Account Reconciliation Process: A Step-by-Step Guide

The account reconciliation process is the structured sequence of steps that verifies whether the balance recorded in your general ledger is accurate, complete, and supported by evidence. It is the primary control that prevents misstatements from compounding across periods, and it is the step in the financial close where most teams spend the majority of their time.

This guide covers the account reconciliation process step by step, from initial data gathering through final sign-off. It is designed for teams that want to formalize their process, train new preparers, or evaluate where automation can reduce effort without sacrificing control quality. For a structured implementation approach, see our account reconciliation implementation guide. If you are evaluating software to automate this process, our account reconciliation software guide covers what to look for.

What is the account reconciliation process?

At its core, account reconciliation answers one question: does the balance in the general ledger match reality?

"Reality" takes different forms depending on the account. For a bank account, reality is the bank statement. For an intercompany receivable, reality is the counterparty's payable balance. For an accrued liability, reality is the supporting calculation and documentation. For a fixed asset, reality is the sub-ledger detail.

The reconciliation process is the series of steps that connects the GL balance to that reality, investigates any differences, and documents the conclusion. When done well, it produces an audit-ready work paper that proves the balance is right. When done poorly, or not done at all,. It leaves misstatements undiscovered until an auditor or a downstream process catches them.

Every organization reconciles. The difference between organizations that close in 5 days and those that close in 15 is how efficiently they move through the process for hundreds of accounts simultaneously.

The 7-step reconciliation process

Regardless of account type, reconciliation method, or level of automation, every reconciliation follows this sequence. Understanding each step is essential for identifying where your team spends time and where process improvement delivers the most value.

Step 1: Gather source data.

Pull the two data sets that will be compared. One is always the general ledger balance or transaction detail for the account being reconciled. The other depends on the account type:

• Bank accounts: bank statement or bank feed
• Intercompany accounts: counterparty balance from the other entity's books
• Fixed assets: sub-ledger detail from the fixed asset register
• Accrued liabilities: supporting calculations, invoices, or schedules
• Revenue accounts: billing system detail, contract schedules

The challenge at scale is not pulling one account's data. It is pulling data for 200 or 500 accounts from multiple source systems, often in different formats. Automated data integration eliminates the manual extraction step by importing source data directly into the reconciliation platform on schedule.

Step 2: Match transactions.

For accounts reconciled at the transaction level, bank accounts, intercompany, accounts payable,, compare individual transactions between the GL and the source data to identify matches.

Matching logic varies in complexity:

• One-to-one matching: A single GL transaction matches a single source transaction by amount, date, and reference. The simplest case.
• One-to-many matching: A single GL transaction (a batch deposit) matches multiple source transactions (individual customer payments). Requires summing the source transactions to verify they equal the GL amount.
• Many-to-many matching: Multiple GL transactions match multiple source transactions where amounts are split or combined differently between systems.

Transaction matching engines handle all three match types automatically, using exact matching, tolerance matching for rounding differences, and fuzzy matching for transactions with similar but not identical descriptions or amounts.

Step 3: Identify discrepancies.

After matching, two categories remain: unmatched items and balance differences.

Unmatched items are transactions that appear in one data set but not the other. A check issued and recorded in the GL but not yet cleared at the bank. An intercompany invoice recorded by one entity but not yet by the counterparty. A cash receipt deposited but not yet posted to the GL.

Balance differences are the net result of all unmatched items. If your GL shows $1,042,500 and the bank shows $1,035,200, the $7,300 difference must be fully explained by specific reconciling items.

The output of this step is a list of every discrepancy, classified by type: timing difference, posting error, missing transaction, or unknown.

Step 4: Investigate unmatched items.

This is where reconciliation consumes the most time. For every unmatched item and every unexplained balance difference, the preparer must determine the cause.

Investigation typically involves:

• Checking posting dates to identify timing differences (transactions recorded in different periods)
• Reviewing supporting documents (invoices, contracts, bank notices) to verify transaction accuracy
• Contacting other departments or entities for information on missing transactions
• Comparing to prior period reconciling items to identify recurring patterns
• Pulling additional data from the ERP or sub-ledger systems for context

For a single complex account, investigation can take 30 minutes to 2 hours. Across 300 accounts, the investigation step alone can consume 200 to 500 hours per close cycle.

