The state of accounting in 2026

Accounting is the financial area with the greatest automation potential. According to a McKinsey study, 42% of accounting tasks are highly automatable, and another 19% are partially automatable. We are talking about 60% of an accounting department's work that can be assisted or executed by AI.

And yet, the reality is that most accounting departments still process invoices manually, perform bank reconciliations in Excel, and dedicate the first week of every month to a close that could be completed in 2 days.

The reason is not technological. The tools exist. The reason is that accounting is a conservative area by nature (and rightly so: errors have legal and tax consequences) and the adoption of new technologies is slow. This module shows you how to adopt accounting automation safely, with controls, and with measurable return on investment.

The accounting automation map

Accounting tasks fall into three categories based on their automation potential:

• Fully automatable (today): invoice data extraction (OCR + AI), bank transaction matching, depreciation calculations, recurring journal entry generation.
• Partially automatable: accounting categorization (AI proposes, human validates), complex item reconciliation, accruals, standard provisions.
• Require human judgment: complex accounting estimates, unusual transaction treatment, standards interpretation (IFRS, local GAAP), auditor relationships.

Intelligent invoice categorization

The process of receiving an invoice, reading it, categorizing it for accounting, and generating the journal entry is the biggest time consumer in an accounting department. A company processing 500 invoices per month dedicates between 80 and 120 hours monthly to this process alone. With AI, it can be reduced to 15-20 hours (the time for review and validation).

How AI categorization works

The automated process has 4 steps:

1. Extraction (OCR + AI): the invoice (PDF, image, or email) is processed with intelligent OCR that extracts: supplier, tax ID, date, taxable base, tax, description, invoice number.
2. Classification: AI categorizes the invoice according to your chart of accounts. For a telecommunications expense, it assigns the appropriate services account. For office supplies, the corresponding sub-account you have defined.
3. Matching: if the invoice corresponds to an existing purchase order, AI links it automatically (three-way matching: purchase order, delivery note, invoice).
4. Journal entry generation: with the extracted data and classification, AI generates the journal entry: expense account, input tax, supplier account.

Accuracy and learning

Categorization accuracy improves over time. A new AI system correctly classifies around 70-80% of categorizations. After 3 months of corrections by the accounting team, it rises to 90-95%. After one year, the best systems reach 97-99% for recurring suppliers.

The key is that AI learns from corrections. If you change the category of an invoice from one account to another, the system remembers it for the next invoice from that supplier.

Tax management

One of the complications specific to many markets is the variety of tax rates (standard, reduced, super-reduced, exempt, intra-community, import). AI can:

• Detect the correct tax rate based on the description and supplier.
• Identify intra-community transactions requiring reverse charge.
• Alert when the tax on an invoice does not match the expected rate for that supplier or service.
• Calculate tax proration when applicable.

Bank reconciliation with AI

Bank reconciliation is the process of matching bank statement movements with accounting records. For a company with 3 bank accounts and 200 monthly movements per account, manual reconciliation can take 2-3 days of work.

Automated matching

AI for bank reconciliation works on three layers:

• Exact matching: the amount and date match exactly between the bank movement and the accounting record. AI resolves this instantly.
• Fuzzy matching: the amount matches but the date differs by 1-3 days (typical for wire transfers), or the amount is slightly different (bank fees). AI proposes the match with a confidence level.
• Complex matching: one bank payment corresponds to multiple invoices (grouped payment) or one invoice is paid in multiple movements (partial payments). AI identifies possible combinations.

Unreconciled items

Items that AI cannot automatically reconcile are the most interesting: they usually indicate errors, duplicates, unrecorded payments, or unidentified income. AI can categorize them by type of issue and prioritize them so the accounting team investigates the most significant ones first.

Intercompany reconciliation

In corporate groups, intercompany reconciliation is a recurring pain point. Transactions between group companies must balance on both sides. AI can cross-reference statements from both entities and identify discrepancies: transactions recorded in one entity but not the other, date differences, exchange rate differences in international operations.

Month-end close automation

The monthly accounting close is the most structured and repetitive process in the finance department. And, paradoxically, the one that generates the most stress. The typical "hard close" in many companies occupies day 1 through day 10 of the following month, with the accounting team working against the clock.

