November 19, 2025 • 48 mins
Fireflies CEO, Krish Ramineni shares how the company is transforming AI-powered note-taking into a deeper layer of knowledge automation. He breaks down the technology behind real-time functionality like Live Assist, the user behavior patterns driving product evolution, and how Fireflies is innovating far beyond meetings. Krish also shares insights on future trends in AI and the potential for hardware integration, emphasizing the ongoing evolution of AI in knowledge work.
Featuring:
- Krish Ramineni – LinkedIn
- Chris Benson – Website, LinkedIn, Bluesky, GitHub, X
- Daniel Whitenack – Website, GitHub, X
Links:
Sponsors:
- Miro – Get the right things done faster with Miro's Innovation Workspace. AI Sidekicks, instant insights, and rapid prototyping—transform weeks of work into days. No more scattered docs or endless meetings. Help your teams get great done at Miro.com.
- Framer – Design and publish without limits with Framer, the free all-in-one design platform. Unlimited projects, no tool switching, and professional sites—no Figma imports or HTML hassles required. Start creating for free at framer.com/design with code `PRACTICALAI` for a free month of Framer Pro.
Upcoming Events:
- Register for upcoming webinars here!
Mark as Played
Transcript
Episode Transcript
Available transcripts are automatically generated. Complete accuracy is not guaranteed.
Jerod (00:04):
Welcome to the Practical AI podcast, where we break down
the real world applications ofartificial intelligence and how
it's shaping the way we live,work, and create. Our goal is to
help make AI technologypractical, productive, and
accessible to everyone. Whetheryou're a developer, business
leader, or just curious aboutthe tech behind the buzz, you're
(00:24):
in the right place. Be sure toconnect with us on LinkedIn, X,
or Blue Sky to stay up to datewith episode drops, behind the
scenes content, and AI insights.You can learn more at
practicalai.fm.
Now onto the show.
Daniel (00:48):
Welcome to another episode of the Practical AI
Podcast. This is DanielWitenack. I am CEO at Prediction
Guard, and I am joined as alwaysby my cohost, Chris Benson, who
is a principal AI researchengineer at Lockheed Martin. How
are you doing, Chris?
Chris (01:04):
I am doing very well. Got to spend the morning with, with
some folks down at Georgia Techtalking AI. Cool. So
Daniel (01:12):
And you're headed my way. After this, we're gonna
meet up at the the Midwest AISummit, which is as of recording
this happening tomorrow and alsoas recording this still
experiencing government shutdownand and travel problems in The
US. So hopefully I'm hoping youmake it, our our way, but, yeah,
(01:32):
excited for the summit. It'sgonna be gonna be a fun time to
to be together in person andmeet some listeners, meet meet
some AI enthusiasts and AIcurious folks. So, excited for
that.
Chris (01:45):
I'm encouraged because our guest was on a plane,
earlier today and and and he gotthrough. So I think we should
dive right into the conversationhere.
Daniel (01:54):
Yeah. Yeah. Sounds good. I am very excited because we
have with us Krish Ramineni, whois co founder and CEO at
fireflies.ai. Welcome.
Krish (02:03):
Hi, Dan. Hi, Chris. It's great to be here. Thanks for
having me.
Daniel (02:07):
Yeah, it's really great to have you on. I see, of
course, Fireflies join mymeetings all the time, which is
cool, sort of Firefly callassistance and recording and
that sort of thing. Of course,this is probably something like
it was somewhat early on, atleast in the public perception
(02:29):
of ways people were using AIthat kind of impacted maybe
their day to day were these kindof meeting assistants that have
taken various forms and peoplehave tried various things with
these. So as we're sitting herein November 2025, like what is
the kind of, like how would youpick apart that industry from
(02:49):
your perspective? Like what arepeople trying?
What are various approaches tothat? What is the state of that
technology, etcetera?
Krish (02:56):
For me to really appreciate how far we've come as
a company and as an industry,it's good to even look just five
years back and maybe even goback a few more years before
that where we started in 2016.We were working on a whole host
of different tools all aroundthe general AI space. And
(03:17):
remember, this was beforeChadGPT existed. This was before
any LLM or crazy AI tools wereavailable. It was when me and my
co founder were reading articleson deep learning and sequence to
sequence applications and how tomake it work.
We were using crude technologiesin today's standards like BERT
(03:38):
and really manual naturallanguage processing libraries
around that time. So it was avery different time when
everyone wished and hoped thatyou could have AI understand
conversations. And that was theessence of what we were trying
to build. At the time, we didn'teven call it an AI notetaker. We
(03:59):
were trying to build thisgeneral AI secretary or EA, and
we needed to pay our bills.
So we said, we've tried a bunchof different products, in this
space, so we need to figure outwhat's the thing that people are
willing to pay for. And wouldn'tit be a great idea if someone
for a $100 a month could get anexecutive assistant with a human
(04:21):
in the loop and some AI in thebackground. So we went about
building that, testing that out,learned a lot and realized
having a human in the loop isthere's no way this is gonna
scale. If we wanna eventuallybuild this into a platform used
by millions of people every day,taking notes for you twenty four
seven, capturing all thosetranscripts, there was just no
(04:44):
way that human in the loop wasgoing to build, this sort of
company. It was also rightaround the same time companies
like Scale AI were created wherethey were doing data labeling
and the human in the loopprocesses.
So we took inspiration from ourpeers at the time, but our
ambitions were, I think a littletoo grandiose. And we said, do
(05:06):
we really want to build thisbusiness? Like the max we might
be able to support is like 5,000people, 10,000 people, and it's
going to be an operations heavybusiness rather than an AI
business. So that's when me andmy co founder said, you know
what? We validated there's amarket need for this.
And within, like, trying it outfor 10 of our close friends who
(05:28):
were paying us enough money topay for rent in San Francisco at
the time. I felt like they werejust they felt bad for us. My co
founder got a place for $750 inSF. That's probably the biggest
hustle at the time.
Daniel (05:42):
Yeah. Serious.
Krish (05:43):
And we were like, this is the only way to make things work
is we just have to, like, rentour time. So as we built that,
we validated it very quickly,and we did it without writing a
line of code. After we did thatexperiment with, like, 10 of our
friends, we said, if we reallywant to build a serious
business, we know there's aproduct market need here. Before
(06:05):
we would write code for sixmonths, we would ship something
and no one would want to use it.And we would be wasting a lot of
time.
This is the first time wherebefore we built anything or
wrote a line of code, wevalidated the market and we
ourselves were the product,right? So fast forward, we ended
up building the product. Therewas no category for AI notetaker
at the time. And that's when wecreated Fred, the AI notetaker.
(06:30):
We started with simple promisesof, hey, one click capture your
meetings.
You don't have to deal with likenative recording on Zoom because
every user, when they were usinglike, local recordings, there
was a limit to how much theycould record on Zoom and other
platforms. Some platforms likeGoogle Meet didn't even offer an
ability to capture meetings. Sowe started with a basic function
of how can I record my calls soI can go back and play through
(06:53):
them? Then we focused ontranscription, but it was really
expensive at the time and notvery accurate. And we said
that's okay.
Even if transcription is notperfect, as long as I can just
search index through theconversations and just go back
to the general time frame whenwe talked about dates or months
or something some importantkeyword that's good enough.
(07:15):
Right? That's how bad the techwas back then. So we built,
like, a way to record yourmeetings, a way to search back
through your meetings. Then asit got better, we built the
transcription layer that you canactually read.
Initially, there was nodashboard. It was just an email
that we would send out. And thenafterwards, we started building
our own task detection andaction item and keynote
(07:39):
detection engine by hand. Thatwas brutal. So we wrote the code
for that, using some off theshelf stuff and some custom
scripts that I wrote at thetime.
And fast forward, we launch in2020. So it was a four year
struggle of trying to figure outwhat we were trying to do. We
launched in 2020. It's still noChatGPT at around that time. But
(08:01):
the core product of being ableto capture, search, and see some
general bullets or sentencefragments was interesting enough
that we got our first couplepaid customers by the 2020.
So it wasn't until 2021 that weactually started making revenue.
And then fast forward, COVIDhappens, right? And so we get
(08:23):
like this proliferation of freeusers and then some of them
turning into paid users. Andthen you accelerate through to
2022, we get early access toOpenAI's GPT 3.5. Vinod Khosla,
who is an investor, alsohappened to invest in OpenAI.
That really opened thefloodgates for us. Again, LLMs
(08:44):
were very expensive at the time,but we said, you know what?
Let's go ahead and bring thisin. It can't be that bad, right?
Like whatever we're doing today,it's probably going to be
better.
It blew our mind. It changed thecomplete technology. And then
from November 2022, so exactlythree years ago to today, it's
been an absolute rocket ship.We've never looked back. Company
(09:08):
has grown, accelerated, crossedseven figures, and then eight
figures in revenue, and scalingbeyond.
So it's been verytransformational. So I think
it's fair to say we were luckyto be at the right place at the
right time, but the more honestanswer was we just showed up a
little too early and we justtried to survive for, what it's
(09:30):
worth. And eventually the buscame around and we were able to
hop on the bus. So, and thencreate this whole category of AI
note taking. And now it seemslike the most obvious generative
AI use case.
Like if your company is workingon something and you need to
pivot because it failed, thefirst thing you look at is,
(09:50):
maybe we should just buildanother AI notetaker. Seems like
it's working for Fireflies. Weshould try to copy them. So it's
a very different space now.
Chris (09:57):
I'm I'm curious. So that that's a fantastic I love the
history of that, and I love thatthe kind of the incremental
development of both the businessand the technology as you
described it. One of the thingsI'm wondering is like, it seems
like early on, you know, youtalked about that in the
beginning, you would write codefor six months and ship it and
people weren't using it. And youlearn the lesson of of kind of
(10:20):
going and making sure you had amarket before you did a major
commitment on that. And so andas you've talked about the
evolution from that point on, atwhat point did you start to see
users changing their ownbehaviors in response to how
your like, whatever state yourproduct was in at the time.
(10:41):
Like, what were the things thatyou noticed where users would
respond and they would changeand adjust their own behavior
based on the integration of theproduct into their own workflow?
And how did that behavior changeevolve over time as your product
got more and more sophisticatedand capable?
