NovaFinance.AI Year-End Report 2025: The Transition from Generative AI to Agentic Finance
By: Tymur Chalbash
As 2025 concludes, the financial sector appears to have moved beyond the initial enthusiasm surrounding generative AI and into a phase of structural integration. The dominant theme of this year is not merely better chatbots or faster document drafting—it is the emergence of agentic AI systems capable of executing multi-step workflows with limited human supervision.
Finance, more than any other industry, is suited to this evolution. It operates on structured data, defined rules, compliance obligations, and repeatable workflows. In that sense, AI is not replacing finance professionals—it is becoming the new operating layer beneath them.
- AI-Native Accounting: The Erosion of the Manual Close
For decades, the “monthly close” has been one of the most labor-intensive accounting processes. Reconciliation, journal entries, exception handling, and reporting traditionally required days or even weeks of manual coordination.
A new generation of AI-native ERP platforms is challenging that model. Startups such as Rillet have positioned themselves as alternatives to legacy accounting systems that function primarily as data repositories rather than intelligent systems.
In 2024, Rillet announced a $70 million Series B round reportedly led by Andreessen Horowitz and ICONIQ. While performance claims vary by client, early adopters report significant reductions in financial close timelines—some citing compression from multi-week cycles to under one business week.
More important than the time reduction itself is the architectural shift:
- Continuous transaction classification
- Real-time anomaly detection
- Automated accrual suggestions
- Exception routing to human review
From my perspective, the real breakthrough is not “automation percentage” (often marketed as 80% or higher), but the transition from periodic accounting to continuous accounting. When AI agents monitor books in real time, the concept of a stressful month-end close becomes structurally obsolete.
If this trajectory continues, by 2027 we may see companies treating financial reporting as a live dashboard rather than a historical snapshot.
- Institutional Integration: JPMorgan’s AI Operating Model
Large financial institutions historically adopt technology cautiously due to regulatory constraints. Yet in 2025, JPMorgan Chase has demonstrated one of the most visible enterprise-scale AI deployments.
The bank has long reported annual technology investments exceeding $15 billion, part of which is allocated to AI research and deployment. Its internal “LLM Suite” has reportedly been rolled out across major divisions, supporting:
- Investment research drafting
- Legal document analysis
- Internal compliance reviews
- Data summarization
While public reporting does not confirm precise user counts or ROI percentages, executives have emphasized measurable productivity gains and operational leverage.
What is structurally important is not chatbot usage—it is workflow orchestration. The shift toward “connected intelligence” suggests AI systems are increasingly embedded into reporting chains, analytics pipelines, and risk management frameworks rather than sitting on the periphery as productivity assistants.
In my assessment, banks will not publicly quantify AI ROI aggressively for competitive and regulatory reasons. However, institutions that successfully integrate agentic systems into risk modeling and regulatory reporting will gain cost advantages that compound over time.
AI in banking is no longer experimental—it is infrastructural.
- Quant Meets Frontier AI: The DeepSeek Signal
One of the more surprising narratives of 2025 has been the visibility of quant-finance-backed AI research.
The lab DeepSeek, reportedly linked to figures associated with High-Flyer Capital, released its R1 model in early 2025. The release attracted attention for its reasoning benchmarks and cost-efficiency claims relative to Western frontier models.
While app store ranking claims vary by region and timing, the broader signal is significant:
quantitative trading infrastructure—historically optimized for latency, scale, and distributed computing—is increasingly intersecting with general AI development.
This convergence suggests a new competitive dynamic. Financial engineering expertise is being redirected toward model training efficiency and inference optimization. If compute-efficient reasoning models proliferate, AI cost structures may compress rapidly, increasing accessibility across mid-sized financial firms.
From a strategic standpoint, the financial sector may become both a major consumer and builder of advanced AI systems.
- AI in Crypto and DeFi: Defense at Scale
With cryptocurrency markets experiencing renewed volatility and growth cycles, security remains paramount.
Major platforms such as PayPal and Binance deploy machine learning models to monitor transaction flows, detect suspicious behavioral clusters, and identify emerging fraud typologies—including so-called “pig butchering” scams.
AI’s value in crypto lies in pattern recognition across millions of wallet interactions—far beyond human analytical capacity.
