Stacks Raises $23M: Agentic AI Agents for Finance Automation

Stacks Raises $23M to Bring Agentic AI to Enterprise Finance

Stacks, a fintech startup focused on autonomous finance operations, has raised $23 million to scale its agentic AI platform for accounts payable, accounts receivable, reconciliation, and financial reporting. The round reflects a growing conviction among investors that finance departments represent one of the highest-value targets for AI agent deployment, combining high transaction volumes, rule-heavy processes, and measurable ROI with relatively low ambiguity compared to other enterprise functions.

Finance teams across enterprises spend enormous amounts of time on manual, repetitive tasks that follow well-defined rules but require judgment to handle exceptions. This combination makes finance operations an ideal candidate for agentic AI that can handle the routine autonomously while intelligently escalating exceptions to human reviewers.

The Problem with Finance Automation Today

Enterprise finance departments have invested heavily in automation over the past decade, deploying tools like robotic process automation (RPA), optical character recognition (OCR) for invoice processing, and enterprise resource planning (ERP) systems. Yet despite this investment, significant manual work persists.

flowchart LR
    CALLER(["Client or Lead"])
    subgraph TEL["Telephony"]
        SIP["Twilio SIP and PSTN"]
    end
    subgraph BRAIN["Financial Services AI<br/>Agent"]
        STT["Streaming STT<br/>Deepgram or Whisper"]
        NLU{"Intent and<br/>Entity Extraction"}
        TOOLS["Tool Calls"]
        TTS["Streaming TTS<br/>ElevenLabs or Rime"]
    end
    subgraph DATA["Live Data Plane"]
        CRM[("CRM and Notes")]
        CAL[("Calendar and<br/>Schedule")]
        KB[("Knowledge Base<br/>and Policies")]
    end
    subgraph OUT["Outcomes"]
        O1(["KYC pre-fill done"])
        O2(["Funding instructions sent"])
        O3(["Compliance officer<br/>escalation"])
    end
    CALLER --> SIP --> STT --> NLU
    NLU -->|Lookup| TOOLS
    TOOLS <--> CRM
    TOOLS <--> CAL
    TOOLS <--> KB
    NLU --> TTS --> SIP --> CALLER
    NLU -->|Resolved| O1
    NLU -->|Schedule| O2
    NLU -->|Escalate| O3
    style CALLER fill:#f1f5f9,stroke:#64748b,color:#0f172a
    style NLU fill:#4f46e5,stroke:#4338ca,color:#fff
    style O1 fill:#059669,stroke:#047857,color:#fff
    style O2 fill:#0ea5e9,stroke:#0369a1,color:#fff
    style O3 fill:#f59e0b,stroke:#d97706,color:#1f2937

The fundamental limitation of existing automation tools is their brittleness. RPA bots follow exact scripts and break when invoice formats change, when vendors use unexpected terminology, or when edge cases arise that were not anticipated during the bot's design. OCR systems achieve high accuracy on clean documents but struggle with handwritten notes, poor scans, and non-standard layouts. ERP systems provide structure but require manual data entry for information that arrives outside their standard input channels.

The result is that finance teams still spend substantial portions of their time on:

• Invoice processing and three-way matching between purchase orders, receiving reports, and invoices
• Exception handling for discrepancies in pricing, quantities, or terms
• Vendor communication to resolve disputes, request missing information, or confirm payment details
• Bank reconciliation matching thousands of transactions against internal records
• Month-end close procedures including accruals, adjustments, and report generation
• Audit preparation gathering documentation and responding to auditor queries

How Stacks' Agentic Approach Differs

Stacks' platform deploys AI agents that go beyond script-following automation. These agents understand the intent behind finance operations and can reason about exceptions, ambiguities, and novel situations that would break traditional automation tools.

Accounts Payable Agents process incoming invoices regardless of format, extracting relevant data, matching against purchase orders and receiving records, identifying discrepancies, and either approving for payment or routing exceptions to the appropriate reviewer with a clear explanation of the issue. Unlike OCR plus rules systems, these agents can handle unusual invoice formats, interpret handwritten notes, and even correspond with vendors to resolve missing information.

Accounts Receivable Agents manage the collections process by analyzing payment patterns, generating customized outreach for overdue accounts, processing incoming payments, applying cash to the correct invoices, and flagging unusual payment behavior. They can adapt their communication tone based on customer relationship value and payment history.

