Every month, someone in your firm sits down with custodian statements and compares them line by line to your internal records. They check balances, spot missing transactions, flag fee discrepancies, and chase down explanations for anything that doesn’t match. It takes hours. It’s tedious. And it’s expensive when you calculate the cost of a paraplanner or operations manager doing this work instead of higher-value tasks.

For a typical advisory firm managing 200 to 500 client accounts, monthly reconciliation can consume 15 to 30 hours of staff time. That’s not just the checking, it’s the follow-up emails to custodians, the internal notes, the corrections in your portfolio management system, and the audit trail you need to keep for compliance. If you’re paying a senior paraplanner $80K to $100K a year, you’re spending $600 to $1,200 every month on this one process. Over a year, that’s $7K to $14K in direct labor, and it doesn’t count the opportunity cost of what that person could be doing instead.

The real problem isn’t the cost alone. It’s the risk. Manual reconciliation is prone to error. A missed transaction or a misread balance can cascade into client reporting mistakes, compliance gaps, or worse, a regulatory issue that costs tens of thousands to fix. One firm in our network described a situation where a fee discrepancy went unnoticed for three months because the reconciliation process was rushed. The error affected 40 clients and took weeks to unwind.

AI can fix this. Not by replacing your team, but by handling the repetitive, high-risk parts of reconciliation automatically. An agent can pull custodian data, compare it to your internal records, flag every discrepancy, and route issues to the right person with context. It runs every month without supervision. It doesn’t miss lines. And it creates a clean audit trail that compliance can review in minutes.

This article walks through how custodian statement reconciliation works today, where the time goes, and what an AI agent doing this work looks like in practice. If you’re spending more than a day a month on reconciliation, this is worth your attention.

The Manual Reconciliation Process

Most advisory firms follow a version of the same workflow. At the end of each month, custodian statements arrive by email or download. Someone on the operations team exports them to PDF or CSV, then opens the firm’s portfolio management system to pull the corresponding internal records. The comparison begins.

They check account balances first. Does the custodian’s closing balance match the internal system? If not, they drill into the transaction history. They look for missing trades, duplicate entries, corporate actions that weren’t recorded, or fees that weren’t captured. They cross-reference dates, amounts, and security identifiers. When they find a mismatch, they note it, investigate the cause, and decide whether to correct the internal system, contact the custodian, or flag it for the adviser.

This process repeats for every account. For firms with multiple custodians, it means navigating different statement formats, different data structures, and different quirks in how each custodian reports information. One custodian might show fees as separate line items. Another might net them into the transaction. A third might report them in a summary section that doesn’t map cleanly to your system.

The time adds up fast. A straightforward account might take five minutes to reconcile. An account with active trading, multiple asset classes, or a recent rollover can take 20 minutes or more. Multiply that by 200 accounts and you’re looking at 16 to 60 hours of work, depending on complexity.

Then there’s the follow-up. If the operations team finds discrepancies, they email the custodian, wait for a response, and track the issue to resolution. If the problem is internal, they update the system and document the change. Every correction needs a note for the audit file. Every unresolved issue needs a status update. The workflow stretches across days or weeks.

Most firms batch this work into a reconciliation sprint at month-end. It’s disruptive. It pulls people off other projects. And it creates a bottleneck, because nothing downstream can happen until reconciliation is done. Client reports wait. Quarterly reviews wait. Compliance checks wait.

Where AI Fits

An AI agent doesn’t replace the judgment your team brings to reconciliation. It replaces the manual checking. It pulls custodian data automatically, compares it to your internal system, and surfaces every discrepancy in a structured format. It runs on a schedule you set, usually the first business day after month-end. It doesn’t need supervision. And it doesn’t miss anything.

Here’s what that looks like in practice. The agent connects to your custodian’s data feed or ingests statements from a shared folder. It reads the data, normalizes it into a consistent format, and maps it to your internal records. It compares balances, transaction histories, and fee entries. When it finds a mismatch, it logs the details, categorizes the issue, and flags it for review.

