From Black-Box Systems to Augmented Decision-Making | Anusha Akkina

The transcripts are edited for clarity, sometimes with AI. If you notice any incorrect information, let us know.

Building trust in AI finance and introducing Auralytix

Alexey: Hi everyone, welcome to our event. This event is brought to you by DataTalks.Club, a community for people who love data. We have weekly events and today is one of them. If you want to learn more, there is a link in the description. Click on it to see all upcoming events. (0.0)

Alexey: Do not forget to subscribe to our YouTube channel so you stay up to date with future streams. We also have an amazing Slack community where you can hang out with other data enthusiasts. During today's interview, you can ask questions using the pinned link in the live chat. Click the link, submit your question, and we will cover it later in the interview. That is the usual intro. (5.0)

Alexey: Now I will stop sharing my screen and open the questions we prepared. There is a new Zoom feature called AI Companion that I have not used for a podcast before. I am giving it a try to see how well it generates a summary. Anusha, if you are ready, we can start. (44.0)

Alexey: This week we will talk about how AI is transforming the finance function. We will discuss how it helps teams move beyond spreadsheets and slow systems toward faster data driven decision making. Our guest today is the co founder of Auralytix. (1:11)

Alexey: How do you pronounce it? Could you say it again? (1:23)

Alexey: Auralytix comes from Aura meaning gold, but also aura as in the spiritual field around a person. It is an AI driven finance platform that gives CFOs and finance teams clarity and speed without adding complexity. Anusha’s work focuses on using AI to augment, not automate, finance and combines advanced analytics with compliance, explainability, and trust. Welcome to our podcast. (1:41)

Anusha: Thank you. Thank you Alexey. (2:17)

From accounting roots to auditing at Deloitte and Paraxel

Alexey: Before we talk about AI and finance, I want to hear your story. I enjoy learning how founders start their journey. Running this community is also a kind of business, so it is always interesting for me to hear how people make the leap from a stable job to founding a company. Tell us about your career so far. (2:22)

Anusha: I am Anusha, based in Hamburg, Germany, originally from India. I am a Chartered Global Management Accountant from the UK. I worked at Deloitte and Parexel in multiple roles. I started my career as an auditor, doing internal and sometimes statutory or external audits. (3:01)

Alexey: What do auditors actually do? I worked at a company where people in suits would show up, usually from Amsterdam for some reason, and they were auditors from Deloitte. I knew they worked long hours and earned good money, but I never really understood what they do. (3:38)

Anusha: Auditors examine a company’s processes, financial data, non financial data, and operational data. Their role is to reassure authorities that the company is paying taxes correctly, following regulations, and not committing fraud. Someone independent must confirm that everything is compliant. Internal auditors help management ensure operations run smoothly, and external auditors, like Deloitte or PwC, verify processes, test data, and issue audited financial statements. (4:15)

Anusha: It is not a value adding role in the usual sense, but it gives a holistic perspective of whether a business is working properly. It brings high trust for the government and stakeholders because they know the company follows compliance requirements. (5:28)

Alexey: It must be a difficult job because you have to check all the books and make sure people are not hiding anything. (5:45)

Anusha: Yes, and sometimes people get offended when you ask certain questions. You might face situations where someone with thirty years of experience wonders why a young auditor is questioning them. But the best part of the role is that you see many businesses and develop a broad understanding of processes. You learn to detect bottlenecks and identify what is not working well. (5:59)

Anusha: It is also very labor intensive. You travel a lot, sometimes to remote places, especially for manufacturing or construction audits. You might stay there for months. It is great experience but can also cause burnout. (7:23)

Moving to Germany and pivoting into corporate finance

Alexey: Did you do this in the UK or India? (8:20)

Anusha: My qualification is from the UK, but I worked as an auditor in India. (8:25)

Alexey: So you did this for a while. What did you do next? (8:31)

