Latest / Private Equity Data Guy / Why Private Equity Operators Still Cannot Find Their Real EBITDA

Transcript

Why Private Equity Operators Still Cannot Find Their Real EBITDA · Private Equity Data Guy

• 0:00
  If AI was not there, then what I have achieved in the last one and a half years, it would have taken me five years to achieve it in some other time.
• 0:08
  So I can very easily say that.
• 0:11
  And what AI has also allowed us to do is to be like, as a founder, I'm very close to the product.
• 0:19
  The agent is really your reimagined process.
• 0:23
  So you have this current process that has all of these issues.
• 0:27
  The agent is really the reimagined process in today's world, the way we are building it today.
• 0:33
  What I also observed was all of this gap around data silos, broken processes, all of these conflicting business objectives that impacts the overall transformation goals that the organization has.
• 0:50
  So that is the learning that brought me into this thesis that AI adoption is not AI transformation.
• 0:58
  Behind every value creation plan, there's a data problem nobody wants to talk about.
• 1:02
  Fragmented systems, metrics nobody trusts, and decisions made on gut feel dressed up as analysis.
• 1:07
  Welcome to the PE Data Guy.
• 1:09
  Each week, host Graham Crawford talks to the operating partners, advisors and practitioners who are doing the work inside portfolio companies.
• 1:16
  If you care about what actually drives returns in the market, then you're in the right place.
• 1:20
  The PE Data Guy starts now.
• 1:24
  Afrozi Ara is the founder and CEO of Luminadata, an AI transformation platform for finance operations backed by techstars and Barclays Skydeck.
• 1:33
  Her agents run the workflows that most people never see.
• 1:36
  But every CFO bleeds on reconciliations, invoicing, cash application, lease accounting, powered by a patent pending language model trained on gap socks and lease accounting standards rather than on the open Internet.
• 1:48
  But what makes a Frozen interesting for this audience is what she did before.
• 1:51
  She spent over a decade in enterprise consulting, advising Fortune 500 companies at MU Sigma and building out diagnostic practices, benchmarking how finance work actually happens inside large organizations.
• 2:03
  She's an engineer by training, a diagnostic person by trade, and her phoning thesis is one that I'd frame on my wall.
• 2:09
  AI adoption is not AI transformation.
• 2:12
  Today we get into whether EBITDA actually, actually comes from Afrozi.
• 2:16
  Welcome to the show.
• 2:17
  It's so great to have you on.
• 2:18
  Hello, Prem.
• 2:19
  Thanks for having me.
• 2:20
  It's wonderful to be here.
• 2:22
  Great.
• 2:22
  It's so good to have you.
• 2:23
  That's such a wonderful and varied career that you've got and, you know, similar to me, you've gone from Fortune 500 to doing your own thing.
• 2:30
  What's your, what's your story behind that?
• 2:32
  And, and how did you come to, you know, strike out on your own so.
• 2:36
  Well, I started my Career as an engineer.
• 2:39
  So, so that's, you know, as an engineering mindset before it was cool, right?
• 2:45
  Same as me.
• 2:48
  So I studied electrical engineering.
• 2:50
  Actually that wasn't even like computer science engineering.
• 2:54
  I was very keen to learn about computer science even during my electrical bachelor's degree.
• 3:00
  So I studied electrical engineering.
• 3:02
  But then there is this big technology boom that's came up and you know, everyone was getting into computer science.
• 3:10
  So then I got into a technology company, start couple of years, worked on building software for printers, then moved into building software on the product engineering side, got into consulting, worked with consulting companies like Mu Sigma, so work
• 3:29
  with Fortune 500.
• 3:30
  And there I, I saw this whole problem around transformation.
• 3:35
  So there is this whole data AI aspect of it, where there is so much of value that's currently that can be leveraged within data.
• 3:46
  And then working with these Fortune 500 companies, working with the executives and helping them get insights from the data.
• 3:52
  So that was the starting point.
• 3:54
  But what I also observed was all of this gap around data silos, broken processes, all of these conflicting business objectives that impacts the overall transformation goals that the organization has.