The output is a reconciliation report. It shows every account that reconciled cleanly and every account with a discrepancy. For each issue, the report includes the custodian’s data, your internal data, the difference, and a suggested next step. The agent can even draft the email to the custodian or the internal correction note, based on the type of mismatch it found.

Your team reviews the report, investigates the flagged issues, and takes action. The agent handles the rest. It updates the audit log, tracks the status of each issue, and generates a summary for compliance. The entire process that used to take 15 to 30 hours now takes two to four hours of human time, mostly spent on judgment calls and follow-up.

We’ve built agents like this for advisory firms using Omni Ops, the AI layer that connects to your existing systems and runs workflows end to end. One firm with 350 accounts cut their monthly reconciliation time from 24 hours to three hours. Another firm with multiple custodians eliminated the need for a dedicated reconciliation role entirely, redeploying that person to client onboarding and advice preparation.

The accuracy improvement is just as important as the time savings. Manual reconciliation depends on focus and consistency. An agent doesn’t get tired. It doesn’t skip lines. It doesn’t rationalize a small discrepancy as immaterial and move on. It flags everything, which means your team catches issues that would have been missed.

What an AI Agent Does

Let’s walk through a specific example. You have 250 client accounts spread across three custodians. At the end of January, the agent runs automatically. It pulls statements from each custodian, either through an API connection or by reading files from a shared folder. It extracts the data, transaction by transaction, and loads it into a comparison table.

The agent then queries your portfolio management system for the corresponding internal records. It matches accounts by custodian account number, compares closing balances, and checks every transaction for date, amount, security, and type. It identifies three categories of discrepancy: missing transactions, balance mismatches, and fee differences.

For missing transactions, the agent flags the custodian entry and checks whether it’s a known issue, like a late-reported trade or a corporate action. If the transaction type is routine, the agent drafts a correction entry for your system. If it’s unusual, it escalates to your operations manager with a summary of the issue.

For balance mismatches, the agent calculates the difference and traces it back to the underlying transactions. It identifies whether the issue is a single missing entry, a cumulative error, or a data format problem. It logs the details and suggests whether to contact the custodian or adjust the internal record.

For fee differences, the agent compares the custodian’s reported fees to your internal fee schedule. It flags accounts where the charged fee doesn’t match the expected amount. It notes the variance and routes it to the adviser for review, because fee discrepancies often require client communication.

The agent compiles all of this into a reconciliation report. The report shows 247 accounts that reconciled cleanly, two accounts with missing transactions, and one account with a fee discrepancy. Your operations manager reviews the three flagged accounts, investigates the issues, and resolves them in under an hour. The agent updates the audit log with the resolution notes and closes the reconciliation for January.

This is the kind of workflow we build with the AI audit for financial advisory firms. We map your current reconciliation process, identify the decision points that need human judgment, and automate everything else. The agent runs monthly, but it can also run on demand if you need to reconcile a single account or investigate a client question.

The Broader Impact

Automating reconciliation doesn’t just save time. It changes how your operations team works. Instead of spending days on manual checking, they focus on exceptions, client service, and process improvement. They become problem solvers, not data validators.

It also tightens your compliance posture. Automated reconciliation creates a complete audit trail. Every comparison, every discrepancy, and every resolution is logged with a timestamp and a reason. When your compliance officer or an external auditor asks to see your reconciliation records, you hand them a structured report, not a folder of spreadsheets and email threads.

For firms that are growing, automation removes a scaling constraint. Adding 50 new accounts doesn’t mean hiring another operations person or stretching your existing team thinner. The agent handles the additional volume without incremental cost. You scale revenue without scaling headcount in proportion.

One firm we worked with was preparing for a merger. They needed to reconcile 600 accounts across five custodians in a compressed timeline to satisfy due diligence requirements. Manual reconciliation would have taken weeks and required temporary staff. The agent completed the work in two days, with the internal team spending about 12 hours reviewing flagged issues. The clean audit trail was a key factor in closing the deal on schedule.