Reconciliation Agents match bank transactions against general ledger entries, identifying matches with high confidence, investigating potential matches that require judgment, and flagging genuinely unmatched items for human review. They learn from each reconciliation cycle, improving their matching accuracy over time as they encounter more transaction patterns.

Reporting Agents automate the generation of financial reports by pulling data from multiple sources, performing calculations, applying accounting standards, and generating narrative explanations of significant variances. They can answer ad-hoc questions about financial data in natural language, replacing the back-and-forth between business leaders and finance analysts.

The 60 Percent Workload Reduction Claim

Stacks reports that its enterprise customers are seeing approximately 60 percent reduction in manual finance workload after deploying its agents. This figure deserves scrutiny, as automation vendors frequently overstate efficiency gains. However, the claim is plausible given the nature of finance work being automated.

The breakdown of workload reduction typically follows this pattern:

• Invoice processing: 70 to 80 percent reduction through automated extraction, matching, and approval for straightforward invoices that represent the bulk of volume
• Exception handling: 30 to 40 percent reduction as agents resolve common exceptions autonomously, such as minor price discrepancies within tolerance thresholds or missing fields that can be inferred from context
• Reconciliation: 60 to 70 percent reduction through automated matching and pattern-based investigation of near-matches
• Reporting: 50 to 60 percent reduction through automated data gathering, calculation, and narrative generation
• Vendor communication: 40 to 50 percent reduction through automated query generation and response processing

The overall 60 percent figure represents a blended average across these categories, weighted by the time each activity consumes in a typical finance operation.

Finance Team Transformation

The $23 million raise is not just about automating existing processes. Stacks' vision extends to fundamentally transforming the role of finance teams within enterprises. As routine processing work is handled by agents, finance professionals can shift their focus toward:

• Strategic analysis examining business performance trends and recommending operational changes
• Business partnering working directly with operational teams to optimize financial outcomes
• Risk management identifying and mitigating financial risks proactively rather than reactively
• Process design architecting better financial workflows and controls
• Technology governance overseeing and optimizing AI agent performance

This shift mirrors what happened in manufacturing when automation displaced assembly-line tasks but created demand for higher-skilled roles in automation management, quality engineering, and process optimization.

Market Context and Competitive Landscape

Stacks enters a market that includes established players like SAP Concur, Coupa, and Bill.com alongside newer AI-native competitors. The competitive advantage of the agentic approach lies in handling the long tail of exceptions and edge cases that rule-based systems cannot address.

Traditional automation tools handle perhaps 60 to 70 percent of finance transactions end-to-end. The remaining 30 to 40 percent, which involve exceptions, ambiguities, and non-standard situations, still require human intervention. Agentic AI pushes automated handling to 85 to 90 percent by reasoning about exceptions rather than failing on them, and this incremental improvement in automation rate has outsized impact on operational efficiency.

Frequently Asked Questions

What specific finance tasks can Stacks' AI agents handle autonomously?

Stacks' agents handle invoice processing and three-way matching, accounts receivable collections and cash application, bank reconciliation, and financial report generation. They process invoices regardless of format, match transactions against records, manage vendor and customer communications, and generate reports with variance explanations. Common exceptions within defined tolerance thresholds are resolved autonomously.

How does agentic AI differ from RPA in finance automation?

RPA bots follow exact scripts and break when encountering unexpected formats or edge cases. Agentic AI agents understand the intent behind finance operations and can reason about exceptions, interpret ambiguous data, and handle novel situations. When an invoice has an unexpected format or a transaction does not match cleanly, agents can investigate and often resolve the issue without human intervention.

Is a 60 percent workload reduction realistic for finance teams?

The figure is a blended average across multiple finance functions. Invoice processing sees the highest reduction at 70 to 80 percent, while exception handling sees lower reduction at 30 to 40 percent. The overall number is achievable for enterprises with high transaction volumes and standardized processes, though organizations with highly complex or unusual financial operations may see lower initial reductions.

What happens to finance professionals when agents automate routine work?

Finance teams shift toward higher-value activities including strategic analysis, business partnering, risk management, and process design. Rather than eliminating roles, agentic AI typically transforms them, much like manufacturing automation created demand for higher-skilled positions in automation management and quality engineering.

Source: TechCrunch - Stacks Funding | Forbes - Finance Automation | Deloitte - Finance Transformation | McKinsey - AI in Finance Operations