Krish (11:00):
All the other products that we built were easily
buildable, but there wasn't aclear market demand for it. And
we always looked at it from, isthis feasible from an
engineering point of view ratherthan is this something that the
market cares about? Is thissolving a killer pain point for
customers? So when we decided tobuild Fireflies the notetaker,
(11:24):
we said, let's forget what iseven theoretically possible
right now. Let's figure out whatcustomers want and we'll work
backwards from there.
That was a fundamental shift. Sowhen we had this experiment
where we bootstrapped it withjust my co founder and I and a
dozen friends where we weretesting it for them, like we
were the software. We wereeverything at at that moment in
(11:46):
time. And that was before therewas a company really built.
There was no business plan, butit was, is this something that's
valuable enough to do?
You validate it, but it's onething to validate it with humans
and now another thing to buildthe technology. Both my co
founder and I were technical. Weboth like solving hard
(12:06):
engineering problems and creditto Sam, my co founder and CTO,
went to MIT, studiedaeronautical engineering. And
for him, computer science islike the easy stuff, right? The
aeronautical engineering, he wasworking on drones, autonomous
vehicles, unmanned vehicles backin 2015, 2016.
So I think this is like, if Ican solve that problem, how hard
(12:29):
can this machine learningproblem be? Was like a Again, we
got a little carried away, butit was a very hard problem. But
I think a lot of the besttechnologists in the world, they
worry about the technology partafterwards. They're like, great.
If it's a hard problem, we'regonna find a lot of really smart
people that wanna get behind itand solve it.
(12:50):
And we did the best we couldwith the the technology that was
available at the time. So toeven be able to capture calls,
stream them into the cloud,record them, store them, index
them, search them, that itselfwas a hard problem, with the
resources that we had back then.And it was also very expensive
(13:11):
to do all of this stuff.Transcription has gone down like
10 x since that time welaunched. So that's been an
enabling factor for us.
We've scaled our owninfrastructure over the last
five years where we manage ourown bare metal servers. The
volume is insane, right? Wemaybe a great We are a multi
(13:32):
cloud platform, but we run moreof our traffic off of our own
servers than on Google or AWS.So we've gone to this economies
of scale and done all of thishard work where the end user,
they just look at Fireflies andsays, Oh wow, it's $10 per
month. And I'm getting thisnotetaker that's going to join
all my meetings all the time,every single day around the
(13:54):
clock.
Doing that in the past wouldhave cost thousands and
thousands of dollars to support.It doesn't make sense to charge
someone $10 if it's going tocost you $1,000 to support that
user. So the businessinnovation, the scaling itself
was, I think, a miracle. Thatwas like a masterclass in how
the engineering team got behindand solved it. Because not only
(14:16):
you have to optimize for qualityand reliability and uptime, but
you also have to make itaffordable enough where
someone's going to be willing topay $10 a month.
And that that was like a bigfeat. And since 2023, we've been
profitable ever since we'vestarted hyperscaling. So we've
done all of this while beingprofitable, and we didn't touch
(14:37):
any of our series a funding thatwe had raised, along the way. It
was just our seed round. Andthat initial seed round was
enough to, like, seed theinitial free users, figure out
our monetization strategy, buildout our own infrastructure, and
get it to a point where we foundproduct market fit, we found
customers, and we found revenue.
(14:58):
So yeah, that was the differentthing this time around was we
did not stop ourselves or limitour dreams from what was
technologically possible at thetime. And we absolutely got
lucky because we might not havebeen able to reach the success
that we have today had it notbeen for LLMs, but one way or
(15:19):
the other, maybe we would havebeen two years delayed, right?
Eventually this sort oftechnology would come out
because that's how fast theindustry was moving. And again,
the credit was definitely to Samon the forethought because every
time we went to investors, theywould tell us there's no way
transcription is going to beaccurate enough. There's no way
transcription is gonna be cheapenough.
In fact, many investors told us,you can make good money with
(15:41):
this human in the loop business.Why don't you just go turn that
into an actual business? Like,you ran this cool experiment
with your friends. You shouldturn that into a business. And
we were very strongly opposed tothat.
We said, we want to build asoftware company, but we don't
want to deal with operations.And this is also from some
battle scars from the past whenwe tried to build a food
(16:02):
delivery app in college and thelogistics behind it. So we said,
never again are we going to doanything that requires like
logistics and on the groundstuff. So it was a series of
optimistic decisions and thetechnology ended up catching up
to where we were vindicated.
sponsors (16:37):
So friends, you know that feeling when your team has
brilliant ideas and you'reexcited, but these ideas, this
innovation gets stuck behindendless meetings, scattered
documentation, and when you'removing at this speed of
bureaucracy instead of the speedof innovation, we've all been
there. The gap between idea,execution, impact, all that good
(16:59):
stuff, it is killing your team'sprogress. But here's the thing,
simply throwing AI at a problemdoesn't help. It just makes
things messier. That's exactlywhy you should check out Miro.
Miro is powered by AI, and ittransforms teamwork that
normally takes weeks into days.Now I've used Miro to plan
different ideas out, map complexworkflows, and I've even
(17:20):
generated fresh ideas frominterviews all in one single
place. And that is the power ofMiro. It is like having a second
brain with AI that actuallythinks like you, like your team.
No more switching between 10different tools.
Help yourself, help your teamget great things done with Miro.
Check out miro.com to find outhow. That is mir0.com. Again,
(17:44):
miro.com.
Daniel (17:50):
Well, Krish, I'm wondering, you kind of alluded
to the evolution of thetechnology, which of course has
played a key role in that andfor our kind of AI practitioners
and those that are reallycurious about more of the kind
of AI engineering side in ouraudience, could you just kind of
in generic terms, I'm sure youcan't share everything about how
(18:14):
everything works, but you hadmentioned this evolution where
it was kind of at first thesetranscription earlier
transcription models, whichweren't that accurate and maybe
you training or building yourown kind of set of other models
that would do kind of taskspecific things like a action
item detector or whatever thatis. So how does the approach to
(18:39):
something like an AI notetakernow with the models that are
available differ both in termsof what is easier, but maybe
also there's certain things thatare challenging now, which you
didn't have to think aboutbefore because of how the
technology has evolved. Wouldlove to hear that side as well.
Krish (19:00):
If you're using transcription engines today like
Whisper or one of the ASRproviders today and even some of
the big ASR providers from theGoogles of the world, Microsofts
of the world, you don't have tomassage the output as much.
That's exactly the word that wehad to think about because the
raw output was pretty crappy andwe had to fix the grammar, we
(19:21):
had to fix the punctuation, wehad to figure out speaker
identification at the time. Soyou're just getting a garbled up
string of text and you're tryingto make sense of that. And, you
also have to figure out how toidentify certain type of words
because for certain industries,certain work domains, there are
(19:42):
different types of words thatare out there. How do you deal
with filler words, pauses, Like,we speak a lot of filler words.
I feel like working on Fireflieshas improved the way I speak and
helped reduce the number offiller words I I use because the
first versions of Fireflies,every time I would read the ums
and the ahs from my own meetingnotes and recordings and
(20:02):
transcripts, I'd be like, wow, Isaid it that many times. So we
ended up building a littlefilter back in the day that told
me like how many filler words Iused in a meeting. And I would
instantly learn from that andsay, oh, wow, I need to learn to
speak slower. I need to be moreenunciated. So the original text
that was generated fromtranscription engines wasn't
(20:22):
good enough.
So we had to build a bunch ofour own filters on top to clean
up the output that wasgenerated. And then we had to
build other layers on top thatwould classify different parts
of the meetings and then pullout what it thought were action
items, what it thought werebullet point notes, what it
thought were key ideas. And atthe time, it was very
(20:44):
extractive, meaning you're justextracting pieces from this
garbled up text and then you'recalling that notes. And it's a
shame, but at the time that wasgood enough for a lot of people.
So we had to build, rebuild, rearchitect things many times
over.
Whereas if a person was doingsomething today off the shelf,
(21:06):
you don't have to do all ofthat. Like you can get to like
80% pretty solid just using offthe shelf parts, right? And that
is something that we didn't havethe luxury of. Now you may ask,
well, Krish, if it's so easy nowto build something 80%, 85% off
the shelf, what's thecompetitive moat? Everyone
(21:28):
should be able to do this.
And I think the part there thatwe were fortunate enough, one is
going really deep on the problembecause the other 15% actually
takes a long time. Thatdifferentiates good versus
great. And you have to reallypolish the product and those
things take a very, very longtime. And you have to learn from
(21:49):
your users. The roadmap thatwe've been able to build out
beyond the core tech for teams,for enterprises, That takes a
lot of work.
You have to think about accesscontrols, privacy, storage. How
do you think about multi tenant?Like we're one of the first
companies to get SOC twocompliance in the AI meeting
assistance space. And then HIPAAfor doctors that want to use
(22:12):
Fireflies. We offer this optionfor enterprises called private
storage where you can storeFireflies meetings inside your
own server storage containers,which was another architectural
change.
If you don't architect itcorrectly in the beginning, it
makes it very, very difficult.All this administrative sharing
(22:32):
features, team features, whichmakes Fireflies more valuable as
more and more people inside yourorganization start using it.
Search was a really big problemthat you have to solve over time
as well. So all of this isenabled by great transcription
and being able to understandlanguage. I absolutely agree
with that.
But then all the use cases youhave to build on top of it, like
(22:54):
our Ask Fred example, where youcan talk, instead of reviewing
the notes or the meeting, youcan just ask Fred questions
about the meeting and it willcatch you up on everything that
happened, right? So thatrequired a lot of stuff like
figuring out search from scratchand making that really
effective. So, yeah, I I thinkit it was a big journey in terms
(23:15):
of building out, like, thesecore building blocks. But as we
were able to do that, when westarted, it was a blue ocean.
And going back to I'm like Pogosticking back between the
commercials and technology partbecause both are really
important.
If you wanna be a great artist,you don't have to think about
the commercialization of whatyou're building. You can just
(23:36):
think about great art. But tobuild a great business, you have
to be a good artist and you alsohave to figure out, can someone
pay me money for this? We werefortunate enough at the time to
establish the AI Notetaker brandand be one of the first
companies to start championingthat word. And now it's an
entire category.