Reconciliation also connects to other workflows in your firm. Once the agent confirms that custodian data matches your internal records, it can trigger downstream processes like client reporting, performance analysis, or billing. You move from a monthly batch process to a continuous workflow where data is always current and always verified.

We see this pattern across the advisory firms we work with. Automating one high-volume, low-judgment task creates capacity for the next improvement. Firms that start with reconciliation often move on to automating meeting prep, advice document drafting, or client onboarding. Each agent compounds the efficiency gain from the previous one.

Building the Agent

The technical work to build a reconciliation agent is straightforward if you have the right foundation. The agent needs three capabilities: data ingestion, comparison logic, and workflow orchestration.

Data ingestion means connecting to your custodians and your portfolio management system. Most custodians offer API access or structured file exports. The agent reads the data, parses it into a standard format, and loads it into a comparison table. If your custodian doesn’t offer an API, the agent can read PDFs or CSVs from a shared folder.

Comparison logic is the core of the agent. It matches accounts, compares balances and transactions, and flags discrepancies based on rules you define. The rules can be simple, like flagging any balance difference over $10, or complex, like checking whether a missing transaction is a known corporate action type that can be auto-corrected.

Workflow orchestration ties everything together. The agent runs on a schedule, generates the reconciliation report, routes flagged issues to the right person, and updates the audit log. It can also send notifications, draft follow-up emails, and track issue resolution.

We build these agents on Omni Ops, which handles the infrastructure, the integrations, and the compliance logging. You don’t need to hire a development team or manage servers. The agent runs in your environment, connects to your systems, and operates under your control.

The build process starts with the Omni Audit. We spend 60 minutes mapping your current reconciliation workflow, identifying the decision points, and defining the rules the agent will follow. We then build a prototype agent that runs on a sample of your data. You see it work, you test it, and you decide whether to deploy it across all accounts.

Most firms go live within two to four weeks. The agent runs in parallel with your manual process for the first month to build confidence. After that, it takes over. Your team shifts to reviewing the agent’s output instead of doing the comparison themselves.

What This Means for Your Firm

If you’re spending 15 to 30 hours a month on reconciliation, you’re spending $7K to $14K a year in direct labor. That’s the floor. It doesn’t count the opportunity cost of what your operations team could be doing instead, the risk of manual errors, or the time lost to follow-up and corrections.

An AI agent brings that cost down to $2K to $4K a year, mostly in the form of review time. The time savings free up capacity for higher-value work. The accuracy improvement reduces compliance risk. And the audit trail makes regulatory reviews faster and cleaner.

For firms with multiple custodians or complex account structures, the savings are higher. One firm with 500 accounts and four custodians was spending 40 hours a month on reconciliation. The agent cut that to six hours. The annual savings exceeded $20K in labor alone, and the firm redeployed the operations manager to lead a client onboarding improvement project that shortened time-to-first-review by two weeks.

The broader pattern is this: advisory firms leak $70K to $200K a year on manual processes that AI can automate. Reconciliation is one piece. Meeting prep is another. Advice document drafting is a third. Each process you automate compounds the efficiency gain and creates capacity for growth.

We’ve written more about how AI fits into advisory operations on the EDNA blog and in our insights library. If you want to explore the full range of agents we build for advisory firms, start with See Omni for financial advisory firms.

If you’re deciding where to start with agents, start here. The free Working With Claude field guide walks through the ecosystem, Claude Code, and a real rollout plan. Get your copy.

Reconciliation is one of those tasks that everyone knows is inefficient but no one has time to fix. AI makes it fixable. The question is whether you want to keep spending 15 to 30 hours a month on manual checking, or whether you’d rather spend two hours reviewing an agent’s output and redeploy the rest of that capacity somewhere it drives revenue.