Expert Insights | Earley Information Science

Earley AI Podcast - Episode 98 Agentic AI in Finance, the Trusted Advisor Advantage, and the Pricing Model Reckoning with Nikita Komarov

Written by Earley Information Science Team | Sep 9, 2026, 8:02:01 PM

What It Takes to Build AI That Is Accurate Enough, Traceable Enough, and Trustworthy Enough for High-Stakes Financial Work

Guest: Nikita Komarov, CEO and Founder at Dobs.AI

Host: Seth Earley, CEO at Earley Information Science

Published on: September 9, 2026

 

In this episode, Seth Earley speaks with Nikita Komarov, CEO and Founder of Dobs.AI, who spent seven years at McKinsey advising Fortune 1000 executives before founding a company that is rebuilding financial due diligence, internal audit, and vendor overpayment recovery from the ground up as agentic AI systems. They explore why financial professionals are the most resistant to AI adoption and why that resistance is rational, how orchestrating teams of AI agents with financial controls built in produces outputs that are deterministic enough for audit, why the difference between an efficiency tool and a production-ready AI system is enormous, and how the trusted advisor status accountants have built over decades becomes a platform for entirely new services in the AI era.

Key Takeaways:

  • Financial professionals are among the most resistant to AI adoption for a rational reason - LLMs are non-deterministic by nature, and accounting requires numbers that are 100% accurate and traceable.
  • Building production-grade financial AI requires three levers working together: orchestrating teams of agents with defined roles, building financial controls and guardrails into the pipeline, and solving for data extraction accuracy before any analysis begins.
  • The difference between an efficiency tool like Claude or ChatGPT and a production-ready AI system is not the model - it is the architecture, the controls, and the product thinking required to get from unstructured input to a final output a human can take to a client.
  • DOBS AI compresses financial due diligence from a six-week engagement to 72 hours for the management meeting - cutting the cycle from week and a half to three days on that critical milestone alone.
  • Accounting firms have more trust with clients than management consultants or lawyers, and that trust combined with recurring access creates a platform for expanding into advisory services that AI now makes possible.
  • The pricing model reckoning is real - time and materials no longer makes sense when AI does the work in hours, and firms need to shift to value-based pricing anchored to the outcome delivered, not the hours spent.
  • The long-term trajectory is positive, but the mid-term transition is the risk - AI is compressing decades of technological change into five to ten years, and organizations and individuals who are not adapting will be left behind.

Insightful Quotes:

"Large language models, they predict the next word. That's why these systems are non-deterministic. You can't say what the output will be next. That's the problem in financial services - you need 100% accuracy, but you don't know what the system is going to tell you." - Nikita Komarov

"That's exactly the difference between an efficiency tool and a production-ready solution. When people say we use AI, they most likely mean Copilot or ChatGPT - and that's 5 to 10% of what's actually possible." - Nikita Komarov

"You can't automate what you don't understand. The first thing you have to do is say, what is the expected output and the outcome, and then how do I verify that I actually get there?" - Seth Earley

Tune in to discover why financial AI is one of the most demanding and highest-stakes applications in the enterprise - and what it actually takes to build systems that are accurate and auditable enough to trust.

Links

LinkedIn: https://www.linkedin.com/in/nikita-komarov/

Website: https://dobs.ai

Ways to Tune In:

Earley AI Podcast: https://www.earley.com/earley-ai-podcast-home Apple Podcast: https://podcasts.apple.com/podcast/id1586654770 Spotify: https://open.spotify.com/show/5nkcZvVYjHHj6wtBABqLbEiHeart Radio: https://www.iheart.com/podcast/269-earley-ai-podcast-87108370/ Stitcher: https://www.stitcher.com/show/earley-ai-podcast Amazon Music: https://music.amazon.com/podcasts/18524b67-09cf-433f-82db-07b6213ad3ba/earley-ai-podcast Buzzsprout: https://earleyai.buzzsprout.com/

 

Podcast Transcript: Agentic AI in Finance, the Trusted Advisor Advantage, and the Pricing Model Reckoning

Transcript introduction

This transcript captures a conversation between Seth Earley and Nikita Komarov of Dobs.AI about what it takes to build AI that financial professionals can actually trust - not just use. They cover why hallucination is a solvable engineering problem, how orchestrating teams of agents with built-in financial controls produces outputs that are traceable to the source, the critical difference between an efficiency tool and a production-ready system, how DOBS AI is compressing due diligence cycles from weeks to days, what the junior talent pipeline looks like when AI handles foundational work, why accountants have a structural advantage over consultants and lawyers in the AI era, and what the mid-term workforce risk of rapid AI adoption actually is.