Anusha: After auditing, I moved into a corporate finance role as a financial controller. A controller sits on the other side of the table from the auditor. I moved to Germany for personal reasons, but getting an auditing job without German language skills is very difficult. You need at least C1 German, which takes time to achieve. So I chose the next best option, a controller role with a specialization in SOX, Sarbanes Oxley. I found exactly that type of job. (8:38)

Alexey: Interesting. How is your German now? (9:34)

Anusha: It is good. My team consisted mostly of amazing women who did not speak much English, so they made sure I learned German quickly. Within one to one and a half years, I became reasonably fluent. (9:39)

Alexey: Do you think your German level now would allow you to work as an auditor? (10:04)

Anusha: Yes, in conversations it should be enough. I am not sure about exams, but I can communicate well. (10:09)

Alexey: I got a B2 certificate a year ago. Speaking is one thing, passing a test is another. Even with the certificate, I sometimes freeze in real situations. (10:31)

Anusha: It is the same for me. I also have B2, but certificates matter less than being able to speak, understand, and communicate. You will improve with time. (10:58)

Alexey: How long have you been in Germany? (11:19)

Anusha: Since around 2018, so about seven years. (11:26)

The data struggle in strategic finance and the need for change

Alexey: You were working in corporate finance as a financial controller. At what point did you decide to go solo? (11:50)

Anusha: After the controller role, I moved into a larger role as lead finance or finance business partner for Central Europe at a UK based company. It was more strategic than controlling. Even though the role was strategic, I still spent about 80 percent of my time not analyzing data but chasing it. (11:58)

Anusha: Every month felt like Groundhog Day. I waited for updates, joined meetings to ask what was happening, and stitched together numbers from different spreadsheets. I always wondered whether the data was complete or current. The workload was fine, but the dependence on other people just to get data was frustrating. (12:37)

How Auralytix was born: bridging AI and financial compliance

Anusha: Finance had become disconnected from what was really happening in operations. Even with modern ERPs, CRMs, and BI tools, finance was still reactive, never ahead. Around 2023, I read about Databricks and how it unified data across silos so companies could move from chaos to insight. It resonated with me because finance faces the same fragmentation but with stricter compliance requirements. (13:20)

Anusha: I thought that if Databricks could transform how enterprises use data, there should be something similar designed for finance, but with compliance and audit trails in mind. When I did not find a solution like that, I decided to create one. Although I am not an engineer, the last two years changed the way I think. I now understand data perspectives and finance pain points. That led to Auralytix. (14:11)

Alexey: You mentioned ERPs. Coming from software engineering, I sometimes had to interact with finance systems for procurement. We used a system called Coupa, and as an end user, it was painful. The finance team had to write long manuals just so we could request and pay for tools. What is the state of ERPs now, and what exactly is an ERP? (15:18)

Why ERP systems fail finance teams and how spreadsheets fill the gap

Anusha: ERP stands for Enterprise Resource Planning. A company has many processes such as finance, procurement, sales, operations, manufacturing, supply chain, and logistics. The data is everywhere, and you need one platform to integrate it so you can build reports and connect the dots. That is the purpose of an ERP. (17:15)

Anusha: In reality, ERPs rarely achieve this. They are black box systems with huge manuals and require expertise to make even small changes. This creates heavy dependence on spreadsheets, and that is where problems begin. (17:53)

Alexey: Let me clarify with a simple example. I run a small company selling shoes. Every sale must be recorded somewhere. Eventually, all this data should end up in the ERP so auditors or finance people can see our spending, revenue, and number of shoes sold. Then someone in a strategic role can pull the numbers and guide the company. Why does this not work in practice? (18:50)

Anusha: ERP systems have rigid, standardized structures. They are one size fits all with small customizations. For a shoe company, management needs insights such as which models sell well, which sizes, which colors, who buys them, and which trends are emerging. Before Christmas, which items should be stocked, and during Black Friday, which items should be discounted. These are strategic questions. (19:45)

Anusha: If you open the ERP, you will not find a report answering any of these questions. ERPs are built for compliance and storing data, not for analyzing it. That is why spreadsheets, BI tools, and many specialized systems like Coupa exist. (20:48)