• 4:11
  So that is the learning that brought me into this thesis that AI adoption is not AI transformation.
• 4:19
  So seeing all of that in the industry, I decided to go out on my own.
• 4:24
  And then around a year and a half I started this company.
• 4:28
  And we started with like a very simple thesis, like a clear problem around reconciliations.
• 4:34
  So especially on the finance transformation side.
• 4:37
  Now there is so much of time that gets spent in the reconciliations at various levels.
• 4:42
  And we started with building an AI agent around that it hit the market.
• 4:48
  We saw quite a bit of traction and picked up.
• 4:52
  But then the problem kept coming back to the transformation problem that you have fixed this small piece of the workflow, but what about the entire workflow?
• 5:02
  So in the end, it is the entire workflow that has to become better.
• 5:06
  And we went back to that same problem that I solved with enterprise consulting, which is helping identify the gaps, kind of gap analysis, defining the future state, building the roadmap.
• 5:20
  And that's what we productized and built through our company.
• 5:24
  I love it.
• 5:25
  I love it.
• 5:25
  And you know, the start in electrical engineering is interesting because that's what all the cool kids are doing apparently is not going to college and going to study to be electricians instead.
• 5:33
  So it's amazing that you've always got that fallback if, you know, if AI eats the world and no code and consulting is required anymore.
• 5:41
  You know, like we read about every day how consultancy and SEO and SaaS and software engineering is all dead.
• 5:47
  If the doom mongers are ever correct, then, you know, no problems for you.
• 5:50
  You can just start an electrician business with your electrical engineering.
• 5:54
  Yes, I can fix my fridge.
• 5:57
  There you go.
• 5:58
  There you go.
• 5:58
  Do you have a screen on your fridge or not?
• 6:01
  Not yet.
• 6:02
  I think that's one of the places in my home that doesn't have a screen yet.
• 6:06
  I think what one of my friends, my wife told me about this, one of her friends actually has a camera inside their fridge and it's not for like ordering, shopping or whatever else, but when they're in the grocery
• 6:17
  store they can like open the app on their phone and look inside the fridge to see what's already there.
• 6:20
  Like that would have saved me a fortune.
• 6:22
  I think that.
• 6:23
  But anyway, anyway, you know, I love looking at systems and seeing what goes in and what comes out.
• 6:28
  And I think, you know, we've got, we've got similar backgrounds, but you know, what is it that you think, Drew, you should like engineering in the first place and understanding systems and how they work.
• 6:36
  Yeah.
• 6:36
  So engineering forces you to be like very methodical in terms of how you solve problems.
• 6:43
  Right.
• 6:44
  So my training from engineering and actually when I go back and think about the electrical engineering, that work that I did, I don't remember any of the theory, you know, like, how does the PNP or the NPN transistor
• 6:58
  work?
• 6:59
  Like I can actually now go and search and figure that out.
• 7:03
  But the learning is around how to focus your thinking and figure out the hard stuff.
• 7:10
  I think that's what makes engineering special.
• 7:13
  The whole problem solving approach that you take this problem, break it down into components and then how do you solve the components and then you assemble the solution together.
• 7:23
  I think that's the training that is very common whether you do electrical or whether you build computers or whether you even build a company.
• 7:31
  So, so all those learnings translate.
• 7:35
  So building your own company or solving for your own customers problems.
• 7:40
  Because see, there are always problems in the world and then they are at different levels of complexity.
• 7:48
  But being able to take a problem, break it down, figure out what are the components of the problem, what is easy versus what is harder to fix, how do you prioritize and sequence that fixing of the problems?
• 8:03
  It's just like writing a program or building any product.
• 8:06
  So I think those are the learnings that have translated in my life so far.
• 8:12
  Graph objects and functions Right.
• 8:15
  Let's talk about AI, because I think your LinkedIn profile has got one of my favorite quotes we hear on all the earnings calls WePilot include.
• 8:22
  We've deployed copilot to every employee we're partnering with, anthropic, whatever it might be.
• 8:27
  And then your punchline, the tools arrived that EBITDA didn't.
• 8:30
  And I think there's a very often quoted MIT study about 95% of Enterprise AI pilots delivering no P and L impact whatsoever.