(23:56):
Build that AI notetaker, get todistribution, get to millions of
people using Fireflies. Today,like tens of millions of people
get notes every month fromFireflies. That in itself,
right, distribution is somethingthat is super important. And
because we started early when itwas less obvious, we were able
to get make the most of it.Like, there's no point in going
(24:18):
to the gold rush after it's beenannounced and after someone's
made a a ton of money off of itor or a killing off of it.
I think, like, being there alittle early helped us maximize
the impact because for every endnotetaker, every additional
notetaker that comes out, itbecomes progressively harder for
them to stand out in the crowd.And there is something to be
(24:42):
said about these markets wheredistribution is one of the most
important things to build in aPLG flywheel. That's why you
don't see so many Calendlycompetitors. People still use
Calendly. Like, it's one of thede facto platforms that everyone
uses.
In our space, there's probably,like, three or so big players
with Fireflies being one ofthem, and then each has its own
(25:04):
merits. Fireflies, for example,is very much focused on teams
and businesses, and if you needrobust integrations and
workflows and admin controls,you come to Firefly. Whereas if
you're looking for more of aprosumer type product, you'll go
to one of the other platforms.So we had to pick what we wanted
to do and go really deep inthere. A good parallel I like to
(25:27):
use also is like the projectmanagement space or the CRM
space, where you had Salesforcebuild a massive enterprise scale
business, and then HubSpot comesalong and takes a huge market
out of the SMB business.
Both are tens of billions ofdollars in market value that's
been captured. In projectmanagement, you have the Asanas
of the world, you have themonday.coms of the world, you
(25:49):
have all of these other projectmanagement systems that follow.
So we've found our niche. Iwouldn't even say it's a niche
because our niche is anyonethat's a knowledge worker that
works inside of a team. So wespecifically say no to consumer
grade use cases like universitystudents teaching, like those
sorts of stuff.
(26:10):
Like our bread and butter hasalways been like, if you're a
team and it doesn't have to bein tech either, like a lot of
our customers are outside oftech, which was also really
fascinating to see. But yeah,it's been a very long tedious
journey. I I wouldn't havebelieved three years ago we'd be
where we are today. But toanswer your question on tech,
(26:31):
yeah, it it's it's been one ofthe most fascinating things
about this space because now itfeels super easy and everyone
can do it.
Chris (26:37):
You hit on so many things there that I that I'm that I'm
interested in. Like, one of thethings I think that I think I
just learned something from youdefinitely on and that it did I
will probably try to share andI'm kind of I'd like to
generalize it towards notstrictly a fireflies thing
because I think there's a lot ofpeople in our audience that can
learn from this in whateverindustry they're in is is your
(27:00):
kind of your pursuit of asustainable competitive
advantage and what that meant toyou. And it sounds like, in your
case, being there early andhaving to solve, you know, the
whole problem and not, know,unlike people that would come in
later and get to that 80%easily, and you have that last
15 or 20%, which is which isreally hard, you were well
(27:21):
positioned to develop theexpertise in your organization,
and to find the niches that youwanted to service and get their
first and best with thatexpertise to do that so that
even as even as the space hasdeveloped other competitors, you
were able to to hold them offand hold your niche and be a
(27:43):
powerhouse in that way.
Is that is that a fairrepresentation of kind of what
you were just saying?
Krish (27:49):
Yeah. This was one of those, situations where being
early definitely helped, andthen having the time to keep
building and refining it andlistening to your users over
time. A lot of people claim thatSaaS is dead. Everyone will
build their own SaaS. But inreality, that extra 15 to 20% of
work you have to do to build it,maintain it, customize it, I
(28:11):
don't think most companies wantto deal with that, right?
They have other they have abusiness to run. They don't want
to be building these tools inhouse. So when we look at our
customers and look at the thingsthat we want to offer to them,
we have 95 plus differentintegrations, for example. That
was a competitive moat for usover time because we wanted to
be the most integrated AImeeting assistant on the
(28:32):
platform, on the market. We alsowanted to really take the
security thing and be the mostsecure AI notetaker for people.
So that means like when you workwith enterprises and businesses,
they have thousandquestionnaires on like all the
security compliance stuff. Youhave to deal with their CIOs.
You have to like build all ofthese like layers and features
like audit logs and all thesecompliance features, that is a
(28:56):
pain to do, right? Forget aboutthe AI, but just building the
SaaS part of it. It's like usergroups, admin controls, data
retention.
Some customers in the financeindustry want their data wiped
every seven days. So buildingall of these sort of things,
help definitely build the moat,but we're shipping ten, twenty
different features orenhancements every week, and you
(29:18):
do that over a span of fiveyears, it compounds. That's why
I feel like whenever a bigplayer gets into a space and
someone says, oh, this companyis dead now, but what ends up
happening is the other companyhas built so much that it's very
hard for you to wanna switchover. And a company that I like
(29:39):
to reference, and I thinkthey're doing a really great job
of it, is Eleven Labs, like,with text to speech. They went
so deep on a problem where manypeople would say, this is
generic.
Like, anyone can do text tospeech now. Like OpenAI or one
of the big players will just dotext to speech and offer it as
an API. But if you look at theamount of stuff that Eleven Labs
(30:00):
provides to their customers andgoing really deep on what your
users need, it's an all in onevery comprehensive platform and
it gives you so much to choosefrom. Sure, you might lose out
on the 10 or 20% of prosumertype users if other players come
in and solve the general usecase, but that's the whole point
of the company, right? Productmarket fit is not about finding
(30:23):
it once and you're good forlife.
You have to keep finding productmarket fit over and over again,
and you have to keep adapting asthe market changes, your company
has to change. Otherwise youbecome stale. Like the reason
you have CRM companies that werein the 90s versus in the 2000s
versus now, they all fell out ofproduct market fit and had to
(30:44):
keep innovating and keep gettingback into product market fit. So
yeah, being early helpful,listening to customers and
having five years to build outthe roadmap, super helpful. And
I'm one of those people thatdefinitely believes SaaS is not
dead and most people don't wantto build their own tools.
Trust me, it's not worth thehassle. You got bigger and
(31:05):
better things to do.
sponsors (31:21):
If your team is still jumping between design tools
just to update the website,that's that's not fun. Maybe
you're stuck behind expensivepaywalls for basic design
features. Here's the thing, mostdesign tools make you pay just
to get started. That's notFramer's way. Our sponsor today
is Framer.
They already built the fastestway to publish beautiful
(31:42):
production ready websites and isnow redefining how we design for
the web with the recent launchof design pages, a free canvas
based design tool. Framer ismore than a site builder. It's a
true all in one design platformfrom social media assets to
campaign visuals to vectors andicons all the way to a live
(32:02):
site. Framer is where ideas goto live, start to finish. Here's
what I love about this tool.
It is an all in one tool. Youdon't have to switch to Figma
for design and then have thesemessy HTML exports. You don't
have separate tools forgraphics. With Framers free
tier, you get unlimitedprojects, unlimited pages, and
(32:24):
you even get unlimitedcollaborators, vector tools,
three d transforms, animations.It's all there.
That's what sets Framer apart.It's not just another site
builder like Webflow or Wix. Itis a complete design tool that
happens to publish productionready websites. It's everything
you need. It's totally free.
So if your team is ready todesign, iterate, and publish all
(32:46):
in a single tool, start creatingfor free today at
framer.com/design and use thecode practical a I and get a
free month of Framer Pro. That'sframer.com/design and use the
promo code practical a I. That'sall in word. Framer.com/design,
promo code practical a I. Rulesand restrictions may apply, but
(33:10):
check it out today.
Daniel (33:16):
Well, Krish, I definitely wanna shift now to
some of the things that are, ofcourse really exciting that you
have either just released orcoming on the roadmap with
Fireflies and kind of some ofwhat's enabled that from the
technology standpoint, some ofthe challenges related to that,
some of the value that thatbrings. One of the big things
that we were talking aboutbefore we hit the record button
(33:38):
was real time functionality. Socould you describe a little bit
of what is coming out inrelation to real time and also
kind of the why of that, likewhy people would want that, what
it enables, and maybe why thathasn't come out yet in terms of
(33:59):
the technology side of things.What has enabled that at this
point in time?
Krish (34:05):
One of the biggest things that I always found when I talk
to customers was how can we helpthem in the moment while the
meeting itself is happening?Today, people were using
Fireflies where they would getmeeting notes a few minutes
after the meeting. It would helpthem be their second brain and
(34:26):
jog their memory after the fact.But what if Fireflies could
really level up yourconversation while the meeting
is happening. It can serve asthat person that taps you on the
shoulder and guides you when youget stuck on a sales call or
you're interviewing a candidateand you need more context on who
that person is or what the pastinterviews with that person
(34:48):
were.
So we've rolled out what we'recalling Live Assist, where
Fireflies will assist you whilethe meeting is happening. It's
like having someone that servesas auto complete for your
meetings. And we'll talk aboutthe technology part after, but
to that enabled all of this. Butthe core piece of this is as I'm
(35:09):
having a meeting, Fireflies willgive me detailed meeting prep
before we even get into theconversation. Who am I talking
to?
What did we talk about last timewe met like three months ago?
Giving me all of that contextbefore a meeting. And then
during a meeting, giving me cuesand live suggestions while we're
talking about different topics.If it's a general meeting, we
(35:29):
might have talked about sometopic in the past, it's being
mentioned right now. Fireflieswill say, Hey, do you want me to
pull this up and prime you onit?
Let's say you got distracted fora few minutes while the meeting
was happening, checking yourphone or notifications, And
instead of asking the team torepeat themselves, you can press
this button called catch me up,and it will catch you up. If you
wanna ask questions about sometopic you're talking about,
(35:52):
let's say we're talking aboutrockets and, how expensive it is
to build one of these SpaceXrockets. I don't have to switch
tabs anymore. Fred is rightthere and I can just ask the
question. Thanks to ourpartnership with Perplexity, I'm
bringing the power of the webinto the meetings while it's
happening through our LiveAssist.