AI investigation fundamentally changes this step. Instead of the preparer manually querying systems and assembling evidence, an AI investigation agent performs the same queries, cross-references the data, identifies the root cause, and presents a structured finding. The preparer reviews the finding and approves or overrides it, a 5-minute review replacing a 45-minute investigation.

Step 5: Record adjusting entries.

Based on investigation findings, some discrepancies require corrective action. A payment recorded in the wrong amount needs a correcting entry. An expense invoiced but never recorded needs an accrual. A duplicate posting needs a reversal.

Each adjusting entry must be:

• Documented with the specific reconciling item that triggered it
• Prepared with the correct account mapping, entity coding, and period assignment
• Approved through the journal entry workflow before posting to the GL

Not every discrepancy requires an adjustment. Timing differences, a check mailed but not yet cleared,, resolve in the next period without any GL entry. The reconciliation documents the timing difference and carries it as an expected reconciling item.

Step 6: Document and certify.

The preparer creates a work paper that documents the reconciliation: the GL balance, the source balance, the matched items, the unmatched items, the investigation findings, the adjusting entries, and the conclusion. The conclusion is typically a statement that the balance is supported, all reconciling items are explained, and the account is ready for review.

Standardized documentation matters. When 20 preparers use 20 different formats, the reviewer spends time understanding the format before they can assess the substance. Consistent templates, or better, automated work papers generated by the reconciliation platform, ensure that every reconciliation package has the same structure, the same level of detail, and the same supporting evidence.

The preparer certifies the reconciliation: asserting that the work is complete, the balance is supported, and the findings are accurate.

Step 7: Review and approve.

A reviewer, typically a manager, senior accountant, or controller,, evaluates the completed reconciliation. The review process includes:

• Verifying that the GL balance ties to the source data after accounting for reconciling items
• Assessing whether investigation findings are reasonable and adequately supported
• Confirming that adjusting entries are appropriate and properly documented
• Evaluating whether any reconciling items represent control deficiencies or require escalation
• Approving or returning the reconciliation for additional work

The preparer-reviewer workflow is a fundamental internal control. The person who prepares the reconciliation should not be the person who approves it. Account reconciliation platforms enforce this separation of duties systematically.

Types of account reconciliation

Different accounts require different reconciliation approaches. Understanding the types helps allocate the right method, and the right level of effort,, to each account.

Bank reconciliation. The most familiar type. The GL cash balance is compared to the bank statement balance. Reconciling items include outstanding checks, deposits in transit, bank fees not yet recorded, and direct debits not yet posted. High transaction volume makes this the reconciliation type most improved by automation.

Balance sheet reconciliation. Broadly, any non-bank balance sheet account reconciled against supporting detail: prepaid expenses against amortization schedules, accrued liabilities against supporting calculations, deferred revenue against contract schedules, debt against loan amortization tables. The reconciliation method depends on the account: balance comparison for accounts with clear independent sources, account analysis for estimated balances.

Intercompany reconciliation. Matching and reconciling balances between entities within the same corporate group. Both sides of every intercompany relationship must agree before consolidation can proceed. The intercompany reconciliation module automates matching across entities and flags disputes for resolution.

Vendor and customer reconciliation. Comparing your AP or AR balances against statements received from vendors or customers. Vendor reconciliation catches duplicate payments, missed credits, and pricing errors. Customer reconciliation catches misapplied cash, disputed invoices, and revenue recognition issues.

Credit card reconciliation. Matching corporate credit card transactions to the card statement and verifying that all charges are properly coded, approved, and recorded in the GL. High transaction volume and varied coding make this reconciliation type particularly tedious when done manually.

Manual vs automated reconciliation

The difference between manual and automated reconciliation is not just speed. It is what your team spends their time on.

Manual reconciliation: The preparer performs every step. They export data, format spreadsheets, match transactions by eye, investigate variances by querying multiple systems, write up findings in a Word document or spreadsheet template, and submit for review. A moderately complex account takes 2 to 4 hours. The preparer's expertise is consumed by mechanical work, data gathering and transaction matching,. That does not require judgment.