Anusha: For a CFO, data is scattered across many tools. At a single moment, it is hard to understand what is happening in the company. By the time controllers or analysts gather all the data and turn it into insights, the opportunity to act is gone. The Black Friday sale is over, or Christmas is over, and the company has already fallen behind competitors. (21:40)

Anusha: Enterprises win because they can afford sophisticated tools. They use them to create demand, refine strategy, increase profits, and act faster. But for small and medium sized businesses, even up to 500 million euros in revenue, it is difficult to understand these patterns. (22:22)

Alexey: The complexity is high. As I understood, correct me if I am wrong, the data is in the ERP, which is the database where we record all transactions. We keep an eye on stock and see our inventory. But this is not enough because the data is just sitting there and it is not easy to extract. For that we need an analyst who can set up a process to get data out and create a dashboard or report so that before Black Friday we know which shoes we need to discount. (22:30)

Alexey: We need an analyst and perhaps a technical person who can pull the data and make this report. Is this how it works (22:50)

Anusha: It is not only that, it is also about incomplete data. The systems are rigid and even small changes require an army of consultants to update the ERP. These consultants are expensive, so companies accept limitations and switch to manual processing. It is not only about hiring analysts to extract and transform data, but also about ERP rigidity where data is never complete. It is only in a fixed form. (22:57)

Alexey: So when I buy SAP or another ERP system, they have a fixed flow for entering data and defining what data we can enter. If I want to go outside the typical process I have problems. Some systems use visual basic or other languages for customization. But finance people are usually not developers and the software is not easy to deal with. (23:15)

Alexey: Someone told me that if you want to make money as a developer you should become an SAP consultant. (23:25)

Anusha: I agree completely. This is the biggest gap I see. As a finance person you do not have the technical knowledge to transform your ideas into tools or data you want to use. That is where external consultants come in. Some are great but some just make money because you do not know how small changes cost a lot. (23:28)

Anusha: Companies spend enormous amounts of money. The problem does not stop after the first change. If you think one add-on will solve everything, it will not. The pace of business is fast and these tools do not have the same agility. They do not scale accordingly and that is why we need AI based ERPs and AI platforms that can run at the speed of business. (23:45)

Alexey: What do you mean by an AI ERP? Instead of explaining what I want, I talk to AI, formalize it, press enter and the AI system modifies my SAP form or whatever system I use. It adds an extra field or changes the workflow. Is this what you mean (24:02)

Anusha: Yes, something like that. We need a flexible system that can scale with the organization. I think this is the future. Because until now this has been the biggest pain. (24:12)

The real cost of ERP rigidity and lessons from failed transformations

Anusha: Maybe I can give an example. In one company where I worked the board decided to change the ERP system. They thought it was the best ERP for manufacturing and project management. They invested around a million euros and with consultants implemented it in two years. But the data quality became so bad that we did not know what was happening. (24:31)

Anusha: Even operations people were frustrated. It went back to the board and they decided to switch again to another ERP system. This new system was three or four times more expensive. After a year we were in a better situation but still not at the level we expected. We started buying new add-ons because the ERP was missing functions. (24:52)

Anusha: This is how finance transformation projects go. More add-ons and more silos because the ERP does not solve everything. What we need is a system flexible and scalable enough to grow with the business. I think AI makes this possible. (25:16)

Alexey: You mentioned your personal experience where you adopted a new ERP system but after two years the data quality dropped and you had to switch again. Why did the quality drop Was it because the company changed but the processes in the ERP did not change, so people stopped entering data correctly Was the data entry process too complex and people were lazy to do it Why was there a drop in quality (25:31)

Anusha: There were multiple reasons. One reason was the way people entered data. Management needs many different perspectives from sales, purchasing and operations. Decisions depend on many data points and someone must input the data correctly so it can be converted into patterns. These patterns become insights for strategic decisions. (25:47)