• 8:38
  And this is one of the thesis that you push on Luminadata.
• 8:42
  So, you know, you've seen this movie before.
• 8:45
  What's actually happening inside those companies that check all the boxes, they deploy code, they roll out copilot to everyone, they put the big announcement they're going to be.
• 8:54
  They might even add AI to the company name.
• 8:56
  I've seen that.
• 8:56
  I've seen people do that.
• 8:57
  But when the value isn't captured, what's actually happening in there?
• 9:01
  So that is fascinating.
• 9:03
  And I have also seen it in bits and pieces.
• 9:06
  So the value is really somewhere.
• 9:09
  So what happens is when you are deploying this copilot or you are deploying Claude, you're always kind of solving a part of the problem, but you're letting one person take a product and then take it to the next
• 9:25
  level, take their work to the next level.
• 9:27
  But organizations are complex.
• 9:30
  They consist of a group of people together.
• 9:33
  So it's like there are teams which are kind of tribes of their own.
• 9:37
  They have their own belief systems and the way they work, policies and procedures.
• 9:42
  So you can have one person work at 100x the speed, but then the speed of any process is actually the speed of the slowest part of that process.
• 9:53
  The bottleneck is what really governs the speed of any process.
• 9:57
  Right?
• 9:57
  So that's the typical goal theory, philosophy.
• 10:01
  So that's exactly what's happening with AI adoption.
• 10:04
  You give people the tools to work faster, to produce 10x more output, 100x more output.
• 10:11
  But what we have not been enabling is that entire infrastructure around making that happen consistently across the organization.
• 10:21
  And that infrastructure is not just technology.
• 10:24
  It is your processes, it is the people who work.
• 10:27
  It is a common alignment among the folks in doing the work.
• 10:32
  I think that is the biggest gap.
• 10:34
  I was talking to one of my customers at Luminadata.
• 10:39
  That's kind of like the core thesis, right, that, you know, we bring in coworkers that will stitch your systems and get the job done for you.
• 10:46
  So today there's so much of this manual work that's happening.
• 10:49
  Our coworkers understand the context and stitch it together.
• 10:54
  And, you know, the best proof of what you're doing right is your customers.
• 10:58
  So I did ask them, like, why would you work with us and not use Claude in the work that you do today?
• 11:04
  And what he told me was pretty insightful.
• 11:08
  He said, you know Afrozi, I can get this work done with Claude, but then if I give Claude to 10 people, 10 people will come with 10 different answers in 10 different directions.
• 11:20
  And there is like no common framework or structure or problem solving methodology that is tying everything together or that is aligned towards the goals.
• 11:32
  And the second thing is Claude is not going to challenge me.
• 11:35
  Claude is going to just keep.
• 11:36
  I have to be able to challenge and ask the right questions to Claude to be able to get the answer.
• 11:42
  So I think those are the two things, like the whole organization as a complex unit which AI adoption doesn't solve for.
• 11:51
  I think that's where the opportunity is, Graham.
• 11:54
  And I've seen it too, like, oh, I turned on Copilot for everyone therefore.
• 11:58
  And I'm just going to wait for all the efficiency gains to roll in.
• 12:00
  And in fact you can actually make efficiency losses because as you say, if all the spiders go and weave their own web, then you just make an organization more complicated rather than less complicated.
• 12:13
  And you mentioned going after the constraint of a process.
• 12:18
  And there's another good thing in process engineering.
• 12:20
  You should always simplify something before you automate it, not make it more complicated.
• 12:25
  Let's not make something more complicated than automated.
• 12:27
  Let's simplify it and understand why we're doing everything and then we'll automate the thing, things that actually make a difference and, and kind of make sense with that.
• 12:33
  So yeah, I've, I've, I've totally seen that too.
• 12:36
  Is there any type of like, say you've got a standard kind of business, let's say a mid market business that you're, that you're working with and you got carte blanche and they just threw the doors open and said,
• 12:46
  you know, come in and do whatever you want, like where would you start?
• 12:49
  So I think.
• 12:50
  So the way we start is it's almost like a T shaped approach.