Automated reconciliation: The platform performs steps 1 through 3 automatically. Data imports from source systems on schedule. Transaction matching runs using multi-layered logic (exact, tolerance, fuzzy, pattern). Discrepancies are identified and classified. For accounts where all items match within tolerance, the reconciliation auto-certifies with a generated work paper. For accounts with unmatched items, AI investigates the discrepancies and presents findings for review.

The numbers tell the story. A team reconciling 300 accounts manually spends 600 to 1,200 hours per month. The same team with auto-reconciliation spends 40 to 80 hours, a 90 to 95 percent reduction. The hours saved are not idle time. They are reallocated to the 5 to 15 percent of accounts that genuinely require expert judgment.

How AI changes reconciliation

The first generation of reconciliation software digitized the manual process. Templates replaced spreadsheets. Workflow replaced email. Dashboards replaced status meetings. This was valuable. It solved consistency and tracking problems. But it did not solve the core problem: investigation.

AI solves the investigation problem. Three capabilities make the difference:

AI investigation agents. When the matching engine identifies an unmatched item or an unexplained variance, the investigation agent goes to work. It queries the ERP for posting details. It checks the bank for clearing information. It compares to prior period patterns. It pulls supporting documents. It assembles a structured finding: "This $15,000 variance is a timing difference. Payment #4521 was issued March 29 and cleared April 2. No adjustment required. Supporting evidence: [attached bank clearing confirmation]." The accountant reads the finding, agrees, and approves. Five minutes instead of forty-five.

Confidence scoring. Not every automated reconciliation is equally reliable. Confidence scoring assigns a quantitative score to each reconciliation based on match completeness, variance magnitude, historical stability, and data quality. Your team sets the threshold: auto-certify above 95, review between 80 and 95, manual preparation below 80. The scoring is transparent. Every factor in the score is visible and auditable.

Continuous learning. When your team overrides an AI finding, the system incorporates the feedback. When a reviewer changes a confidence threshold, the system adjusts. Over time, the auto-reconciliation rate increases and the override rate decreases. The AI calibrates to your data, your accounts, and your team's professional judgment.

The practical result is a reconciliation process where your team's expertise is deployed on the accounts that need it. Not consumed by the 85 percent that reconcile cleanly every period.

Frequently asked questions

How many accounts should be reconciled each period?

Best practice is to reconcile every material balance sheet account every period. What constitutes "material" depends on your organization's risk tolerance and your auditor's expectations. A risk-based approach assigns reconciliation frequency based on balance size, transaction volume, historical error rates, and inherent risk. High-risk accounts reconcile monthly. Lower-risk accounts may reconcile quarterly. The key is that the decision is documented and defensible. Not that every account receives the same treatment.

What is the difference between a reconciliation and a review?

A reconciliation compares two data sets, the GL and an independent source,, and explains the differences. It is a detailed, evidence-based analysis. A review is a higher-level evaluation of the reconciliation's quality: are the findings reasonable, is the documentation adequate, and is the conclusion supported? In the preparer-reviewer model, the preparer reconciles and the reviewer reviews. Separating these roles is a fundamental internal control.

How do you prioritize which accounts to reconcile first?

Start with accounts that are prerequisites for downstream close activities. Bank accounts and intercompany balances must be reconciled before consolidation can proceed. High-materiality and high-risk accounts should be reconciled early in the close window so that issues have time to be investigated and resolved. Low-risk, low-materiality accounts can be reconciled later or, with AI automation, continuously throughout the period.

What is a reconciling item?

A reconciling item is any difference between the GL balance and the independent source balance that has been identified and explained. Common reconciling items include timing differences (transactions recorded in different periods by each source), known adjustments (bank fees not yet recorded), and pending corrections (posting errors identified for adjustment). Every reconciling item should have an owner, an expected resolution date, and documentation of the cause.

How long should a single reconciliation take?

Manual reconciliation of a moderately complex account (200 to 500 transactions, multiple sub-ledger sources, some investigation required), takes 2 to 4 hours. With AI-powered automation, the same reconciliation takes 5 to 15 minutes of human time: reviewing the AI-generated work paper, evaluating flagged items, and approving or overriding findings. Simple accounts with clean matching (under 50 transactions, no variances) auto-certify without any human time at all.

Related reading

• Account Reconciliation Software Guide
• Best Account Reconciliation Software 2026
• Account Reconciliation - automated matching and AI investigation
• Cortex Investigation Agent - how AI investigates variances autonomously