So our live assist knows fromyour past knowledge, from all of
(36:15):
your past meetings, which is aunique advantage to Fireflies
because if you've had years,months, hundreds, thousands of
meetings on Fireflies, thatsecond brain is now available to
you in real time while themeeting is happening. You will
be like the most knowledgeableperson with perfect memory while
the call is happening, not justafter the call. And then the
(36:35):
power of what's happening inreal time during the meeting
because we get distracted, butFireflies has perfect attention
span during the meeting andremembers everything that's
going on. And it will also giveyou real time notes while the
call is happening, real timetranscripts as the call is
happening so you can refer backto it. So on top of suggestions,
you're getting real time notesand real time transcripts.
(36:59):
And the best part of all of thisis also you're getting the power
of the web available to you atyour fingertips. So that is our
Live Assist product. And thenwe've built different versions
of this. So if you want, ifyou're on a sales call, you can
enable Sales Assist and you canupload all of your sales docs,
FAQs, wikis. And imagine I'm ona very important deal with a
(37:23):
prospect.
They ask me about what is yourenterprise offering? How does
your enterprise offering differfrom XYZ competitor? And you
don't want to get stuck. Youwant to know what to say.
Fireflies will give you realtime suggestions based on all of
your knowledge bases.
Hey, this is how you answer thisquestion. So real time sales
coaching. Or if I'm recruitingand interviewing a candidate and
(37:46):
I want some more context onthings that they've said in
their resume or things thatthey've said in past meetings,
it'll tell me you shouldprobably dive deeper into this
experience. That will besomething that the last
interviewer didn't go into, youshould do that. So making sure
you have more effectivemeetings, hopefully less
repetitive meetings, and ifyou're fully attentive, if every
(38:07):
person could be fully attentivewith a click of a button, that's
what Live Assist is.
And then to add to that, we feltthe best form factor for Live
Assist was going to be through adesktop application. And that's
also a big announcement becausetoday, everyone knows about the
Fireflies meeting bot that joinsyour meetings, the notetaker
(38:29):
bot. We also have customers thatwould like to have a experience
that doesn't involve a bot, andthat has been something that our
customers have requested. So thedesktop app serves a couple
different functions. One isyou'll be able to capture your
meetings, get notes withouthaving a bot.
You'll be able to capturemeetings on platforms beyond
traditional video conferencingplatforms. So a lot of people
(38:51):
have spontaneous meetings onSlack huddles or on Discord or
any platforms where the botusually could not join, you
could do that on top of theZoom, the Teams, the Google
Meets. You also have a muchcleaner, slicker, real time UI
where you can see all of theseLive Assist suggestions in a
panel right then and there. Thenice thing about Fireflies is
(39:13):
you can use Live desktop,whether it's on mobile, whether
it's on web or even our Chromeextension. We're always been
multi platform, but the desktopapp offers this really nice,
extension of our surface areabecause our ultimate goal is
work happens everywhere and it'shappening when you're having
(39:33):
scheduled meetings or in personmeetings.
That's why you use the mobileapp to capture in person
meetings. Or it's happening,well, very impromptu where you
tell a team, let's get on ahuddle and let's have a call. So
we want to be everywhere, whereyou're having these
conversations so that we canhelp you capture that knowledge.
And that's why we're superexcited both with Live Assist
(39:54):
helping you in real time, andthen two, having a desktop
platform where getting the mostout of that experience will be
really seamless.
Chris (40:02):
I I'm curious as you guys have been testing this
internally with the team andeverything, what has like,
you've kinda gone through awhole bunch of kind of
behavioral adjustments and usecases, which which I would have
asked if you hadn't offered themup. And I'm curious as as you
guys have experienced ityourself prior to going to
market here, like what surprisedyou about it in terms of your
(40:25):
own reaction. So there's like,there's the vision that you have
that your team is realizing, asthey're putting the product
together. But when you'reactually using it, what has made
you what has surprised you asthe leader of this team in a way
that maybe wasn't exactly whatyou're expecting, maybe gave you
an extra superpower that youhadn't really counted on. Any
(40:46):
insights there into into yourown moment of kind of wow?
Krish (40:50):
When we're looking at the initial Live Assist data, what
fascinated me, my initialhypothesis was everything will
be based on the suggestions weprovide them. We're gonna be
suggesting things proactively.The proactive suggestions is
where all the magic is gonnahappen. People are engaging with
that. But what's superinteresting is the manual
(41:10):
engagement with Fred on LiveAssist.
Being able to ask manual querieshas shot up even more than what
we had in the past. We thoughtmanual queries would go down
because everyone would just usethe suggested live assist. In
fact, the suggestions are thisgreat fodder for them to
actually, I want to dive deeperinto that topic. So they'll
(41:31):
click on the suggested tile, butthen they'll go ask a bunch of
follow-up questions, even more.So we're seeing increased usage
of Ask Fred and an increasedusage of follow-up questions
because the Live Assist isserving as a great nudge.
And that's a really interestingbehavioral change because we
thought, yeah, maybe they'lllook at one or two suggestions.
It's something passive. But whensomeone opens that panel, the
(41:56):
intensity of usage is a lothigher. And then the
distribution of usage where themanual engagement, like the
manual queries, is equal orsurpassing the automated
suggestions that are happening.So that means the automated
suggestions are doing a good jobof piquing someone's curiosity
to want to dive deeper.
(42:17):
When you see a suggested searchresult on Google, you kind of go
down that rabbit chain. Orsimilarly, when you see what to
watch next on YouTube, you godown that rabbit chain or rabbit
hole. So that's something thatwas super interesting to us
where our suggestions areactually helping people talk to
fireflies more. And this givesme that like Her type, the movie
(42:39):
sort of example where you'rehaving this AI that's like
helping you and it knows andit's learning like, okay, this
is relevant to you. You probablywant to catch up on this topic.
You might not know about thistopic. Do you want me to like
pull this information up? Sobeing able to have your IQ
points jump up by another 10 or20 on a meeting because you now
(43:00):
have perfect memory and perfectawareness and you know about the
context of everything that'sgoing on. It's like having these
super special glasses thatyou're wearing that lets you see
everything. Yeah.
Daniel (43:12):
Well, I'm really excited to try the Live Assist. I think
that's amazing. I've definitelyneeded that assist that I
haven't had in meetings becauseof my own cognitive limitations.
But yeah, I'm wondering, Krish,as you look forward, I mean,
you've had quite a journey thusfar. You've released of course
(43:35):
some amazing stuff, even justthis last week, but as you look
to kind of the future,especially kind of maybe even
from a broader context of wherethe industry is going, how
companies are being influencedby this AI kind of driven
workflows, or maybe specificthings with Fireflies.
(43:56):
What is most exciting for you asyou kind of look to the next
year of things that are openchallenges that you're looking
forward to digging into orthings that are positive and
interesting that you're seeingin terms of how people are using
the technology or where it couldgo? Any thoughts?
Krish (44:16):
We try not to hold super long term road maps. I know that
sounds contrarian. If you'reworking in technology, you have
to have a vision of the future.We believe where the technology
trends are gonna be going. Like,we understand that.
But so much can change in ayear, so much can change in six
months, heck, in like six weeks,so much can change with AI. We
(44:38):
have a general sense of thedirection that we wanna go, but
no like fixed long term plans.You have a plan and then you
make things up as you go. That'show how we do things. But a
couple things that are coming inthe near future that I'm
personally excited about is ourinvolvement in hardware.
When I said that I wantFireflies to be everywhere,
(44:58):
whether you're capturing inperson meetings on your phone
with the mobile app or you're onyour Chrome extension or you
have the meeting bot on the webor the desktop app, we are going
to be announcing somethingreally exciting that hopefully
will be available on 10,000,000devices, sometime next year,
which brings the power ofFireflies to everyday devices
(45:22):
that you probably are alreadyusing on with some well known
brands. So that's something I'mvery excited about. Like I
personally didn't want to getinto hardware. I didn't think
about hardware at this time, butit increases our surface area
tremendously. And whatever wetalked about with that movie Her
or this Ambient AI that's alwaysavailable assisting you, that's
(45:43):
the general trend that I do see,the market going.
So I also believe when you lookat where these LLMs have gotten,
GPD four, like the affordabilityof it has cut down by a 1000x.
So I do believe at some point intime, we will have technology,
really powerful technology, LLMsthat can run on device on edge
(46:07):
20 fourseven, low cost, lowlatency all the time. And I
think that will open upincredible amounts of use cases
for people. So that's a generaltrend I believe in and we're
heading towards. So the hardwareangle is interesting for us.
And then for us as a company, welook at our own processes and
tools that we built internally,and we realize some of these
(46:30):
tools could actually be valuablefor companies beyond Fireflies
because we built a very uniqueset of tools that help us
operate really quickly andexecute really fast with just
100 people. And what I reallybelieve was Fireflies was one of
the first AI agents. We neverused the word agent at the time.
We used bots and stuff. But ifwe can do this sort of value add
(46:54):
for meetings, what other partsof knowledge work can Fireflies
provide knowledge, can providevalue at?
That's something that I'mthinking a lot about. And
hopefully next year, we'll beable to announce a few products
that takes Fireflies well beyondmeetings and that brings this
concept of AI agents and humansworking side by side to reality.
(47:18):
So that's something I'm veryexcited about that we are in the
works on right now. Yeah, thosewill be two big things that I'm
looking forward to for thefuture.
Daniel (47:27):
Cool, well, make sure you shoot us a message, come
back on the show to let us knowhow all of that worked out and
talk about those things that youcan't quite share yet, but sound
very exciting. And, yeah, thankyou for serving as an example,
early example of just reallydigging in and making something
(47:47):
like the AI note taking,something that is actually
bringing value to people's livesthrough AI, which is of course
encouraging and definitely goesbeyond the kind of AI hype or AI
bubble or however you wanna putit to kind of real value and
real revenue. And, yeah, just anamazing example. So thanks for
(48:10):
taking time to join us. Hope totalk to you again soon.
Krish (48:12):
That was a lot of fun. Thank you. Thank you, guys.
Jerod (48:22):
Alright, that's our show for this week. If you haven't
checked out our website, head topracticalai.fm and be sure to
connect with us on LinkedIn, X,or Blue Sky. You'll see us
posting insights related to thelatest AI developments, and we
would love for you to join theconversation. Thanks to our
partner Prediction Guard forproviding operational support
for the show. Check them out atpredictionguard.com.
(48:44):
Also, thanks to BreakmasterCylinder for the beats and to
you for listening. That's allfor now, but you'll hear from us
again next week.