• 12:54
  You know, we look broad and we say, hey, these are your divisions.
• 12:58
  You know, let's look at say the top level metrics for these divisions, right?
• 13:02
  What is the KPIs and how do you track across all?
• 13:06
  So we always start with the KPIs and the numbers, many times the numbers don't exist.
• 13:11
  But even the back of envelope calculations, what's your margin like, what's your resource footprint like and what are your goals?
• 13:20
  And then once we identify Areas where the biggest opportunities are, then we can go deeper and then we can do an assessment, end to end assessment.
• 13:29
  So it will typically start with a current state assessment.
• 13:32
  We look at your workflows, we look at, we have conversations and interviews.
• 13:39
  And I know there is conversation that today you can automate a lot of this, you can have AI interview people, but it's not the same thing.
• 13:49
  Even though I have AI that can interview people, I would never want AI to interview people because people want to be heard.
• 13:57
  And the way they will converse with an AI is very different with the way they will converse with humans.
• 14:03
  A lot of the value is in that hidden conversations, the hidden gems in the conversations that surface.
• 14:11
  When you have a conversation, you connect with the person, they really say, hey, let me tell you what exactly is the problem here?
• 14:20
  And they are generally right.
• 14:21
  The intuition of people is generally right.
• 14:24
  That interviews, process mapping is the base.
• 14:27
  And then we start with the current state, we do a gap assessment on what could be the future state.
• 14:34
  What are the benchmarks?
• 14:35
  If today your conversion rate is 10% and the industry is converting at 20%, then we know you have that opportunity.
• 14:44
  If your cost per Invoice is say $25, industry cost per invoicing process is $5, then we know that, okay, you have this gap which you can achieve and you can go over.
• 14:57
  And with AI, the reality is that the benchmarks are also changing.
• 15:01
  Right.
• 15:02
  Like yesterday, if the benchmark is cost per dollar, cost per invoice is $5.
• 15:07
  But then with better AI processes, you can actually bring it down to even more.
• 15:12
  So I think that opportunity helps us create a roadmap and then you have to do something real and show the results really fast to have people believe in what you are saying or what you.
• 15:27
  The vision that you are showing becomes real when you build something that could be really small really quickly and show them the impact.
• 15:35
  So I think there's a three step playbook, like look across, you know, broadly, pick one area, go deep, and the third is show quick results however small it is to get alignment along the way.
• 15:49
  Yeah, magic.
• 15:50
  And that's where it all comes together.
• 15:52
  And totally agree.
• 15:53
  The same thing is true of databases, data and databases.
• 15:55
  Like a company could hand to my team the keys and access to every single database in the company.
• 16:02
  But would I understand what's happening at that company from that?
• 16:04
  Absolutely not.
• 16:05
  Yeah.
• 16:05
  Oh, here's the monthly board report.
• 16:07
  That's probably the least useful thing.
• 16:10
  That just shows.
• 16:11
  That just shows me what the board want to see rather than what's Actually happening sometimes.
• 16:16
  Sometimes.
• 16:17
  And there's a case study that I read about from you of like, you know, companies uncovering like between 7 and 15% of EBITDA, like through your work in, you know, finance operations.
• 16:28
  Like where is that, where's that money hiding?
• 16:30
  Like what does it come from?
• 16:32
  So in finance the money is hiding in the data, right?
• 16:37
  So in any other kind of division, if there is data problems, then oh, okay, you will not be able to acquire the customers well, or there are other problems, you don't target them as well.
• 16:49
  But in finance, data problems have real implications, right?
• 16:54
  You will bill customers, but then you never collected the revenue.
• 16:59
  So especially in the order to cash process, there is so much of problems that come in, and they come in from the manual work, the human touch where you are doing something repeatedly, repetitively, copy paste, and then the
• 17:14
  mistakes happen.
• 17:15
  So there are so many mistakes around in the invoicing process where you have duplicate invoices, invoices were sent out, never paid.
• 17:25
  So tracking all of that, especially if the bank and when payments come in, because order to cash is a very, it's a very customer driven, sensitive process because it's a customer, especially for large customers.
• 17:39
  You may have your own custom process, right?