Anusha: People did not know where to put the data because the ERP was rigid. You could only use the available fields. For strategy we could not understand what was going on. Let me give an example. Our company worked on contract basis. A major criteria to understand performance was contract renewals, new contracts and closed contracts. (26:10)

Anusha: This was basic information to understand where to focus and why customers were leaving. The system could not provide it because it was not built for that. It was an ERP for a trading company, not for project management or contract-based business. If systems are rigid you cannot get the data you want. You start using an Excel file again. (26:31)

Anusha: You fill Excel manually and one mistake breaks everything. Another example is fixed assets. The ERP did not even have a fixed assets module. All fixed assets and depreciation were in Excel. So you enter some data in ERP and extra data in Excel. You keep the link in your mind but it is easy to make mistakes when everything is manual. (26:52)

Anusha: There is also the issue of turnover. Accounting teams have high attrition. When someone leaves it is hard to replace them and by the time a new person comes the knowledge is gone. Then you hire a consultant to set everything up again. There is no continuity and no one knows where the Excel started. (27:16)

Anusha: People leave, new people join with no idea about the spreadsheet, so they create their own file with a different format. Then they leave and someone else creates another version. You end up with many Excel files and nobody knows how they work or what the current state is. Historical data becomes a mess. (27:36)

Alexey: For me it is interesting to know how this led you personally to think you can fix these problems. How did it happen for you What was the process Did you see that Databricks was doing this for analytics but nobody was doing it for finance and thought you could be the one How did the timeline look Because leaving a well paid role and taking the leap to start your own company is brave. (27:53)

Anusha: It is a good question. This happened around 2023 when I was frustrated with my work. We completed our ERP migration to a new tool and also had a consolidation system. In global companies you have a local ERP and on top of it a consolidation system. (28:12)

Anusha: A consolidation system gives management one single source of truth. Every country has a different way of putting accounts together, but for management everything must be uniform. Systems like SAP EPM, Tagetik or HFM are used for consolidation, budget and forecast. Budgeting and forecasting are strategic tasks. (28:31)

Anusha: You must understand data patterns, new customers, customer retention, suppliers and many considerations for forecasting sales, cost of sales, working capital and cash flow. You analyze everything, talk to many people and by the time you upload the data you have 20 to 30 versions of Excel sheets. (28:52)

The hidden risks of spreadsheet dependency and knowledge loss

Anusha: In a meeting when someone asks why a specific product revenue changed you must go through all these Excel files. You spend so much time but still cannot answer exactly. This should be solved by proper processes. Around that time the first public version of ChatGPT came out. It was not great yet but I was a tech enthusiast. (29:10)

Anusha: I knew how the data looked, so I prepared a mockup file because I could not upload company data. I loaded it and asked AI to transform it and find patterns. It did surprisingly well. I would rate it maybe 55 or 60 out of 100 which already saved 60 percent of my time during tough budgeting periods. (29:30)

Anusha: I gave two different datasets, sales orders and data from another system, and asked it to build a bridge between them. It did well. It felt like magic. I wondered if it was real. (29:52)

Alexey: Did it write code or how was it doing this for you Was it code (30:00)

Anusha: Yes, mostly by writing code. I learned basic Python but before you could not just upload spreadsheets like now. Today you can upload the file to ChatGPT, see the code and look only at the final result. At that time you had to copy the code, execute it and prepare the environment. Installing Python is easy for technical people but not for many finance people. (30:02)

Anusha: This was based on many data points, and someone had to input them in a specific way so they could be converted into patterns and turned into insights. As a strategic role, you look at these insights and understand what is happening. You think about missing mitigation and how to deal with the situation. At that time people did not know where to put the data because the ERP system was rigid. You could not change anything and only saw the available fields where you entered the data. (30:04)

Anusha: For strategy, you still could not understand what was going on. For example, our company worked on a contract basis, so one of the major criteria to understand performance was the number of renewed contracts, new contracts and closed contracts. This basic information helped us understand where to concentrate and why people were leaving. But the system could not provide it because it was not built for that purpose. It was an ERP system for a trading company, not for project management or contract-based work. (30:26)