Transcript

Seth Earley: Welcome to the Earley AI Podcast, brought to you by VKTR. I'm Seth Earley, your host, and in each episode we talk about how AI and data are transforming business operations and business strategy throughout the industry and across industries. Today, we're talking about what happens when you take one of the most time-intensive, high-stakes financial analysis workflows - the kind of work that would take weeks and a team of CPAs - and rebuild it from the ground up as an agentic AI system. Financial due diligence, internal audit, vendor overpayment recovery, not assisted by AI, but done by AI with humans overseeing the output. Joining me today is Nikita Komarov, CEO and Founder of Dobs.AI. Nikita spent seven years at McKinsey advising Fortune 1000 executives on digital transformation, and worked with private equity on commercial due diligence and portfolio transformation before founding DOBS. His company serves Fortune 1000 enterprise clients as well as top accounting firms looking to automate their core client services. Nikita, welcome to the show.

Nikita Komarov: Thank you, Seth, for hosting me. Excited for this.

Seth Earley: Let's talk about common misconceptions about AI in finance and accounting. When you talk to CFOs, accounting firm partners, and financial executives about AI, what do they most consistently get wrong?

Nikita Komarov: Financial professionals are among the most conservative in the market, and I understand why - I have been in their shoes. I am a company auditor by training. You work with client data, the numbers need to be 100% precise.

When you talk about AI specifically, what is it? Large language models predict the next word. That is why these systems are non-deterministic - you cannot say what the output will be next. It is a kind of black box. That is the problem in financial services. You need 100% accuracy, but you do not know what the system is going to tell you. It can hallucinate. Every AI solution, based on all of our experience over the last couple of years, is helpful but sometimes provides something you cannot rely on 100%.

That is one of the biggest concerns of financial professionals about how to leverage AI in their work. But here is the thing: it is just about the way you use AI. If you use Copilot, ChatGPT, and similar tools, you will see hallucinations from time to time. But to build a really production-ready system you can rely on is much more complicated. That is what we focus on - taking the flexibility of AI systems while providing the accuracy you need so you can trust the number.

Seth Earley: So how do you put guardrails around it? If it is probabilistic and not deterministic, and you need deterministic answers - financial systems require the numbers to add up - how do you take something that is inherently statistical and make it bulletproof from the perspective of financial audit and controls?

Nikita Komarov: That is our superpower, and that is where we spend a lot of time building IP around. Several things.

First, we orchestrate teams of agents. Think about financial due diligence - you typically have a director, a manager, and analysts running it. We reproduce this in AI. You have a director who oversees the scope and the core analysis. You have a manager who decomposes the task. You have analysts who execute, and you can run 50, 60, 100 cycles until the output is ready for human review. That is one lever - iterations and compute.

Second is controls and guardrails. The same way I would work with my team - I am the manager, I have an analyst, I have a set of checkpoints for how I review their work. This number should reconcile here, the total should be zero here. I put a set of controls and guardrails that I give the agents to follow.

Third is data quality. In financial diligence, every deal is unique - you can get messy files. We spend a lot of time building models for data extraction. How do you digest 500 pages of scanned PDF bank statements, or millions of rows of general ledger data in a legacy TXT format? If you just put a PDF into Claude, it will extract some numbers, but you will never understand the accuracy of the extraction. We measure this specifically. Those three levers are the core components.

Seth Earley: The fourth element is specific tools - because LLMs are not good at math.

Nikita Komarov: Exactly. Large language models are a good reasoning layer - they can orchestrate other agents, make judgments, read context. But for precise math calculations, you need code that executes the numbers. Most of our analysis is Excel-based with formulas, so every number is traceable and the formulas verify themselves. Understanding the process, picking the right tools for every process step, and then having humans look at the critical controls and checks - that is the full picture.

Seth Earley: The difference between an efficiency tool and a production-ready system is significant. What does it mean to build an AI employee versus an AI tool that assists a human?

Nikita Komarov: With an efficiency tool like Claude Code, you prompt back and forth. Every time the output is different. You can standardize some prompts, but there is no consistency, and every edge case requires manual intervention.

With an AI employee, we take all your input data and provide you the final output. It can be a 70-page Excel workbook, everything formula-based - proof of cash, expense analysis, working capital, revenue waterfalls, EBITDA adjustments, vendor and customer concentrations. We provide you the final output that you can take to the client. It is ready for human review. You do not need to interact back and forth with the AI to get there.

For larger teams, we help standardize best practices - what the template looks like, how formatting should be done, what typical analyses should be run. That keeps a consistent level of quality and granularity, helps onboard new employees, and helps train junior people. The system drives the work all the way to a ready result where you can then add your judgment as the professional.

When people say we use AI, they most likely mean Copilot or ChatGPT - and that is 5 to 10% of what is actually possible. Building a production-ready system means taking all those components, teaching them so the system can validate itself, and orchestrating the right architecture so you get a final output, not just back-and-forth prompting.