• 17:42
  So when cash comes in, being able to track the cash accurately, many times there is remittance information that's there in emails that is not available.
• 17:52
  And you know, there is so much of like, you know, look up the check and then you open the email and then you look at the bank statement and there's so much of this manual verification that's being done,
• 18:04
  all of that.
• 18:05
  And then there are system limitations and it's to nobody's fault.
• 18:10
  Like for example, companies that have grown through acquisitions, then they have these different systems that need to come together and to have like a combined order to cash process.
• 18:23
  But if that doesn't happen, then the result is spreadsheet billing which comes back to the same manual problems.
• 18:29
  So I mean, I can just keep talking on the list of issues and problems that can happen, but automation is not everything because see a lot of these problems also exist because of systemic issues in terms of there's
• 18:47
  an acquisition that has happened.
• 18:49
  So disparate systems exist.
• 18:51
  Sometimes it's very hard to merge the systems because of compliance requirements.
• 18:57
  In one example, there is this company that is issuing insurance policies and because there are different rules for different states, these two systems, they cannot be merged together or the companies have been working very hard to do that.
• 19:15
  So those are some of the things that hold you back Then there are non standard contracts, especially if you work with really large customers.
• 19:28
  Amazon will not do a standard contract with you.
• 19:31
  Amazon will have its own contract and everyone wants Amazon as their customer.
• 19:38
  So they will not pay on the terms that are there for you.
• 19:42
  They will have their own terms, they will have their own portal.
• 19:44
  You have to follow their process.
• 19:46
  And if your invoice doesn't follow the process, they will not pay you.
• 19:49
  Or they can even cancel you as a vendor.
• 19:51
  90 I heard as a standard.
• 19:53
  So you're well aware.
• 19:54
  Yeah, yeah, yeah.
• 19:56
  And they will not pay you if the invoice is not correct for two or three quarters.
• 20:00
  I mean, that's a lot of responsibility.
• 20:02
  And it's just so there are so many of these things happen and the leakage happens because of these reasons.
• 20:14
  But the answer is not just straight up, go automate every piece.
• 20:20
  Many times these things are interconnected with each other.
• 20:23
  For example, in one case I was seeing this team that was building this report and it was pulling data from the sources and building this report.
• 20:32
  Then there was another team that was taking the same report and breaking it into the three pieces that they were building it from.
• 20:38
  That repetitive things happen because people don't know the work happens in silos.
• 20:44
  So, like a more holistic view of the ecosystem is what's really important for you to know how you can really transform.
• 20:53
  Yeah, absolutely correct.
• 20:56
  Humans, though, we talked about humans a few times.
• 20:58
  So I want to talk about the humans in the loop because I think I've seen you describe your AI, your agents, AI co workers, and as part of that, you design kind of human oversight into that which is
• 21:09
  designed to do exception handling and any of that stuff.
• 21:12
  But nobody in a private equity portfolio finance team is going to hear AI agent and feel calm about that agent making decisions on their finances.
• 21:22
  As you correctly said before, there's there are, you know, real consequences to that.
• 21:26
  So, you know, for a CFO or an operating partner who's a partner who's just starting their journey with AI, like how do you land that as a people story to get them comfortable?
• 21:35
  Yeah, that's a great question.
• 21:37
  See, while we say agents, agents is a, is a very overused word, I feel like, and it's a buzzword.
• 21:45
  You got to use it.
• 21:46
  But it's not necessarily only agents that we are using as we get the work done.
• 21:53
  Like for example, especially something like invoicing.
• 21:56
  Right.
• 21:58
  You need a company to have a very standard way of invoicing.
• 22:02
  You know, when you say agent, an agent will make decisions.
• 22:05
  So I Don't want my agent to decide how to invoice every single month.
• 22:10
  I want my agent to follow a very deterministic process.
• 22:13
  However, wherever there are exceptions or however, if the data is different or I have to do some resolution, then I need AI to make the decision for me.
• 22:23
  Right.
• 22:23
  So that's how the process has to be structured.
• 22:27
  Now, for any PE operating partner who is thinking of bringing in AI into their finance specifically, what IT can help accelerate for you is it can help accelerate the build out of the agent agents or the workflows.