Anusha: If systems are rigid you cannot get data in the way you want, and then you start using an Excel file again. You fill the Excel file manually and one mistake can make everything wrong. This was one example. Our company was also heavy in fixed assets, and the ERP system did not even have a fixed assets module. All fixed assets and depreciation were in Excel. (30:52)

Anusha: You had the ERP for what it could handle, and an additional Excel spreadsheet for data without fields. You kept the link between ERP and spreadsheet in your mind. It was very easy to make mistakes when entering everything manually. It was not an ETL or automated process. Everything was manual and that was the biggest issue. (31:14)

Anusha: Accounting teams also have a high attrition rate. People leave quickly, and when someone leaves it is hard to replace them. There is a gap before a new person comes, and the knowledge leaves with them. Then you need a consultant to set things up again, and later another consultant when someone else leaves out of frustration. There is no continuity in the data. (31:33)

Anusha: Someone starts an Excel spreadsheet and leaves, and the new person has no idea about it. They start their own spreadsheet with a different format. Then they leave and another person starts yet another version. You end up with many Excel files. Nobody knows how they work, what the current state is, or how to get historical data. It becomes a mess. (31:55)

Alexey: For me it is interesting to know how you saw all these problems and thought you could fix them. How did this happen for you? What was the process? You mentioned Databricks, and you thought Databricks does this for data analytics but nobody does this for finance, so maybe you could be the person who tries. How did this work for you? (32:19)

Alexey: Also, leaving a company and a well-paid role to start your own company is brave. I want to know more about that. (32:45)

Anusha: It is a good question. This happened around 2023 when I was very frustrated with my work. We completed our ERP migration to a new tool and also had another consolidation system. In global companies you have a local ERP system and on top of it a consolidation system. A consolidation system provides management with one single source of truth. (33:00)

Anusha: Every country records accounts differently, but management needs a unified top-level view. Systems like SAP EPM, Tagetik or HFM are consolidation tools for management reporting. You do forecasting and budgeting at the consolidation level, not at the local level. Budgeting and forecasting are strategic tasks and you need to understand data patterns. You think about new customers, existing customers and suppliers. (33:25)

Anusha: There are many considerations when forecasting sales, cost of sales, working capital and cash flow. You analyze everything and speak to many people. By the time you upload the data, you have 20 to 30 versions of Excel sheets. When you sit in a meeting and someone asks why a particular service revenue changed, you go through all the versions to answer. You spend so much time and still cannot explain exactly why. (33:50)

Anusha: This should be solved by proper processes. Around that time the first public version of ChatGPT came out. It was not great yet, but I was a tech enthusiast. I prepared a dummy mockup file because I could not upload company data. I loaded the mock data and asked AI to transform it and understand the patterns. (34:15)

Anusha: It did surprisingly well. I would give it around 55 or 60 out of 100, which already meant saving about 60 percent of my time during budgeting and forecasting. Then I provided two different sets of data, like sales orders and Datar data, and asked it to bridge them. It did well and felt like magic. I thought it was amazing that something could do this. (34:38)

Alexey: Was it writing code for you? How was it doing the work? Was it generating code? (35:02)

Anusha: Yes, it was generating code. I learned basic Python and it was mostly writing code. Today you can upload spreadsheets directly and get analysis without looking at code. You see the result and adjust it. Before, tools were not available, so you needed to run the code and prepare the environment. (35:10)

Anusha: For technical people installing Python is easy, but for someone in finance it is not. I was lucky because my husband is a software engineer. He knows Python and helped me install open-source models. He also shared videos so I could learn. At the time I had my baby and was on maternity leave. (35:32)

Anusha: So again you have systems like SAP, EPM, Tagetic or HFM. These are consolidation tools for management reporting. You do your forecast, budget and everything at the consolidation system level, not at the local level. When I used to work on budgeting and forecasting, it was always a strategic task, not a day-to-day task. You have to understand the patterns in the data. (35:54)

