Agentic accounting FabrCore Surface

We Built Our Own Accounting System. Then We Put Agents Inside It.

Impact started as internal accounting software. Now it is becoming a governed agentic accounting platform, with ReconAgent working directly inside reconciliation.

Eric Brasher | July 9, 2026 | 5 min read

Liverunning our books
1stproduction accounting agent
15+planned squad agents
.NETagent system foundation

At Vulcan365 and Vulcan365 AI, we spend a lot of time telling clients that practical AI should live inside the real workflows of the business. Not in demos. Not in slide decks. In the places where money moves, work gets approved, and decisions stack up.

So we built Impact, our own accounting system, in house.

We used a combination of OpenAI Codex and Anthropic Claude coding tools to move fast, reason through domain complexity, generate working code, review edge cases, and keep momentum. But the important part is not that AI helped us write software. The important part is that we used AI to build software we now run our company on.

Impact is live.

This is not a prototype or a sales demo. It is software we trust inside one of the most sensitive parts of our company: accounting.

Impact as the Source of Truth

At its core, Impact is a full accounting platform built around double-entry accounting, a general ledger, and operational subledgers.

General ledger

Chart of accounts, journal entries, posting rules, transfers, deposits, credits, and audit history.

Customer subledger

Customer invoices, payments, receivables, billing activity, and account-level payment history.

Vendor subledger

Vendor bills, payments, credits, cost tracking, and operational detail close to the workflow.

Reporting foundation

Balanced books with the G/L as truth and subledgers preserving the business context.

But we did not want a generic accounting system with generic reconciliation bolted on at the end. Our business has its own flow, and our accounting software needed to understand that flow.

Impact Reconciliation Is a Workflow

Reconciliation Flow
statement to balanced books
ImportRead statements and bring transaction detail into the accounting workflow.
IdentifyDetect customer payments, vendor payments, transfers, fees, and timing differences.
AllocateMap amounts to the correct G/L accounts and connect activity to customers and vendors.
BalanceFlag exceptions, review adjustments, and keep the reconciliation tied to the books.

Impact reconciliation is not just “match bank transaction to ledger entry.” It understands customer payments, vendor payments, deposits, transfers, fees, adjustments, and timing differences without forcing our team into someone else's accounting model.

Then We Added ReconAgent

On the reconciliation page, ReconAgent can read statements, interpret entries, suggest allocations, connect customer payments to open receivables, identify vendor payments, flag exceptions, and help balance the reconciliation.

Reads statement detail

The agent starts from the same financial artifact the accounting user is reviewing.

Suggests allocations

It proposes where amounts belong, while the accounting system keeps the final rules.

Connects entities

Customer and vendor context stays attached to the financial activity.

Flags exceptions

Ambiguity, timing differences, and odd entries become reviewable work.

It is not a chatbot floating off to the side. It is an agent embedded directly in the accounting workflow, with the context it needs and the tools to act where the work is happening.

A generic AI assistant can explain reconciliation.
ReconAgent helps do reconciliation.

Built with FabrCore and Surface

We built this with FabrCore, our open-source .NET framework for building AI agent systems.

Agent runtime

Lifecycle management, chat persistence, tool resolution, and multi-model orchestration.

Governance

Monitoring, access control, and verifiable execution for real application work.

Surface UI

Structured agent output and action routing inside the application page.

Surface lets us connect agents to real app pages, render structured agent output, route actions back into the system, and build experiences where the agent is part of the workflow instead of a separate tab.

From ReconAgent to AccountingAgent

ReconAgent is our first production accounting agent inside Impact. It will not be the last.

Now that the first agent is live, we are adding cost allocation agents, subscription and billing agents, and the one we are most excited about: AccountingAgent.

Reconciliation

Match, allocate, explain, and balance statement activity.

Payables

Review vendor bills, payments, cost detail, and exceptions.

Receivables

Connect payments, invoices, subscriptions, and customer history.

Close support

Help with reporting checks, review queues, and month-end readiness.

Audit prep

Preserve explanations, evidence, and traceable decisions.

Exception review

Surface unusual entries and route them to the right human decision.

That is the future we believe in: not one giant magic prompt, but a team of focused agents with roles, tools, memory, permissions, and accountability.

Why This Matters for Our Clients

This is also why Vulcan365 AI exists. We build practical AI systems that businesses own. We capture how the business actually runs, connect the systems already in place, and put AI against the bottlenecks that slow the company down.

Impact is proof of that philosophy.

We are not asking clients to believe in a theory. We are using the same approach internally. We are building with our own platform, shipping our own agents, and trusting them inside accounting.

Practical AI Belongs in Real Workflows

Impact started as an accounting system. Now it is becoming an agentic accounting platform.

And this is just the beginning.

About Vulcan365 AI: We build practical AI systems, integrations, and agent software for businesses that want ownership over their workflows and data. Based in Birmingham, Alabama.