11:33
  So we write a bunch of queries and we're like, oh, well we gotta
• 11:36
  break out the, the criminal law versus the PI versus the family law.
• 11:39
  And we're like, well actually this criminal law is having a, this PI's
• 11:42
  firm's having a signed retainers.
• 11:43
  This one isn't so.
• 11:45
  When we do that, we have to really compartmentalize the data to make sure
• 11:48
  we're able to get some good results.
• 11:50
  One of the things we have learned from that is we have a, we have this something
• 11:54
  called the invite pop, where basically if someone hasn't been on the site for about
• 11:58
  15 seconds, I think it's 16, actually, we'll prompt them to start a chat
• 12:01
  conversation and we had some good results from the AB test, but we're able to really
• 12:05
  feed that data into AI to pull it from.
• 12:09
  Like-minded.
• 12:10
  'cause at first we did all the clients, but it wasn't fair to
• 12:12
  do CRI versus family versus pi.
• 12:14
  And then we compartmentalize by only pi and then only PI that had the same similar
• 12:19
  set of rules where everybody get some really good learnings from that as well.
• 12:22
  That's cool.
• 12:22
  That's cool.
• 12:23
  So where do you see chat going in the next couple of years?
• 12:26
  I think that people are gonna continually in AI, how do you
• 12:28
  continue to differentiate?
• 12:30
  Sure.
• 12:31
  AI is the future and it's here.
• 12:33
  It's about who can utilize AI with.
• 12:37
  Right now it's with people to maximize the leads you get for your clients.
• 12:41
  Because, when I go to these marketing conferences, chip, it's every other
• 12:44
  booth is selling AI something or other.
• 12:46
  Some of them are good, some of them, a lot of 'em are great.
• 12:49
  Some of them aren't AI, but they're just saying they're AI.
• 12:52
  'cause they have this one little AI component.
• 12:54
  As long as law firms care about maximizing leads, it's the chat company or whoever
• 13:00
  that can help them achieve that.
• 13:02
  So it's gonna be who's ever at the forefront maximizing leads for the
• 13:06
  clients and being able to prove that.
• 13:08
  If there's three or four competing AI companies out there.
• 13:10
  But one's better.
• 13:11
  It's gonna be the one that's tested against the other
• 13:13
  ais maximizing the leads.
• 13:15
  And right now we're going head to head against a bunch of ais in the market, and
• 13:19
  right now we're still coming up on top.
• 13:21
  So when that day changes, we hope to be in the forefront of it.
• 13:25
  But right now for us, we're seeing still a dramatic increase
• 13:28
  in leads when you use people.
• 13:30
  That's great.
• 13:31
  That's what we see too.
• 13:32
  Because we track that data also.
• 13:33
  And yeah, we work with a number of chat companies.
• 13:35
  We're not, our bias comes from our data, right?
• 13:38
  We are successful when our clients are successful.
• 13:41
  Yeah.
• 13:41
  And so what we do is we try to pitch them towards the thing that's
• 13:44
  gonna move the needle the most.
• 13:45
  And yeah.
• 13:46
  Our experience has been very, has been awesome working with
• 13:48
  you and with your team I love it.
• 13:50
  Thank you.
• 13:50
  I appreciate you taking the time to do this.
• 13:52
  Any final thoughts?
• 13:54
  If you're using an AI chat and you wanna see the difference, check
• 13:57
  out juvoleads.com juvoleads.com.
• 14:00
  We give you a free 30 day trial so you can AB test for yourself for free.
• 14:04
  We just want to prove to you that you get more leads with humans
• 14:07
  versus anybody else in the market.
• 14:08
  I.
• 14:09
  Awesome.
• 14:11
  Thank you so much to Ted for joining us today on Legal Marketing Radio.
• 14:15
  You can connect with ted at juvoleads.com or find him on LinkedIn.
• 14:20
  Legal Marketing Radio is a production of LaFleur Marketing.
• 14:23
  Find us at LaFleur.Marketing or follow us on LinkedIn, Facebook, or Instagram.