Anusha: You need to understand how you will get new customers, how many customers will stay with you and which suppliers you need to take care of. There are multiple considerations when you forecast sales, cost of sales, working capital and cash flow. You do everything, analyze everything and speak to hundreds of people. By the time you upload the data, you have 20 to 30 versions of Excel sheets. When you sit in a meeting and someone asks why you changed a particular product’s service revenue, you have to search through all these Excel sheets. (36:16)

Anusha: You have already spent so much time, yet you still cannot explain exactly why you did something. This should be solved by better processes. Around that time the first public version of ChatGPT came out. It was not great yet, but I was already a bit of a tech enthusiast. I knew how the data looked, so I prepared a dummy mockup file because I could not upload my company’s data. (37:03)

Experimenting with ChatGPT and coding the first AI finance prototype

Anusha: I prepared mockup data for some XYZ company and loaded it into ChatGPT. I asked AI to transform it and understand the patterns. It did pretty well from my perspective. I would give it 55 or 60 points out of 100. (37:30)

Anusha: From my perspective that already meant saving 60 percent of my time, especially during the difficult forecasting and budgeting hours. Then I gave it two different sets of data from two different sources, like sales orders data and data records. I asked it to form a bridge between the two and it did really well. I felt it was like magic at that moment. I thought someone can actually do this. (38:06)

Alexey: Was it writing code for you? Or how was it doing it? Was it code? (38:54)

Anusha: Yes. I learned basic Python and it was mostly writing code I would say. Before, you could not just upload spreadsheets to ChatGPT and get an analysis. Now you can see the code if you want, but you can also look only at the final result. You can give feedback like this is not what I meant or this is good. (39:00)

Alexey: So you do not even need to know Python now, but before the tool did not exist. You needed to get the code, copy it and execute it. You also had to prepare the environment. For someone technical, installing Python is easy, but for someone in the finance department it is very hard. (39:31)

Anusha: I was lucky because my husband is a software engineer. He works in a different tech stack but he knows Python. He helped me install an open source model and he shared videos so I could learn. At that time I also had my baby and I was on maternity leave. (39:59)

Anusha: I used that time really well to learn everything, including some technical skills. By the time white coding started, it opened a new world for me. As a nontechnical person, with the help of white coding I could do whatever I wanted. It was a real wow moment for me. (40:27)

Alexey: That is cool. What did you use? Was it just ChatGPT or did you also start using tools like Cursor? (40:55)

Anusha: I used ChatGPT, Cursor and Lovable. At that time Lovable was still very small. I met someone who told me they had released version one and I think I was the first non-technical user who tried it. I also used Cursor. For UI and UX I used Framer and started preparing mockups. (41:02)

Alexey: Can you say it again (41:28)

Alexey: Framer is really good for understanding UI and UX. When I want to understand how my UI may look, I usually ask Lovable to design it and it creates a clickable mockup. It helps me see whether this is how I imagine the idea or if I want to iterate. Then I iterate directly in Lovable and ask it to place elements in different positions. It does not always do the best job because it needs to change the code, so I thought maybe there are better solutions focused more on the design stage than on the coding stage. (41:33)

Alexey: Maybe Framer is the answer. I need to check it. Thanks for sharing that. (41:40)

Anusha: What I used to do is first put my idea on paper because I am a pen and paper person. I drew how I would feel comfortable using any tool and where each element should be. Then I designed it in Framer and sent the Framer screenshot to Lovable. Lovable translated it into UI and UX. That worked quite well. (41:45)

Alexey: And you figured all this out by yourself? (41:52)

Anusha: I figured out everything by myself with no help from anyone. After I met Dimma and showed him my prototype, he told me it was already a product and that we did not need to build anything from scratch. (41:58)

Alexey: So Dimma is your co-founder? (43:06)

Anusha: Yes, Dimma is my co-founder. (43:12)

Alexey: Maybe tell us how this started. You had an idea, you built a prototype, and you believed you could transform the way things were done. What exactly did you build in this mockup before looking for co-founders? (43:12)