Additionally, AI-assisted smart contract auditing tools have matured. While fully autonomous auditing remains aspirational, current systems can:
- Identify common vulnerability patterns
- Flag reentrancy risks
- Detect unusual gas consumption logic
- Simulate adversarial interactions
In decentralized finance (DeFi), where billions of dollars can be locked in immutable code, AI-assisted pre-deployment review is becoming a risk management standard rather than a luxury.
Structural Shift: From Tools to Agents
The defining transformation of 2025 is the shift from AI as a tool to AI as a workflow executor.
Generative AI drafts memos.
Agentic AI:
- Pulls relevant datasets
- Cross-checks compliance rules
- Generates draft outputs
- Routes approvals
- Archives documentation
This is a different operational paradigm.
Finance is rule-based, data-rich, and high-frequency—ideal conditions for agentic systems. Unlike creative industries, where subjectivity dominates, financial processes often follow explicit regulatory and accounting frameworks.
From my perspective, three structural outcomes are likely:
- Headcount will not collapse—but role composition will shift.
Junior roles focused on reconciliation and document preparation will decline, while oversight, interpretation, and systems governance roles will grow. - Audit and compliance will become AI-supervised ecosystems.
Human auditors will increasingly validate AI decisions rather than manually inspect transactions. - Competitive differentiation will move from “having AI” to “orchestrating AI.”
The firms that build cohesive agent ecosystems—rather than isolated deployments—will dominate.
Conclusion: AI as Finance’s Operating System
In 2023, AI in finance meant chat interfaces and productivity enhancements.
In 2024, it meant pilot programs.
In 2025, it increasingly means embedded infrastructure.
The financial industry is not witnessing hype—it is witnessing systems redesign.
Agentic AI is becoming an invisible layer beneath accounting closes, regulatory reporting, fraud detection, and quantitative research. While some adoption metrics remain promotional and precise ROI figures are rarely disclosed publicly, the directional shift is undeniable.
Finance has always been a technological frontier—from electronic trading to algorithmic risk modeling. The transition to agentic AI may represent the next foundational upgrade.
If current trajectories hold, 2025 may ultimately be remembered not as the year of better chatbots—but as the year finance quietly rewired itself around autonomous systems.
Sources & References
Corporate Disclosures & Company Announcements
- Rillet
- Series B Funding Announcement (Reported $70M round, 2024–2025)
- Company blog and investor communications
- Executive interviews discussing AI-native accounting architecture
- JPMorgan Chase
- Annual Reports (Form 10-K, 2024–2025)
- Investor Day Presentations on Technology Strategy
- Earnings Call Transcripts discussing AI and digital transformation initiatives
- DeepSeek
- DeepSeek-R1 Technical Report: Training Reasoning Models via Reinforcement Learning (2025)
- Official model release documentation and benchmark disclosures
- High-Flyer Capital
- Public corporate background materials and leadership biographies
- PayPal
- Annual Reports (Form 10-K)
- Risk and Security Infrastructure Disclosures
- Public statements regarding AI-based fraud prevention systems
- Binance
- Security and Compliance Reports
- Public blog posts on machine learning-driven risk monitoring
Industry Research & Consulting Reports
- McKinsey & Company
- The State of AI in 2024/2025: From Pilots to Enterprise Value
- AI adoption surveys across financial services
- PwC
- AI in Financial Services Reports
- Digital Finance Transformation Outlook
- Deloitte
- Banking & Capital Markets AI Adoption Reports
- Tech Trends in Financial Services
- KPMG
- Global Banking Technology Surveys
- AI Risk & Governance Research
Regulatory & Institutional Context
- U.S. Securities and Exchange Commission
- Public filings (10-K, 8-K disclosures referencing AI strategy)
- Regulatory commentary on AI governance and risk oversight
- Bank for International Settlements
- BIS Papers on Artificial Intelligence in Banking
- Basel Committee digital risk supervision guidelines
Market & Crypto Research
- Chainalysis
- Crypto Crime Reports (2024–2025)
- Analysis of “pig butchering” scams and fraud typologies
- CoinMetrics
- Network Activity Reports
- On-chain volume statistics
- Messari
- DeFi Security & Smart Contract Risk Reports
Technical & Academic Context
- OpenAI
- Model system cards and reasoning benchmarks
- Research on reinforcement learning and reasoning models
- Stanford University — AI Index Report 2024/2025
- Global AI investment trends
- Enterprise AI deployment data
- World Economic Forum
- Reports on AI in Financial Services and Risk Governance
Year-End Report 2025: The Transition from Generative AI to Agentic Finance
As 2025 concludes, the financial sector appears to have moved beyond the initial enthusiasm surrounding generative AI and into a phase of structural integration. The dominant theme of this year is not merely better chatbots or faster document drafting—it is the emergence of agentic AI systems capable of executing multi-step workflows with limited human supervision.