Seth Earley: Walk us through how you compress financial due diligence from six weeks to hours, and what it takes to make everything traceable.

Nikita Komarov: A typical deal takes four to six weeks. The first week is data exploration and data book preparation. Close to the end of the second week, you typically have a management meeting.

We can take the entire data room - all the unstructured documents received from the client - and in three to four hours produce the data book that is ready for the management meeting. Most of our clients cut the cycle from a week and a half to 72 hours to reach that management meeting milestone. That accelerates the deal.

After that, any changes, adjustments, or roll-forwards are automated. You provide additional data, or upload a transcript of your call, and the system automatically incorporates those changes. It also identifies data gaps and manages back-and-forth with the target - because one of the biggest frustrations private equity firms have is sending the same data requests multiple times. We eliminate that friction by tracking what has been provided, what is missing, and what is critical.

Seth Earley: The business model has to change as well. Time and materials no longer makes sense.

Nikita Komarov: For buyers, it does not matter whether they are paying 10% more or 30% more for the service. What matters is having an accurate output that surfaces discrepancies and EBITDA adjustments the seller may be obscuring. A $100,000 EBITDA adjustment at a 10x multiple is a million-dollar impact. That is why the value is in the outcome, not the hours.

Accounting firms need to shift to value-based pricing. That is sometimes hard because time and materials is how the industry has historically operated - many CFOs on the other side were previously at accounting firms and know exactly how these services have been priced. But the logic for change is clear: I would rather pay more and have you find those discrepancies than pay less for a report that missed them. The firms that lock in value-based pricing and then continuously deliver more insight will be in a very strong position.

Seth Earley: Accounting firms have always developed senior talent by running juniors through years of foundational work. If AI is now doing that foundational work, how does the next generation of managers, directors, and partners get built?

Nikita Komarov: It is going to be different, and I would say every profession is going through the same cycle. You still need junior people who will progress and become the next senior. But the pattern is different.

I want people to have ownership over what they deliver. I want them to use AI, but they should be the ones responsible for the final outcome. They should know how to orchestrate AI, how to verify it is working correctly, and how to learn much faster with it. Before, if you were an accounting analyst, you needed to do multiple deals yourself, in Excel, and get feedback over years. Right now, you can get much more exposure and iterate through many more deals in the same time, seeing the reasoning logic and learning from it. The educational and professional development system is being transformed. Every company and leadership needs to figure out how to retrain their people and reward the new talent - but it is a new way of learning, not the elimination of learning.

Seth Earley: Accountants have something that consultants and lawyers sometimes do not - deep, trusted, recurring access to the full financial picture of a client. How does AI change what is possible with that access?

Nikita Komarov: There is survey data showing that accountants have significantly more trust than management consultants or lawyers. That opens enormous opportunities. Consultants have traditionally structured information, highlighted important pieces, and run certain analyses - but most of that work can now be done by AI.

What accountants have that consultants do not is recurring client relationships. Consulting is mostly one-time projects. Accounting firms work with clients every month, every year. With that access and recurring engagement, and with AI making them much more analytically capable, I see a major opportunity for accounting firms to become true advisory firms across a much wider range of services. They have the trust. They have the access. Now they have the tools.

Seth Earley: You believe the long-term outcome is positive, but you are concerned about the mid-term transition. What does that mean for organizations and individuals who are not adapting fast enough?

Nikita Komarov: In the long term, I believe we are going to live in a better world - more opportunities, more products and services, more value created. But I am concerned about the midterm. Previous technical revolutions unfolded over a hundred years across different parts of the world. With AI, we may compress that into five to ten years. Change management is important, and it is much harder to change how people work and how organizations are structured within that time frame. That is why I am concerned that many people will find themselves without work, not because there are fewer opportunities overall, but because the pace of change exceeds the ability to absorb it.

There is no opportunity to bury your head in the sand anymore. You need to embrace this. If you are not ready to embrace it, be prepared to be out of the market - because there are companies who want your market share, whether incumbents who are willing to move fast or AI-native companies that have rebuilt their business model from scratch. The best recommendation to any organization right now is to start today. It can be small use cases. But start. Your people need to learn how to work with these systems. The people who know how to leverage the technology are your most valuable asset.

Seth Earley: Nikita, thank you so much for joining me and sharing what it looks like to rebuild financial analysis from the ground up with agentic AI, and what it means for the professionals and organizations navigating that shift. And a final thank you to our sponsor, VKTR. To our listeners, thank you for tuning in to the Earley AI Podcast. Be sure to subscribe for more conversations on how AI is shaping the future of business.

Nikita Komarov: Thank you, Seth. That was a wonderful conversation. Happy to be a partner in your AI journey.