Identifying finance’s biggest pain points through user research

Anusha: I started with identifying finance’s biggest pain points through user research and surveying. I knew my own pain points as a head of finance and as a finance business partner, but I needed to check whether others had the same issues or if it was only my case. I needed this understanding before making any decisions about leaving my job or switching roles. So I started surveying people in my field. (43:34)

Anusha: I have many friends in finance who work in different roles, from accountants and analysts to managers, senior managers and CFOs. I interviewed them with simple questions about what they do in real life as accountants or controllers or CFOs. I asked them to walk me through their typical day and tell me their biggest one or two pain points. I asked why those problems happen from their perspective and what they would solve first if given a magic wizard. These were basic friendly questions. (44:00)

Anusha: I interviewed many people and 70 percent gave me three big points that were common across groups. When I asked accountants, one common point appeared. When I asked controllers, one common point appeared. From this I understood there were three major pain points that needed to be solved. Then I created mockups for these three. (44:57)

Alexey: What were the problems? (45:30)

Anusha: The first was the reconciliations problem, which takes away 50 to 60 percent of an average accountant’s time. I transformed reconciliations into a tool that takes data from different sources, interprets how it looks and produces an output. Once it is done it has to go back into the ERP system without manual intervention. So I built a prototype showing how this could work. (45:36)

Anusha: The second prototype was consolidation. Management wants to see consolidated reports at the top level and every entity uploads data from around the world. There are many problems in this process because the tools are rigid and not scalable, and the scalable ones are very expensive. I explored how to solve this. (46:18)

Anusha: The third problem was the most common among all finance directors and CFOs I spoke to, and it was the problem of converting data into insights. I faced the same challenge in my strategic role. When I created this mockup and later the prototype, this was the one everyone loved. Out of the three, we decided to start with this one because you cannot develop everything at once. (46:53)

Empowering finance teams with AI-driven, real-time decision insights

Alexey: How does it look? Do you upload a bunch of files and it tells you what is interesting? (47:24)

Anusha: It connects to your ERP and takes data from your CRM systems, expense tools and T and E systems. It pulls the key data, interprets it and surfaces insights. The goal is not to duplicate data but to make existing systems finally usable for decision making. That is why I call it augmenting the stack, not automating the stack. One example use case is forecasting. (47:36)

Anusha: Let us take the shoes example. Suppose Alex forecasted that he would sell one million units worth of shoes in a month. In the middle of the month the AI sees that only 100,000 have been sold. It then checks the order pipeline in the CRM system to see if there are orders to be converted. If there are none it checks the invoicing module to see if anything is ready to be invoiced. (48:16)

Anusha: It also checks manufacturing stages. Then it gives you an insight saying you have a risk of not meeting your forecast because nothing is in the sales pipeline or invoicing stage. It warns you about the potential impact on cash flow and working capital. That is what an insight looks like. (48:56)

Alexey: How does the system know what to look for? Do you define it using your domain knowledge? (49:40)

Anusha: Yes. We have an onboarding process to understand the context of the company because every company is different. We learn what the company does, its revenue sources and whether external data like FX or raw material prices affects it. We identify competitors and other context. Then we understand the company’s pain points and train the AI to look for those patterns and deliver outputs. (49:47)

Developing an entrepreneurial mindset through strategy and learning

Alexey: When you were telling your story it seemed you understood what you were doing from the beginning. You had an idea and wanted to check whether it was relevant. You wanted to help people solve their problems rather than blindly building something and hoping it would work. Did you know how to do this? Where did the knowledge come from? Did you study business? (50:59)

Anusha: CGMA is a chartered global management accountant qualification, one of the highest in finance for management roles. You learn strategic management there. Surveying is also important within companies, especially in change management, to understand what to change, how to change and when to change. It is a powerful tool to see things from employees’ or the organization’s perspective. For me it feels natural. (51:33)