Welcome to the Practical AI podcast, where we break down
the real world applications ofartificial intelligence and how
it's shaping the way we live,work, and create. Our goal is to
help make AI technologypractical, productive, and
accessible to everyone. Whetheryou're a developer, business
leader, or just curious aboutthe tech behind the buzz, you're
(00:24):
in the right place. Be sure toconnect with us on LinkedIn, X,
or Blue Sky to stay up to datewith episode drops, behind the
scenes content, and AI insights.You can learn more at
practicalai.fm.
Now onto the show.
Daniel (00:48):
Welcome to another episode of the Practical AI
Podcast. This is DanielWitenack. I am CEO at Prediction
Guard, and I am joined as alwaysby my cohost, Chris Benson, who
is a principal AI researchengineer at Lockheed Martin. How
are you doing, Chris?
Chris (01:04):
I am doing very well. Got to spend the morning with, with
some folks down at Georgia Techtalking AI. Cool. So
Daniel (01:12):
And you're headed my way. After this, we're gonna
meet up at the the Midwest AISummit, which is as of recording
this happening tomorrow and alsoas recording this still
experiencing government shutdownand and travel problems in The
US. So hopefully I'm hoping youmake it, our our way, but, yeah,
(01:32):
excited for the summit. It'sgonna be gonna be a fun time to
to be together in person andmeet some listeners, meet meet
some AI enthusiasts and AIcurious folks. So, excited for
that.
Chris (01:45):
I'm encouraged because our guest was on a plane,
earlier today and and and he gotthrough. So I think we should
dive right into the conversationhere.
Daniel (01:54):
Yeah. Yeah. Sounds good. I am very excited because we
have with us Krish Ramineni, whois co founder and CEO at
fireflies.ai. Welcome.
Krish (02:03):
Hi, Dan. Hi, Chris. It's great to be here. Thanks for
having me.
Daniel (02:07):
Yeah, it's really great to have you on. I see, of
course, Fireflies join mymeetings all the time, which is
cool, sort of Firefly callassistance and recording and
that sort of thing. Of course,this is probably something like
it was somewhat early on, atleast in the public perception
(02:29):
of ways people were using AIthat kind of impacted maybe
their day to day were these kindof meeting assistants that have
taken various forms and peoplehave tried various things with
these. So as we're sitting herein November 2025, like what is
the kind of, like how would youpick apart that industry from
(02:49):
your perspective? Like what arepeople trying?
What are various approaches tothat? What is the state of that
technology, etcetera?
Krish (02:56):
For me to really appreciate how far we've come as
a company and as an industry,it's good to even look just five
years back and maybe even goback a few more years before
that where we started in 2016.We were working on a whole host
of different tools all aroundthe general AI space. And
(03:17):
remember, this was beforeChadGPT existed. This was before
any LLM or crazy AI tools wereavailable. It was when me and my
co founder were reading articleson deep learning and sequence to
sequence applications and how tomake it work.
We were using crude technologiesin today's standards like BERT
(03:38):
and really manual naturallanguage processing libraries
around that time. So it was avery different time when
everyone wished and hoped thatyou could have AI understand
conversations. And that was theessence of what we were trying
to build. At the time, we didn'teven call it an AI notetaker. We
(03:59):
were trying to build thisgeneral AI secretary or EA, and
we needed to pay our bills.
So we said, we've tried a bunchof different products, in this
space, so we need to figure outwhat's the thing that people are
willing to pay for. And wouldn'tit be a great idea if someone
for a $100 a month could get anexecutive assistant with a human
(04:21):
in the loop and some AI in thebackground. So we went about
building that, testing that out,learned a lot and realized
having a human in the loop isthere's no way this is gonna
scale. If we wanna eventuallybuild this into a platform used
by millions of people every day,taking notes for you twenty four
seven, capturing all thosetranscripts, there was just no
(04:44):
way that human in the loop wasgoing to build, this sort of
company. It was also rightaround the same time companies
like Scale AI were created wherethey were doing data labeling
and the human in the loopprocesses.
So we took inspiration from ourpeers at the time, but our
ambitions were, I think a littletoo grandiose. And we said, do
(05:06):
we really want to build thisbusiness? Like the max we might
be able to support is like 5,000people, 10,000 people, and it's
going to be an operations heavybusiness rather than an AI
business. So that's when me andmy co founder said, you know
what? We validated there's amarket need for this.
And within, like, trying it outfor 10 of our close friends who
(05:28):
were paying us enough money topay for rent in San Francisco at
the time. I felt like they werejust they felt bad for us. My co
founder got a place for $750 inSF. That's probably the biggest
hustle at the time.
Daniel (05:42):
Yeah. Serious.
Krish (05:43):
And we were like, this is the only way to make things work
is we just have to, like, rentour time. So as we built that,
we validated it very quickly,and we did it without writing a
line of code. After we did thatexperiment with, like, 10 of our
friends, we said, if we reallywant to build a serious
business, we know there's aproduct market need here. Before
(06:05):
we would write code for sixmonths, we would ship something
and no one would want to use it.And we would be wasting a lot of
time.
This is the first time wherebefore we built anything or
wrote a line of code, wevalidated the market and we
ourselves were the product,right? So fast forward, we ended
up building the product. Therewas no category for AI notetaker
at the time. And that's when wecreated Fred, the AI notetaker.
(06:30):
We started with simple promisesof, hey, one click capture your
meetings.
You don't have to deal with likenative recording on Zoom because
every user, when they were usinglike, local recordings, there
was a limit to how much theycould record on Zoom and other
platforms. Some platforms likeGoogle Meet didn't even offer an
ability to capture meetings. Sowe started with a basic function
of how can I record my calls soI can go back and play through
(06:53):
them? Then we focused ontranscription, but it was really
expensive at the time and notvery accurate. And we said
that's okay.
Even if transcription is notperfect, as long as I can just
search index through theconversations and just go back
to the general time frame whenwe talked about dates or months
or something some importantkeyword that's good enough.
(07:15):
Right? That's how bad the techwas back then. So we built,
like, a way to record yourmeetings, a way to search back
through your meetings. Then asit got better, we built the
transcription layer that you canactually read.
Initially, there was nodashboard. It was just an email
that we would send out. And thenafterwards, we started building
our own task detection andaction item and keynote
(07:39):
detection engine by hand. Thatwas brutal. So we wrote the code
for that, using some off theshelf stuff and some custom
scripts that I wrote at thetime.
And fast forward, we launch in2020. So it was a four year
struggle of trying to figure outwhat we were trying to do. We
launched in 2020. It's still noChatGPT at around that time. But
(08:01):
the core product of being ableto capture, search, and see some
general bullets or sentencefragments was interesting enough
that we got our first couplepaid customers by the 2020.
So it wasn't until 2021 that weactually started making revenue.
And then fast forward, COVIDhappens, right? And so we get
(08:23):
like this proliferation of freeusers and then some of them
turning into paid users. Andthen you accelerate through to
2022, we get early access toOpenAI's GPT 3.5. Vinod Khosla,
who is an investor, alsohappened to invest in OpenAI.
That really opened thefloodgates for us. Again, LLMs
(08:44):
were very expensive at the time,but we said, you know what?
Let's go ahead and bring thisin. It can't be that bad, right?
Like whatever we're doing today,it's probably going to be
better.
It blew our mind. It changed thecomplete technology. And then
from November 2022, so exactlythree years ago to today, it's
been an absolute rocket ship.We've never looked back. Company
(09:08):
has grown, accelerated, crossedseven figures, and then eight
figures in revenue, and scalingbeyond.
So it's been verytransformational. So I think
it's fair to say we were luckyto be at the right place at the
right time, but the more honestanswer was we just showed up a
little too early and we justtried to survive for, what it's
(09:30):
worth. And eventually the buscame around and we were able to
hop on the bus. So, and thencreate this whole category of AI
note taking. And now it seemslike the most obvious generative
AI use case.
Like if your company is workingon something and you need to
pivot because it failed, thefirst thing you look at is,
(09:50):
maybe we should just buildanother AI notetaker. Seems like
it's working for Fireflies. Weshould try to copy them. So it's
a very different space now.
Chris (09:57):
I'm I'm curious. So that that's a fantastic I love the
history of that, and I love thatthe kind of the incremental
development of both the businessand the technology as you
described it. One of the thingsI'm wondering is like, it seems
like early on, you know, youtalked about that in the
beginning, you would write codefor six months and ship it and
people weren't using it. And youlearn the lesson of of kind of
(10:20):
going and making sure you had amarket before you did a major
commitment on that. And so andas you've talked about the
evolution from that point on, atwhat point did you start to see
users changing their ownbehaviors in response to how
your like, whatever state yourproduct was in at the time.
(10:41):
Like, what were the things thatyou noticed where users would
respond and they would changeand adjust their own behavior
based on the integration of theproduct into their own workflow?
And how did that behavior changeevolve over time as your product
got more and more sophisticatedand capable?
Krish (11:00):
All the other products that we built were easily
buildable, but there wasn't aclear market demand for it. And
we always looked at it from, isthis feasible from an
engineering point of view ratherthan is this something that the
market cares about? Is thissolving a killer pain point for
customers? So when we decided tobuild Fireflies the notetaker,
(11:24):
we said, let's forget what iseven theoretically possible
right now. Let's figure out whatcustomers want and we'll work
backwards from there.
That was a fundamental shift. Sowhen we had this experiment
where we bootstrapped it withjust my co founder and I and a
dozen friends where we weretesting it for them, like we
were the software. We wereeverything at at that moment in
(11:46):
time. And that was before therewas a company really built.
There was no business plan, butit was, is this something that's
valuable enough to do?
You validate it, but it's onething to validate it with humans
and now another thing to buildthe technology. Both my co
founder and I were technical. Weboth like solving hard
(12:06):
engineering problems and creditto Sam, my co founder and CTO,
went to MIT, studiedaeronautical engineering. And
for him, computer science islike the easy stuff, right? The
aeronautical engineering, he wasworking on drones, autonomous
vehicles, unmanned vehicles backin 2015, 2016.
So I think this is like, if Ican solve that problem, how hard
(12:29):
can this machine learningproblem be? Was like a Again, we
got a little carried away, butit was a very hard problem. But
I think a lot of the besttechnologists in the world, they
worry about the technology partafterwards. They're like, great.