• 22:50
  So we call it agents, but it is actually workflows with intelligence embedded in it, plus agents on the exception management side.
• 22:58
  So you can start with some of your workflows where you are seeing that you could be losing money.
• 23:05
  Like invoicing is a very common one.
• 23:07
  For example, I am a startup myself and invoicing every month I have to go and I have to check my schedule and then I have to go somewhere else.
• 23:18
  I have breaks.
• 23:19
  I will generate the invoice from there.
• 23:21
  Then I have to kind of reconcile it in QuickBooks.
• 23:24
  So while these systems exist, there is still human effort that is needed.
• 23:29
  So I was talking to somebody around like two weeks ago.
• 23:33
  They said towards the month end they have to spend like four or five hours reconciling the payments between Ramp and QuickBooks.
• 23:41
  And they were duplicate payments and I don't want to make duplicate payments.
• 23:44
  So Ramp was telling me just yesterday that all you need to do is have Ramp Pro and then that will all be taken care of for you.
• 23:53
  Okay.
• 23:54
  See?
• 23:55
  Yeah.
• 23:56
  Yeah.
• 23:56
  So our worlds are colliding.
• 23:58
  Looks like the problem, the solutions are.
• 24:02
  Well, I don't know.
• 24:03
  I've not tried to run pro myself.
• 24:05
  They do try and get me to do it every week.
• 24:07
  And I work with an accountant and a bookkeeper, so they take care of all the syncing for me.
• 24:11
  But I do see occasionally the tasks pop up in Ramp and you have to sync across the QuickBooks.
• 24:16
  So, for example, that's a good one.
• 24:18
  I do love Ramp.
• 24:19
  And you said you use Brexit for expense management as well.
• 24:22
  Like, as a Capital One alumni, I'm very interested in your Brex experience because Capital One just acquired Brex earlier this year.
• 24:30
  I think Brex is heading on the same chart as Ramp is.
• 24:32
  But the thing I love the most about it, like, I've been like you at Fortune 500 companies all over my career.
• 24:38
  So submitting expense reports and doing expenses and taking photos of receipts and emailing them to myself has like been a Part of my daily life, longer than I care to remember.
• 24:47
  But now the card gets swiped behind the counter.
• 24:52
  I get an instant text message, I take the photo of the check and then logs does all the stuff.
• 24:59
  It's a really good example of progress in this space from the AI native platforms.
• 25:05
  What value coworkers can bring in is being able to connect across these different platforms and give you a view of your business, how your business is going.
• 25:17
  Like standardizing your portfolio view on what is the margin of your portfolio and then what are the KPIs.
• 25:25
  I think those are some of the key things that can help any partner, any portfolio, pe, operating partner.
• 25:32
  And the other thing is if you look at in terms of problems, problems will have patterns, right?
• 25:41
  If you have like a consulting company, you have the same pattern of the KPIs, the problems.
• 25:47
  Or if you have to work on your resource cost, then you have to look at things like Pyramid Mix and you have to look at things like your location strategy.
• 25:58
  So those things are some of the playbook elements that a diagnostic LED process can help you identify.
• 26:06
  And then we have our library of agents that you can plug in wherever you see the opportunity.
• 26:12
  So that's kind of like a two part framework where the discovery is the first part of it.
• 26:17
  Where you identify where the opportunities are and then where the gaps lie.
• 26:22
  And the agent is really your reimagined process.
• 26:27
  So you have this current process that has all of these issues the way we are building it today.
• 26:33
  And here's the interesting thing again I think we probably have this in common.
• 26:37
  There are no doubt people with vibe coded software platforms out there that will claim to take care of all of this with one software license.
• 26:45
  You know that they can't and I know that they can't.
• 26:48
  Like how can people who are shopping for this type of work and looking at solving these types of problems tell the difference?
• 26:55
  Yes.
• 26:55
  See there is, there is this.
• 26:58
  There are a lot of.
• 26:59
  There are different, there are tools of different types, right?
• 27:02
  There are these copilot's intelligence in a box, let's say that.