Anusha: Before even thinking about starting a company, I researched and listened to many podcasts. Thanks to people like you who create amazing podcasts, I gained a holistic view of whether such a tool is possible and whether the technology is ready. I also learned what the biggest failures and challenges are for founders. Many founders said they missed understanding market signals or built tools for themselves instead of real users. (52:33)

Alexey: That happens to many developers. We solve our own problem but nobody else needs it. So podcasts helped you avoid that. (53:29)

Anusha: Yes exactly. That was one part of how I started the journey. After that it is a mix of gut-based and data-based decisions. You look at who is doing what in the market, how you can be different and what your unique selling point could be. When the time comes you need to clearly explain that. That is where the idea of surveying came from. (53:53)

Essential resources and finding the right AI co-founder

Alexey: Do you remember any podcast names you listened to? (54:31)

Anusha: I listened to Y Combinator podcasts which are very good, especially founder-to-founder. On the tech side I listened to Prediction Machines by Ajay Agrawal and Joshua Gans. On the human side I love The Fearless Organization by Amy Edmondson because it is about creating psychological safety, which is relevant for finance teams who fear being replaced by AI. (54:37)

Anusha: I also love Naval Ravikant for unit economics and strategy. And of course DataTalks.Club to understand conversations about data and how you transform data into insights. I have listened to your podcast too. They are very pleasant to hear. (55:21)

Alexey: Last question before we finish. How did you meet Dmitri? He is a friend of mine. He lives in Berlin and you live in Hamburg. How did that happen? (55:51)

Anusha: We met some years ago but did not know much about each other then. He was already a founder of another company because he is a serial founder. I met him at a meetup. Later when I was searching for a co-founder there was a co-founder connect event in Berlin about half a year ago. We met again there. (56:08)

Anusha: We talked and I told him about the idea. He said he was also looking for an opportunity. He is great in AI and ML and MCPs. We immediately clicked and decided to start. He already had something in mind related to using AI for finance, though from a slightly different angle. (56:47)

Alexey: That is cool that you met. (57:22)

Anusha: Yes, it was random. We clicked right away and it was fun to work with him. We registered our company together. (57:30)

Alexey: Where did you register, Berlin or Hamburg? (57:41)

Anusha: In Hamburg. (57:48)

Alexey: What is this co-founder connect event? Some sort of school? (57:48)

Anusha: It is called SIBB Co-Founder Meet. You register and they run an algorithm based on questions about what you want in a co-founder. It suggests around five to ten people you can meet. Then you meet them and talk. (57:54)

Anusha: The algorithm worked well for me because I met the right person. I know some programs where you give equity to the school, but this one was free. Before meeting Dimma I met another guy at a hackathon. My suggestion for non-technical people who want to meet technical people is to go to hackathons. Do not worry if you are non-technical. Many great developers would love to talk and transform ideas into products. (58:20)

Alexey: We have to wrap up. I really enjoyed this conversation. It is always interesting to talk to people who decide to start something on their own. I did not even notice how time passed but we have to finish. (59:54)

Alexey: Before we end, do you have any course, video or book that you would recommend that had an impact on you when starting a company or deciding on your idea? (1:00:12)

Anusha: Yes. One book I loved is The Almanac of Naval Ravikant. It is great. The podcasts I mentioned also helped me a lot. (1:00:38)

Anusha: To develop an entrepreneurial mindset, the famous book Rich Dad Poor Dad also helped me. It changes your mindset from being an employee to being an entrepreneur. It transformed how I think. (1:00:59)

Alexey: This is not the first time I hear this recommendation. I need to check that book. (1:01:38)

Anusha: It explains how a rich dad thinks like an entrepreneur and how a poor dad thinks like an employee. There is probably an audio version. It is very famous. (1:01:46)

Alexey: I need something interesting on Audible. Thanks. (1:01:58)

Alexey: Anusha, thanks a lot for joining us today and sharing your story. I wish you all the best on your journey. We will meet soon at Web Summit in Lisbon. I am looking forward to that. (1:02:05)

Anusha: It was amazing talking to you Alexey. I am so happy my first podcast was such a cool conversation. See you soon. (1:02:23)