Finance, more than any other industry, is suited to this evolution. It operates on structured data, defined rules, compliance obligations, and repeatable workflows. In that sense, AI is not replacing finance professionals—it is becoming the new operating layer beneath them.
- AI-Native Accounting: The Erosion of the Manual Close
For decades, the “monthly close” has been one of the most labor-intensive accounting processes. Reconciliation, journal entries, exception handling, and reporting traditionally required days or even weeks of manual coordination.
A new generation of AI-native ERP platforms is challenging that model. Startups such as Rillet have positioned themselves as alternatives to legacy accounting systems that function primarily as data repositories rather than intelligent systems.
In 2024, Rillet announced a $70 million Series B round reportedly led by Andreessen Horowitz and ICONIQ. While performance claims vary by client, early adopters report significant reductions in financial close timelines—some citing compression from multi-week cycles to under one business week.
More important than the time reduction itself is the architectural shift:
- Continuous transaction classification
- Real-time anomaly detection
- Automated accrual suggestions
- Exception routing to human review
From my perspective, the real breakthrough is not “automation percentage” (often marketed as 80% or higher), but the transition from periodic accounting to continuous accounting. When AI agents monitor books in real time, the concept of a stressful month-end close becomes structurally obsolete.
If this trajectory continues, by 2027 we may see companies treating financial reporting as a live dashboard rather than a historical snapshot.
- Institutional Integration: JPMorgan’s AI Operating Model
Large financial institutions historically adopt technology cautiously due to regulatory constraints. Yet in 2025, JPMorgan Chase has demonstrated one of the most visible enterprise-scale AI deployments.
The bank has long reported annual technology investments exceeding $15 billion, part of which is allocated to AI research and deployment. Its internal “LLM Suite” has reportedly been rolled out across major divisions, supporting:
- Investment research drafting
- Legal document analysis
- Internal compliance reviews
- Data summarization
While public reporting does not confirm precise user counts or ROI percentages, executives have emphasized measurable productivity gains and operational leverage.
What is structurally important is not chatbot usage—it is workflow orchestration. The shift toward “connected intelligence” suggests AI systems are increasingly embedded into reporting chains, analytics pipelines, and risk management frameworks rather than sitting on the periphery as productivity assistants.
In my assessment, banks will not publicly quantify AI ROI aggressively for competitive and regulatory reasons. However, institutions that successfully integrate agentic systems into risk modeling and regulatory reporting will gain cost advantages that compound over time.
AI in banking is no longer experimental—it is infrastructural.
- Quant Meets Frontier AI: The DeepSeek Signal
One of the more surprising narratives of 2025 has been the visibility of quant-finance-backed AI research.
The lab DeepSeek, reportedly linked to figures associated with High-Flyer Capital, released its R1 model in early 2025. The release attracted attention for its reasoning benchmarks and cost-efficiency claims relative to Western frontier models.
While app store ranking claims vary by region and timing, the broader signal is significant:
quantitative trading infrastructure—historically optimized for latency, scale, and distributed computing—is increasingly intersecting with general AI development.
This convergence suggests a new competitive dynamic. Financial engineering expertise is being redirected toward model training efficiency and inference optimization. If compute-efficient reasoning models proliferate, AI cost structures may compress rapidly, increasing accessibility across mid-sized financial firms.
From a strategic standpoint, the financial sector may become both a major consumer and builder of advanced AI systems.
- AI in Crypto and DeFi: Defense at Scale
With cryptocurrency markets experiencing renewed volatility and growth cycles, security remains paramount.
Major platforms such as PayPal and Binance deploy machine learning models to monitor transaction flows, detect suspicious behavioral clusters, and identify emerging fraud typologies—including so-called “pig butchering” scams.
AI’s value in crypto lies in pattern recognition across millions of wallet interactions—far beyond human analytical capacity.