• 27:08
  Right?
• 27:09
  Copilot cloud.
• 27:11
  Then there are your workflow tools, Crew AI, others.
• 27:14
  They stitch your workflow together.
• 27:16
  And we also did start with that view too that you bring everything and run the workflow.
• 27:24
  But then there are two things that are important I think.
• 27:28
  One is your platform that you're using and it's not necessarily a platform, it could be platform plus people.
• 27:37
  Because there is diminishing cost of software development today is right.
• 27:42
  So it does it understand, do they Understand your business problem.
• 27:48
  Are they really aligned towards your business goals?
• 27:52
  You know, if you have this EBITDA and you want to move this EBITDA goal or you know, you have this transformation program going on and these, this, these are your goals.
• 28:02
  So do they understand your business goals?
• 28:05
  Because in the end I deeply believe that software, people, processes, everything has to be aligned to the business goal.
• 28:14
  I think that's the first thing.
• 28:16
  And the second thing that's important is your data, right?
• 28:21
  What happens today is you have this ERP that's the source of a lot of things.
• 28:27
  Of course that's where you want that to be your centralized source of truth.
• 28:32
  But then you have all of these non standard contracts and then different kind of products or these things around the ERP that takes so much space and the spreadsheet, billing and all of that happens around the erp.
• 28:48
  But then what happens when this thing, you do this work around the erp, you have these humans doing the work.
• 28:53
  There is no audit trail, there is no data, there is no context that is being captured.
• 29:00
  My second point would be that, does your product bring in that data context?
• 29:08
  Does it accumulate that data context for you today?
• 29:12
  Today when we bring in these agents, you may have been uploading invoicing data from Excel sheets which is like or emails from a couple of different systems.
• 29:23
  But then we bring everything down to that data model and that unique data context which becomes the truth for all the other agents.
• 29:32
  So somebody asked me just last week, they were like, are you building any rp?
• 29:38
  Because you are in some ways building an erp, you're kind of bringing everything together and no, I'm not.
• 29:46
  Because there are enough ERPs out there.
• 29:49
  We don't need another one.
• 29:51
  What we are really building is we are building for that space around the ERP that has so many of these problems that are impossible to solve.
• 30:02
  Not only if you just get good technology is not enough.
• 30:06
  There are different.
• 30:07
  It could be due to organizational issues, it could be due to compliance problems and all of these reasons why that data or that contract cannot be out of the crp.
• 30:18
  That's the problem that we are solving.
• 30:20
  We are solving for the problems where humans are stitching and bringing things together and it's good.
• 30:28
  I'm also not saying that you need to get rid of your humans.
• 30:31
  We absolutely need to keep.
• 30:32
  Glad to hear that.
• 30:34
  See, because I have not seen one place or one, one accounting firm that has said, hey, we got rid of humans.
• 30:42
  I think that getting AI replacing human story has not really played out.
• 30:48
  And I'm not very sure how well it will play out.
• 30:52
  Efficiencies will come, but a system that will completely run autonomously by itself in finance, I do not think that's going to happen for a long time.
• 31:04
  I have seen the story for the last two years as we go, we are actually moving farther away from that story.
• 31:11
  Then closer we will have the humans, but then the humans will not do the running, the macro and the copy paste and all of those things that result in the errors.
• 31:24
  The humans will focus on the judgment task, the follow up with the customer, the customer relationship building, you know, the customer csat.
• 31:35
  So all of those things that you just don't have time for because you just have to get the invoices out.
• 31:42
  I think that's the part where humans will always be there, but that's the part where the weight of the work is going to move with what we are building.
• 31:51
  Yeah.
• 31:53
  Speaking about humans, you know, people like us who spent lot of time with AI are often doing cool things at home in the evenings and weekends.
• 32:02
  So like, have you done any sort of personal projects that have been valuable to you with AI when you've been kind of playing around with things at the weekend?
• 32:09
  Or have you heard of a cool thing that somebody else has done?
• 32:12
  So actually I think the company is a good enough personal project for me.
• 32:17
  Yeah, I know it's pretty all consuming.
• 32:18
  I get that.