If it's a hard problem, we'regonna find a lot of really smart
people that wanna get behind itand solve it.
(12:50):
And we did the best we couldwith the the technology that was
available at the time. So toeven be able to capture calls,
stream them into the cloud,record them, store them, index
them, search them, that itselfwas a hard problem, with the
resources that we had back then.And it was also very expensive
(13:11):
to do all of this stuff.Transcription has gone down like
10 x since that time welaunched. So that's been an
enabling factor for us.
We've scaled our owninfrastructure over the last
five years where we manage ourown bare metal servers. The
volume is insane, right? Wemaybe a great We are a multi
(13:32):
cloud platform, but we run moreof our traffic off of our own
servers than on Google or AWS.So we've gone to this economies
of scale and done all of thishard work where the end user,
they just look at Fireflies andsays, Oh wow, it's $10 per
month. And I'm getting thisnotetaker that's going to join
all my meetings all the time,every single day around the
(13:54):
clock.
Doing that in the past wouldhave cost thousands and
thousands of dollars to support.It doesn't make sense to charge
someone $10 if it's going tocost you $1,000 to support that
user. So the businessinnovation, the scaling itself
was, I think, a miracle. Thatwas like a masterclass in how
the engineering team got behindand solved it. Because not only
(14:16):
you have to optimize for qualityand reliability and uptime, but
you also have to make itaffordable enough where
someone's going to be willing topay $10 a month.
And that that was like a bigfeat. And since 2023, we've been
profitable ever since we'vestarted hyperscaling. So we've
done all of this while beingprofitable, and we didn't touch
(14:37):
any of our series a funding thatwe had raised, along the way. It
was just our seed round. Andthat initial seed round was
enough to, like, seed theinitial free users, figure out
our monetization strategy, buildout our own infrastructure, and
get it to a point where we foundproduct market fit, we found
customers, and we found revenue.
(14:58):
So yeah, that was the differentthing this time around was we
did not stop ourselves or limitour dreams from what was
technologically possible at thetime. And we absolutely got
lucky because we might not havebeen able to reach the success
that we have today had it notbeen for LLMs, but one way or
(15:19):
the other, maybe we would havebeen two years delayed, right?
Eventually this sort oftechnology would come out
because that's how fast theindustry was moving. And again,
the credit was definitely to Samon the forethought because every
time we went to investors, theywould tell us there's no way
transcription is going to beaccurate enough. There's no way
transcription is gonna be cheapenough.
In fact, many investors told us,you can make good money with
(15:41):
this human in the loop business.Why don't you just go turn that
into an actual business? Like,you ran this cool experiment
with your friends. You shouldturn that into a business. And
we were very strongly opposed tothat.
We said, we want to build asoftware company, but we don't
want to deal with operations.And this is also from some
battle scars from the past whenwe tried to build a food
(16:02):
delivery app in college and thelogistics behind it. So we said,
never again are we going to doanything that requires like
logistics and on the groundstuff. So it was a series of
optimistic decisions and thetechnology ended up catching up
to where we were vindicated.
sponsors (16:37):
So friends, you know that feeling when your team has
brilliant ideas and you'reexcited, but these ideas, this
innovation gets stuck behindendless meetings, scattered
documentation, and when you'removing at this speed of
bureaucracy instead of the speedof innovation, we've all been
there. The gap between idea,execution, impact, all that good
(16:59):
stuff, it is killing your team'sprogress. But here's the thing,
simply throwing AI at a problemdoesn't help. It just makes
things messier. That's exactlywhy you should check out Miro.
Miro is powered by AI, and ittransforms teamwork that
normally takes weeks into days.Now I've used Miro to plan
different ideas out, map complexworkflows, and I've even
(17:20):
generated fresh ideas frominterviews all in one single
place. And that is the power ofMiro. It is like having a second
brain with AI that actuallythinks like you, like your team.
No more switching between 10different tools.
Help yourself, help your teamget great things done with Miro.
Check out miro.com to find outhow. That is mir0.com. Again,
(17:44):
miro.com.
Daniel (17:50):
Well, Krish, I'm wondering, you kind of alluded
to the evolution of thetechnology, which of course has
played a key role in that andfor our kind of AI practitioners
and those that are reallycurious about more of the kind
of AI engineering side in ouraudience, could you just kind of
in generic terms, I'm sure youcan't share everything about how
(18:14):
everything works, but you hadmentioned this evolution where
it was kind of at first thesetranscription earlier
transcription models, whichweren't that accurate and maybe
you training or building yourown kind of set of other models
that would do kind of taskspecific things like a action
item detector or whatever thatis. So how does the approach to
(18:39):
something like an AI notetakernow with the models that are
available differ both in termsof what is easier, but maybe
also there's certain things thatare challenging now, which you
didn't have to think aboutbefore because of how the
technology has evolved. Wouldlove to hear that side as well.
Krish (19:00):
If you're using transcription engines today like
Whisper or one of the ASRproviders today and even some of
the big ASR providers from theGoogles of the world, Microsofts
of the world, you don't have tomassage the output as much.
That's exactly the word that wehad to think about because the
raw output was pretty crappy andwe had to fix the grammar, we
(19:21):
had to fix the punctuation, wehad to figure out speaker
identification at the time. Soyou're just getting a garbled up
string of text and you're tryingto make sense of that. And, you
also have to figure out how toidentify certain type of words
because for certain industries,certain work domains, there are
(19:42):
different types of words thatare out there. How do you deal
with filler words, pauses, Like,we speak a lot of filler words.
I feel like working on Fireflieshas improved the way I speak and
helped reduce the number offiller words I I use because the
first versions of Fireflies,every time I would read the ums
and the ahs from my own meetingnotes and recordings and
(20:02):
transcripts, I'd be like, wow, Isaid it that many times. So we
ended up building a littlefilter back in the day that told
me like how many filler words Iused in a meeting. And I would
instantly learn from that andsay, oh, wow, I need to learn to
speak slower. I need to be moreenunciated. So the original text
that was generated fromtranscription engines wasn't
(20:22):
good enough.
So we had to build a bunch ofour own filters on top to clean
up the output that wasgenerated. And then we had to
build other layers on top thatwould classify different parts
of the meetings and then pullout what it thought were action
items, what it thought werebullet point notes, what it
thought were key ideas. And atthe time, it was very
(20:44):
extractive, meaning you're justextracting pieces from this
garbled up text and then you'recalling that notes. And it's a
shame, but at the time that wasgood enough for a lot of people.
So we had to build, rebuild, rearchitect things many times
over.
Whereas if a person was doingsomething today off the shelf,
(21:06):
you don't have to do all ofthat. Like you can get to like
80% pretty solid just using offthe shelf parts, right? And that
is something that we didn't havethe luxury of. Now you may ask,
well, Krish, if it's so easy nowto build something 80%, 85% off
the shelf, what's thecompetitive moat? Everyone
(21:28):
should be able to do this.
And I think the part there thatwe were fortunate enough, one is
going really deep on the problembecause the other 15% actually
takes a long time. Thatdifferentiates good versus
great. And you have to reallypolish the product and those
things take a very, very longtime. And you have to learn from
(21:49):
your users. The roadmap thatwe've been able to build out
beyond the core tech for teams,for enterprises, That takes a
lot of work.
You have to think about accesscontrols, privacy, storage. How
do you think about multi tenant?Like we're one of the first
companies to get SOC twocompliance in the AI meeting
assistance space. And then HIPAAfor doctors that want to use
(22:12):
Fireflies. We offer this optionfor enterprises called private
storage where you can storeFireflies meetings inside your
own server storage containers,which was another architectural
change.
If you don't architect itcorrectly in the beginning, it
makes it very, very difficult.All this administrative sharing
(22:32):
features, team features, whichmakes Fireflies more valuable as
more and more people inside yourorganization start using it.
Search was a really big problemthat you have to solve over time
as well. So all of this isenabled by great transcription
and being able to understandlanguage. I absolutely agree
with that.
But then all the use cases youhave to build on top of it, like
(22:54):
our Ask Fred example, where youcan talk, instead of reviewing
the notes or the meeting, youcan just ask Fred questions
about the meeting and it willcatch you up on everything that
happened, right? So thatrequired a lot of stuff like
figuring out search from scratchand making that really
effective. So, yeah, I I thinkit it was a big journey in terms
(23:15):
of building out, like, thesecore building blocks. But as we
were able to do that, when westarted, it was a blue ocean.
And going back to I'm like Pogosticking back between the
commercials and technology partbecause both are really
important.
If you wanna be a great artist,you don't have to think about
the commercialization of whatyou're building. You can just
(23:36):
think about great art. But tobuild a great business, you have
to be a good artist and you alsohave to figure out, can someone
pay me money for this? We werefortunate enough at the time to
establish the AI Notetaker brandand be one of the first
companies to start championingthat word. And now it's an
entire category.
(23:56):
Build that AI notetaker, get todistribution, get to millions of
people using Fireflies. Today,like tens of millions of people
get notes every month fromFireflies. That in itself,
right, distribution is somethingthat is super important. And
because we started early when itwas less obvious, we were able
to get make the most of it.Like, there's no point in going
(24:18):
to the gold rush after it's beenannounced and after someone's
made a a ton of money off of itor or a killing off of it.
I think, like, being there alittle early helped us maximize
the impact because for every endnotetaker, every additional
notetaker that comes out, itbecomes progressively harder for
them to stand out in the crowd.And there is something to be
(24:42):
said about these markets wheredistribution is one of the most
important things to build in aPLG flywheel. That's why you
don't see so many Calendlycompetitors. People still use
Calendly. Like, it's one of thede facto platforms that everyone
uses.
In our space, there's probably,like, three or so big players
with Fireflies being one ofthem, and then each has its own
(25:04):
merits. Fireflies, for example,is very much focused on teams
and businesses, and if you needrobust integrations and
workflows and admin controls,you come to Firefly. Whereas if
you're looking for more of aprosumer type product, you'll go
to one of the other platforms.So we had to pick what we wanted
to do and go really deep inthere. A good parallel I like to
(25:27):
use also is like the projectmanagement space or the CRM
space, where you had Salesforcebuild a massive enterprise scale
business, and then HubSpot comesalong and takes a huge market
out of the SMB business.