Additionally, AI-assisted smart contract auditing tools have matured. While fully autonomous auditing remains aspirational, current systems can:
- Identify common vulnerability patterns
- Flag reentrancy risks
- Detect unusual gas consumption logic
- Simulate adversarial interactions
In decentralized finance (DeFi), where billions of dollars can be locked in immutable code, AI-assisted pre-deployment review is becoming a risk management standard rather than a luxury.
Structural Shift: From Tools to Agents
The defining transformation of 2025 is the shift from AI as a tool to AI as a workflow executor.
Generative AI drafts memos.
Agentic AI:
- Pulls relevant datasets
- Cross-checks compliance rules
- Generates draft outputs
- Routes approvals
- Archives documentation
This is a different operational paradigm.
Finance is rule-based, data-rich, and high-frequency—ideal conditions for agentic systems. Unlike creative industries, where subjectivity dominates, financial processes often follow explicit regulatory and accounting frameworks.
From my perspective, three structural outcomes are likely:
- Headcount will not collapse—but role composition will shift.
Junior roles focused on reconciliation and document preparation will decline, while oversight, interpretation, and systems governance roles will grow. - Audit and compliance will become AI-supervised ecosystems.
Human auditors will increasingly validate AI decisions rather than manually inspect transactions. - Competitive differentiation will move from “having AI” to “orchestrating AI.”
The firms that build cohesive agent ecosystems—rather than isolated deployments—will dominate.
Conclusion: AI as Finance’s Operating System
In 2023, AI in finance meant chat interfaces and productivity enhancements.
In 2024, it meant pilot programs.
In 2025, it increasingly means embedded infrastructure.
The financial industry is not witnessing hype—it is witnessing systems redesign.
Agentic AI is becoming an invisible layer beneath accounting closes, regulatory reporting, fraud detection, and quantitative research. While some adoption metrics remain promotional and precise ROI figures are rarely disclosed publicly, the directional shift is undeniable.
Finance has always been a technological frontier—from electronic trading to algorithmic risk modeling. The transition to agentic AI may represent the next foundational upgrade.
If current trajectories hold, 2025 may ultimately be remembered not as the year of better chatbots—but as the year finance quietly rewired itself around autonomous systems.
Sources & References
Corporate Disclosures & Company Announcements
- Rillet
- Series B Funding Announcement (Reported $70M round, 2024–2025)
- Company blog and investor communications
- Executive interviews discussing AI-native accounting architecture
- JPMorgan Chase
- Annual Reports (Form 10-K, 2024–2025)
- Investor Day Presentations on Technology Strategy
- Earnings Call Transcripts discussing AI and digital transformation initiatives
- DeepSeek
- DeepSeek-R1 Technical Report: Training Reasoning Models via Reinforcement Learning (2025)
- Official model release documentation and benchmark disclosures
- High-Flyer Capital
- Public corporate background materials and leadership biographies
- PayPal
- Annual Reports (Form 10-K)
- Risk and Security Infrastructure Disclosures
- Public statements regarding AI-based fraud prevention systems
- Binance
- Security and Compliance Reports
- Public blog posts on machine learning-driven risk monitoring
Industry Research & Consulting Reports
- McKinsey & Company
- The State of AI in 2024/2025: From Pilots to Enterprise Value
- AI adoption surveys across financial services
- PwC
- AI in Financial Services Reports
- Digital Finance Transformation Outlook
- Deloitte
- Banking & Capital Markets AI Adoption Reports
- Tech Trends in Financial Services
- KPMG
- Global Banking Technology Surveys
- AI Risk & Governance Research
Regulatory & Institutional Context
- U.S. Securities and Exchange Commission
- Public filings (10-K, 8-K disclosures referencing AI strategy)
- Regulatory commentary on AI governance and risk oversight
- Bank for International Settlements
- BIS Papers on Artificial Intelligence in Banking
- Basel Committee digital risk supervision guidelines
Market & Crypto Research
- Chainalysis
- Crypto Crime Reports (2024–2025)
- Analysis of “pig butchering” scams and fraud typologies
- CoinMetrics
- Network Activity Reports
- On-chain volume statistics
- Messari
- DeFi Security & Smart Contract Risk Reports
Technical & Academic Context
- OpenAI
- Model system cards and reasoning benchmarks
- Research on reinforcement learning and reasoning models
- Stanford University — AI Index Report 2024/2025
- Global AI investment trends
- Enterprise AI deployment data
- World Economic Forum
- Reports on AI in Financial Services and Risk Governance