• 32:20
  So we have built pretty much and our entire company around AI So we started and that's like from the time when I was brainstorming on the idea to actually like doing customer discovery, I have to say 100%.
• 32:37
  Our company is based on the foundation of AI.
• 32:41
  If AI was not there, then what I have achieved in the last one and a half years, it would have taken me five years to achieve it in some other time.
• 32:51
  So I can very easily say that.
• 32:54
  And what AI has also allowed us to do is to be like, as a founder, I'm very close to the product.
• 33:01
  I pretty much wipe code a lot of the product.
• 33:05
  A lot.
• 33:06
  The entire front end, the design and how it looks like, what the customer experience will be.
• 33:12
  I'm building it myself.
• 33:14
  So AI has enabled us to do that.
• 33:16
  So that's kind of like my pet project.
• 33:18
  Honestly, I did try.
• 33:21
  So I also write on the side.
• 33:24
  So I have this standing journal and I have these unwritten set of books.
• 33:31
  Do you write for yourself or do you write for an audience?
• 33:35
  It's all writing for myself.
• 33:37
  It's not for an audience yet.
• 33:39
  It is more of like I have these Unfinished books.
• 33:43
  And they are like chapters of my.
• 33:45
  Life, like Mozart's unfinished symphonies.
• 33:48
  Yeah, I hope that.
• 33:53
  So, you know, I've always, like, wanted to write them in a more structured way, so I thought I will use AI to do that.
• 34:00
  But the writing turned out to be so bad.
• 34:02
  Like, it was so bad.
• 34:05
  Like, even a word, like a sentence or a paragraph that I wrote in my sleep would be better than what Claude wrote it for me.
• 34:15
  So somehow, especially on the personal side.
• 34:19
  Right.
• 34:20
  On the professional side, the writing is great.
• 34:23
  You can take what is there and then you can publish it.
• 34:26
  But if you're writing something like, that's personal.
• 34:29
  So I use AI for like structuring my arguments and then kind of structuring my chapters, critiquing the work and then saying, hey, these are the ideas that I could have put in.
• 34:42
  Or that's kind of like how I use AI for personal writing.
• 34:47
  Personally, I have not found it that useful.
• 34:50
  It's mostly on the professional side that it's really helped me a lot.
• 34:53
  It operates best in systems with rules and logic and decision making.
• 34:58
  Logic, exactly.
• 34:59
  Like wherever there's a little bit of emotion that's there or.
• 35:04
  Yeah, it just.
• 35:05
  It's very strange to me.
• 35:07
  Like, I can immediately read what's written by AI and know it's written by AI.
• 35:11
  It just lacks the soul that's needed for good writing.
• 35:16
  Yeah, I think.
• 35:16
  I think that's absolutely true.
• 35:18
  I really appreciate you coming on.
• 35:19
  If folks are interested in learning more about Luminar data or more about you, or just want to connect to your space so that when the unfinished books finally get finished and published, that they're there to snap them up.
• 35:29
  Where's the best place for people to find you?
• 35:31
  You can find me on LinkedIn.
• 35:33
  You can search for my name, Afrozi Ara, and I'll be there.
• 35:38
  My email is afrozieluminadataai so I can be found over email too.
• 35:44
  Perfect.
• 35:45
  That is magical.
• 35:46
  Thank you so much for coming on the show.
• 35:48
  Good luck with the continued writing.
• 35:49
  I'm glad there are people out there still writing and doing it for yourself, I find is a really good way to get your thoughts onto paper and I think it's a good mental health exercise as well.
• 35:58
  I certainly agree with that.
• 35:59
  So good luck with the continued writing and I'm sure I'll speak to you again very soon.
• 36:04
  Thank you, Graham.
• 36:05
  It was wonderful talking to you.
• 36:07
  Yeah, likewise.
• 36:08
  We'll definitely be in touch.
• 36:10
  Okay, take care.
• 36:11
  Take care.
• 36:12
  Thanks for listening to the PE Data guy.
• 36:14
  The place where private equity meets data.
• 36:17
  Please forward this episode to your favorite private equity friend.
• 36:20
  Thanks for listening.
• 36:21
  See you next time.