Both are tens of billions ofdollars in market value that's
been captured. In projectmanagement, you have the Asanas
of the world, you have themonday.coms of the world, you
(25:49):
have all of these other projectmanagement systems that follow.
So we've found our niche. Iwouldn't even say it's a niche
because our niche is anyonethat's a knowledge worker that
works inside of a team. So wespecifically say no to consumer
grade use cases like universitystudents teaching, like those
sorts of stuff.
(26:10):
Like our bread and butter hasalways been like, if you're a
team and it doesn't have to bein tech either, like a lot of
our customers are outside oftech, which was also really
fascinating to see. But yeah,it's been a very long tedious
journey. I I wouldn't havebelieved three years ago we'd be
where we are today. But toanswer your question on tech,
(26:31):
yeah, it it's it's been one ofthe most fascinating things
about this space because now itfeels super easy and everyone
can do it.
Chris (26:37):
You hit on so many things there that I that I'm that I'm
interested in. Like, one of thethings I think that I think I
just learned something from youdefinitely on and that it did I
will probably try to share andI'm kind of I'd like to
generalize it towards notstrictly a fireflies thing
because I think there's a lot ofpeople in our audience that can
learn from this in whateverindustry they're in is is your
(27:00):
kind of your pursuit of asustainable competitive
advantage and what that meant toyou. And it sounds like, in your
case, being there early andhaving to solve, you know, the
whole problem and not, know,unlike people that would come in
later and get to that 80%easily, and you have that last
15 or 20%, which is which isreally hard, you were well
(27:21):
positioned to develop theexpertise in your organization,
and to find the niches that youwanted to service and get their
first and best with thatexpertise to do that so that
even as even as the space hasdeveloped other competitors, you
were able to to hold them offand hold your niche and be a
(27:43):
powerhouse in that way.
Is that is that a fairrepresentation of kind of what
you were just saying?
Krish (27:49):
Yeah. This was one of those, situations where being
early definitely helped, andthen having the time to keep
building and refining it andlistening to your users over
time. A lot of people claim thatSaaS is dead. Everyone will
build their own SaaS. But inreality, that extra 15 to 20% of
work you have to do to build it,maintain it, customize it, I
(28:11):
don't think most companies wantto deal with that, right?
They have other they have abusiness to run. They don't want
to be building these tools inhouse. So when we look at our
customers and look at the thingsthat we want to offer to them,
we have 95 plus differentintegrations, for example. That
was a competitive moat for usover time because we wanted to
be the most integrated AImeeting assistant on the
(28:32):
platform, on the market. We alsowanted to really take the
security thing and be the mostsecure AI notetaker for people.
So that means like when you workwith enterprises and businesses,
they have thousandquestionnaires on like all the
security compliance stuff. Youhave to deal with their CIOs.
You have to like build all ofthese like layers and features
like audit logs and all thesecompliance features, that is a
(28:56):
pain to do, right? Forget aboutthe AI, but just building the
SaaS part of it. It's like usergroups, admin controls, data
retention.
Some customers in the financeindustry want their data wiped
every seven days. So buildingall of these sort of things,
help definitely build the moat,but we're shipping ten, twenty
different features orenhancements every week, and you
(29:18):
do that over a span of fiveyears, it compounds. That's why
I feel like whenever a bigplayer gets into a space and
someone says, oh, this companyis dead now, but what ends up
happening is the other companyhas built so much that it's very
hard for you to wanna switchover. And a company that I like
(29:39):
to reference, and I thinkthey're doing a really great job
of it, is Eleven Labs, like,with text to speech. They went
so deep on a problem where manypeople would say, this is
generic.
Like, anyone can do text tospeech now. Like OpenAI or one
of the big players will just dotext to speech and offer it as
an API. But if you look at theamount of stuff that Eleven Labs
(30:00):
provides to their customers andgoing really deep on what your
users need, it's an all in onevery comprehensive platform and
it gives you so much to choosefrom. Sure, you might lose out
on the 10 or 20% of prosumertype users if other players come
in and solve the general usecase, but that's the whole point
of the company, right? Productmarket fit is not about finding
(30:23):
it once and you're good forlife.
You have to keep finding productmarket fit over and over again,
and you have to keep adapting asthe market changes, your company
has to change. Otherwise youbecome stale. Like the reason
you have CRM companies that werein the 90s versus in the 2000s
versus now, they all fell out ofproduct market fit and had to
(30:44):
keep innovating and keep gettingback into product market fit. So
yeah, being early helpful,listening to customers and
having five years to build outthe roadmap, super helpful. And
I'm one of those people thatdefinitely believes SaaS is not
dead and most people don't wantto build their own tools.
Trust me, it's not worth thehassle. You got bigger and
(31:05):
better things to do.
sponsors (31:21):
If your team is still jumping between design tools
just to update the website,that's that's not fun. Maybe
you're stuck behind expensivepaywalls for basic design
features. Here's the thing, mostdesign tools make you pay just
to get started. That's notFramer's way. Our sponsor today
is Framer.
They already built the fastestway to publish beautiful
(31:42):
production ready websites and isnow redefining how we design for
the web with the recent launchof design pages, a free canvas
based design tool. Framer ismore than a site builder. It's a
true all in one design platformfrom social media assets to
campaign visuals to vectors andicons all the way to a live
(32:02):
site. Framer is where ideas goto live, start to finish. Here's
what I love about this tool.
It is an all in one tool. Youdon't have to switch to Figma
for design and then have thesemessy HTML exports. You don't
have separate tools forgraphics. With Framers free
tier, you get unlimitedprojects, unlimited pages, and
(32:24):
you even get unlimitedcollaborators, vector tools,
three d transforms, animations.It's all there.
That's what sets Framer apart.It's not just another site
builder like Webflow or Wix. Itis a complete design tool that
happens to publish productionready websites. It's everything
you need. It's totally free.
So if your team is ready todesign, iterate, and publish all
(32:46):
in a single tool, start creatingfor free today at
framer.com/design and use thecode practical a I and get a
free month of Framer Pro. That'sframer.com/design and use the
promo code practical a I. That'sall in word. Framer.com/design,
promo code practical a I. Rulesand restrictions may apply, but
(33:10):
check it out today.
Daniel (33:16):
Well, Krish, I definitely wanna shift now to
some of the things that are, ofcourse really exciting that you
have either just released orcoming on the roadmap with
Fireflies and kind of some ofwhat's enabled that from the
technology standpoint, some ofthe challenges related to that,
some of the value that thatbrings. One of the big things
that we were talking aboutbefore we hit the record button
(33:38):
was real time functionality. Socould you describe a little bit
of what is coming out inrelation to real time and also
kind of the why of that, likewhy people would want that, what
it enables, and maybe why thathasn't come out yet in terms of
(33:59):
the technology side of things.What has enabled that at this
point in time?
Krish (34:05):
One of the biggest things that I always found when I talk
to customers was how can we helpthem in the moment while the
meeting itself is happening?Today, people were using
Fireflies where they would getmeeting notes a few minutes
after the meeting. It would helpthem be their second brain and
(34:26):
jog their memory after the fact.But what if Fireflies could
really level up yourconversation while the meeting
is happening. It can serve asthat person that taps you on the
shoulder and guides you when youget stuck on a sales call or
you're interviewing a candidateand you need more context on who
that person is or what the pastinterviews with that person
(34:48):
were.
So we've rolled out what we'recalling Live Assist, where
Fireflies will assist you whilethe meeting is happening. It's
like having someone that servesas auto complete for your
meetings. And we'll talk aboutthe technology part after, but
to that enabled all of this. Butthe core piece of this is as I'm
(35:09):
having a meeting, Fireflies willgive me detailed meeting prep
before we even get into theconversation. Who am I talking
to?
What did we talk about last timewe met like three months ago?
Giving me all of that contextbefore a meeting. And then
during a meeting, giving me cuesand live suggestions while we're
talking about different topics.If it's a general meeting, we
(35:29):
might have talked about sometopic in the past, it's being
mentioned right now. Fireflieswill say, Hey, do you want me to
pull this up and prime you onit?
Let's say you got distracted fora few minutes while the meeting
was happening, checking yourphone or notifications, And
instead of asking the team torepeat themselves, you can press
this button called catch me up,and it will catch you up. If you
wanna ask questions about sometopic you're talking about,
(35:52):
let's say we're talking aboutrockets and, how expensive it is
to build one of these SpaceXrockets. I don't have to switch
tabs anymore. Fred is rightthere and I can just ask the
question. Thanks to ourpartnership with Perplexity, I'm
bringing the power of the webinto the meetings while it's
happening through our LiveAssist.
So our live assist knows fromyour past knowledge, from all of
(36:15):
your past meetings, which is aunique advantage to Fireflies
because if you've had years,months, hundreds, thousands of
meetings on Fireflies, thatsecond brain is now available to
you in real time while themeeting is happening. You will
be like the most knowledgeableperson with perfect memory while
the call is happening, not justafter the call. And then the
(36:35):
power of what's happening inreal time during the meeting
because we get distracted, butFireflies has perfect attention
span during the meeting andremembers everything that's
going on. And it will also giveyou real time notes while the
call is happening, real timetranscripts as the call is
happening so you can refer backto it. So on top of suggestions,
you're getting real time notesand real time transcripts.
(36:59):
And the best part of all of thisis also you're getting the power
of the web available to you atyour fingertips. So that is our
Live Assist product. And thenwe've built different versions
of this. So if you want, ifyou're on a sales call, you can
enable Sales Assist and you canupload all of your sales docs,
FAQs, wikis. And imagine I'm ona very important deal with a
(37:23):
prospect.
They ask me about what is yourenterprise offering? How does
your enterprise offering differfrom XYZ competitor? And you
don't want to get stuck. Youwant to know what to say.
Fireflies will give you realtime suggestions based on all of
your knowledge bases.
Hey, this is how you answer thisquestion. So real time sales
coaching. Or if I'm recruitingand interviewing a candidate and
(37:46):
I want some more context onthings that they've said in
their resume or things thatthey've said in past meetings,
it'll tell me you shouldprobably dive deeper into this
experience. That will besomething that the last
interviewer didn't go into, youshould do that. So making sure
you have more effectivemeetings, hopefully less
repetitive meetings, and ifyou're fully attentive, if every
(38:07):
person could be fully attentivewith a click of a button, that's
what Live Assist is.
And then to add to that, we feltthe best form factor for Live
Assist was going to be through adesktop application. And that's
also a big announcement becausetoday, everyone knows about the
Fireflies meeting bot that joinsyour meetings, the notetaker
(38:29):
bot. We also have customers thatwould like to have a experience
that doesn't involve a bot, andthat has been something that our
customers have requested. So thedesktop app serves a couple
different functions. One isyou'll be able to capture your
meetings, get notes withouthaving a bot.
You'll be able to capturemeetings on platforms beyond
traditional video conferencingplatforms. So a lot of people
(38:51):
have spontaneous meetings onSlack huddles or on Discord or
any platforms where the botusually could not join, you
could do that on top of theZoom, the Teams, the Google
Meets. You also have a muchcleaner, slicker, real time UI
where you can see all of theseLive Assist suggestions in a
panel right then and there. Thenice thing about Fireflies is
(39:13):
you can use Live desktop,whether it's on mobile, whether
it's on web or even our Chromeextension. We're always been
multi platform, but the desktopapp offers this really nice,
extension of our surface areabecause our ultimate goal is
work happens everywhere and it'shappening when you're having
(39:33):
scheduled meetings or in personmeetings.
That's why you use the mobileapp to capture in person
meetings. Or it's happening,well, very impromptu where you
tell a team, let's get on ahuddle and let's have a call. So
we want to be everywhere, whereyou're having these
conversations so that we canhelp you capture that knowledge.
And that's why we're superexcited both with Live Assist
(39:54):
helping you in real time, andthen two, having a desktop
platform where getting the mostout of that experience will be
really seamless.
Chris (40:02):
I I'm curious as you guys have been testing this
internally with the team andeverything, what has like,
you've kinda gone through awhole bunch of kind of
behavioral adjustments and usecases, which which I would have
asked if you hadn't offered themup. And I'm curious as as you
guys have experienced ityourself prior to going to
market here, like what surprisedyou about it in terms of your
(40:25):
own reaction. So there's like,there's the vision that you have
that your team is realizing, asthey're putting the product
together. But when you'reactually using it, what has made
you what has surprised you asthe leader of this team in a way
that maybe wasn't exactly whatyou're expecting, maybe gave you
an extra superpower that youhadn't really counted on. Any
(40:46):
insights there into into yourown moment of kind of wow?
Krish (40:50):
When we're looking at the initial Live Assist data, what
fascinated me, my initialhypothesis was everything will
be based on the suggestions weprovide them. We're gonna be
suggesting things proactively.The proactive suggestions is
where all the magic is gonnahappen. People are engaging with
that. But what's superinteresting is the manual
(41:10):
engagement with Fred on LiveAssist.
Being able to ask manual querieshas shot up even more than what
we had in the past. We thoughtmanual queries would go down
because everyone would just usethe suggested live assist. In
fact, the suggestions are thisgreat fodder for them to
actually, I want to dive deeperinto that topic. So they'll
(41:31):
click on the suggested tile, butthen they'll go ask a bunch of
follow-up questions, even more.So we're seeing increased usage
of Ask Fred and an increasedusage of follow-up questions
because the Live Assist isserving as a great nudge.
And that's a really interestingbehavioral change because we
thought, yeah, maybe they'lllook at one or two suggestions.
It's something passive. But whensomeone opens that panel, the
(41:56):
intensity of usage is a lothigher. And then the
distribution of usage where themanual engagement, like the
manual queries, is equal orsurpassing the automated
suggestions that are happening.So that means the automated
suggestions are doing a good jobof piquing someone's curiosity
to want to dive deeper.
(42:17):
When you see a suggested searchresult on Google, you kind of go
down that rabbit chain. Orsimilarly, when you see what to
watch next on YouTube, you godown that rabbit chain or rabbit
hole. So that's something thatwas super interesting to us
where our suggestions areactually helping people talk to
fireflies more. And this givesme that like Her type, the movie
(42:39):
sort of example where you'rehaving this AI that's like
helping you and it knows andit's learning like, okay, this
is relevant to you. You probablywant to catch up on this topic.
You might not know about thistopic. Do you want me to like
pull this information up? Sobeing able to have your IQ
points jump up by another 10 or20 on a meeting because you now
(43:00):
have perfect memory and perfectawareness and you know about the
context of everything that'sgoing on. It's like having these
super special glasses thatyou're wearing that lets you see
everything. Yeah.
Daniel (43:12):
Well, I'm really excited to try the Live Assist. I think
that's amazing. I've definitelyneeded that assist that I
haven't had in meetings becauseof my own cognitive limitations.
But yeah, I'm wondering, Krish,as you look forward, I mean,
you've had quite a journey thusfar. You've released of course
(43:35):
some amazing stuff, even justthis last week, but as you look
to kind of the future,especially kind of maybe even
from a broader context of wherethe industry is going, how
companies are being influencedby this AI kind of driven
workflows, or maybe specificthings with Fireflies.
(43:56):
What is most exciting for you asyou kind of look to the next
year of things that are openchallenges that you're looking
forward to digging into orthings that are positive and
interesting that you're seeingin terms of how people are using
the technology or where it couldgo? Any thoughts?
Krish (44:16):
We try not to hold super long term road maps. I know that
sounds contrarian. If you'reworking in technology, you have
to have a vision of the future.We believe where the technology
trends are gonna be going. Like,we understand that.
But so much can change in ayear, so much can change in six
months, heck, in like six weeks,so much can change with AI. We
(44:38):
have a general sense of thedirection that we wanna go, but
no like fixed long term plans.You have a plan and then you
make things up as you go. That'show how we do things. But a
couple things that are coming inthe near future that I'm
personally excited about is ourinvolvement in hardware.
When I said that I wantFireflies to be everywhere,
(44:58):
whether you're capturing inperson meetings on your phone
with the mobile app or you're onyour Chrome extension or you
have the meeting bot on the webor the desktop app, we are going
to be announcing somethingreally exciting that hopefully
will be available on 10,000,000devices, sometime next year,
which brings the power ofFireflies to everyday devices
(45:22):
that you probably are alreadyusing on with some well known
brands. So that's something I'mvery excited about. Like I
personally didn't want to getinto hardware. I didn't think
about hardware at this time, butit increases our surface area
tremendously. And whatever wetalked about with that movie Her
or this Ambient AI that's alwaysavailable assisting you, that's
(45:43):
the general trend that I do see,the market going.
So I also believe when you lookat where these LLMs have gotten,
GPD four, like the affordabilityof it has cut down by a 1000x.
So I do believe at some point intime, we will have technology,
really powerful technology, LLMsthat can run on device on edge
(46:07):
20 fourseven, low cost, lowlatency all the time. And I
think that will open upincredible amounts of use cases
for people. So that's a generaltrend I believe in and we're
heading towards. So the hardwareangle is interesting for us.
And then for us as a company, welook at our own processes and
tools that we built internally,and we realize some of these
(46:30):
tools could actually be valuablefor companies beyond Fireflies
because we built a very uniqueset of tools that help us
operate really quickly andexecute really fast with just
100 people. And what I reallybelieve was Fireflies was one of
the first AI agents. We neverused the word agent at the time.
We used bots and stuff. But ifwe can do this sort of value add
(46:54):
for meetings, what other partsof knowledge work can Fireflies
provide knowledge, can providevalue at?
That's something that I'mthinking a lot about. And
hopefully next year, we'll beable to announce a few products
that takes Fireflies well beyondmeetings and that brings this
concept of AI agents and humansworking side by side to reality.
(47:18):
So that's something I'm veryexcited about that we are in the
works on right now. Yeah, thosewill be two big things that I'm
looking forward to for thefuture.
Daniel (47:27):
Cool, well, make sure you shoot us a message, come
back on the show to let us knowhow all of that worked out and
talk about those things that youcan't quite share yet, but sound
very exciting. And, yeah, thankyou for serving as an example,
early example of just reallydigging in and making something
(47:47):
like the AI note taking,something that is actually
bringing value to people's livesthrough AI, which is of course
encouraging and definitely goesbeyond the kind of AI hype or AI
bubble or however you wanna putit to kind of real value and
real revenue. And, yeah, just anamazing example. So thanks for
(48:10):
taking time to join us. Hope totalk to you again soon.
Krish (48:12):
That was a lot of fun. Thank you. Thank you, guys.
Jerod (48:22):
Alright, that's our show for this week. If you haven't
checked out our website, head topracticalai.fm and be sure to
connect with us on LinkedIn, X,or Blue Sky. You'll see us
posting insights related to thelatest AI developments, and we
would love for you to join theconversation. Thanks to our
partner Prediction Guard forproviding operational support
for the show. Check them out atpredictionguard.com.
(48:44):
Also, thanks to BreakmasterCylinder for the beats and to
you for listening. That's allfor now, but you'll hear from us
again next week.
Popular Podcasts
Las Culturistas with Matt Rogers and Bowen Yang
Ding dong! Join your culture consultants, Matt Rogers and Bowen Yang, on an unforgettable journey into the beating heart of CULTURE. Alongside sizzling special guests, they GET INTO the hottest pop-culture moments of the day and the formative cultural experiences that turned them into Culturistas. Produced by the Big Money Players Network and iHeartRadio.
Crime Junkie
Does hearing about a true crime case always leave you scouring the internet for the truth behind the story? Dive into your next mystery with Crime Junkie. Every Monday, join your host Ashley Flowers as she unravels all the details of infamous and underreported true crime cases with her best friend Brit Prawat. From cold cases to missing persons and heroes in our community who seek justice, Crime Junkie is your destination for theories and stories you won’t hear anywhere else. Whether you're a seasoned true crime enthusiast or new to the genre, you'll find yourself on the edge of your seat awaiting a new episode every Monday. If you can never get enough true crime... Congratulations, you’ve found your people. Follow to join a community of Crime Junkies! Crime Junkie is presented by audiochuck Media Company.
Stuff You Should Know
If you've ever wanted to know about champagne, satanism, the Stonewall Uprising, chaos theory, LSD, El Nino, true crime and Rosa Parks, then look no further. Josh and Chuck have you covered.