Automatable close checklist

A typical monthly close has between 30 and 80 tasks. Those AI can automate or assist:

• Recurring journal entries: depreciation, insurance accruals, rent, subscriptions. AI generates them automatically every month.
• Payroll provision: if payroll is paid with a lag (March work is paid in April), AI calculates and books the provision.
• Debt reclassification: the portion of long-term debt maturing within the next 12 months is reclassified to current. Automatable.
• Revenue accruals: for annual service contracts, AI distributes revenue linearly and generates the monthly journal entry.
• Foreign exchange adjustments: for items denominated in foreign currency, AI applies the closing exchange rate and generates the adjustments.
• Balancing checks: verify that assets = liabilities + equity, that the cash flow statement reconciles with the cash variation, that intercompany balances match.

Reducing the close cycle

Companies that automate their close with AI report significant reductions:

• From 10 days to 5 days (first year of implementation).
• From 5 days to 3 days (with advanced automation and direct ERP connection).
• Some technology companies with simple operations achieve 1-2 day closes ("flash close").

The key is not just speed: it is quality. An automated close has fewer manual errors, automatic balancing checks, and complete traceability of every journal entry.

Intelligent expense management

Employee expense management (travel, meals, entertainment) is a process that creates friction across the entire organization. Employees hate filling out expense reports. Managers hate approving them. And the accounting department hates processing them.

Automated expense workflow

• Capture: the employee takes a photo of the receipt with their phone. AI extracts the merchant, amount, date, description, and tax rate.
• Categorization: AI classifies the expense according to the company's expense policy (meals, transportation, accommodation, entertainment) and the corresponding accounting account.
• Policy validation: automatic check against the expense policy. If the meal exceeds the limit, if the hotel exceeds the maximum rate per night, if the expense was incurred on a weekend without justification.
• Approval: automatic workflow to the appropriate manager, with all information already processed. The manager only approves or rejects.
• Booking: once approved, the journal entry is generated automatically and recorded in the ERP.

Expense anomaly detection

AI can identify anomalous patterns that manual controls do not detect:

• Receipts with suspiciously round amounts (exactly 50.00 EUR, multiple times per month).
• Duplicate expenses (same date, similar amount, same merchant).
• Expenses outside normal working hours.
• Spending patterns that change abruptly (an employee going from 200 EUR/month to 800 EUR/month without apparent reason).
• Receipts from establishments that do not match the employee's location.

AP/AR automation

Accounts Payable (AP) and Accounts Receivable (AR) are two complementary processes that AI transforms in different ways.

Accounts Payable (supplier payments)

• Receipt and registration: invoices arrive by email, supplier portal, or postal mail. AI centralizes all sources and registers them automatically.
• Three-way matching: AI cross-references the invoice with the purchase order and the delivery receipt. If all three match, the invoice is automatically approved for payment.
• Duplicate detection: same invoice sent twice (with a different PDF format, different email, same invoice number).
• Payment optimization: AI suggests which invoices to pay first to maximize early payment discounts and optimize cash flow.
• Payment forecasting: based on received invoices and typical terms, AI generates a payment forecast for the coming weeks.

Accounts Receivable (customer collections)

• Automatic issuance: invoice generation from ERP data (orders, delivery notes, service contracts).
• Collections tracking: AI monitors due dates and sends automatic escalating reminders (7 days before, due date, 7 days after, 15 days after).
• Default risk scoring: based on each customer's payment history, AI assigns a risk score and prioritizes collection actions.
• Collection prediction: for each outstanding invoice, AI estimates the probable collection date based on the customer's historical behavior.
• Provision suggestion: when a customer accumulates delays, AI suggests the bad debt provision according to accounting standards (local GAAP/IFRS).

OCR and AI: from paper to journal entry

OCR (Optical Character Recognition) with AI has evolved enormously. It is no longer just character recognition: it is document understanding.

From classic OCR to Document AI

Classic OCR reads text from an image. Document AI understands the structure of the document: it knows that a number in the bottom-right corner of an invoice is probably the total, that the number next to "Tax ID" is the fiscal identifier, that the central table contains the detail lines.

Relevant Document AI tools for accounting:

• Google Document AI: invoice, receipt, and financial document processing with high accuracy. Per-page pricing.
• AWS Textract: document data extraction with table and form understanding.
• Azure Form Recognizer: pre-trained models for invoices, receipts, and identity documents.
• Nanonets: specific to accounting automation. Pre-trained models for various invoice formats.

Processing invoices in multiple formats

Invoices have particularities that OCR must handle:

• Multiple tax rates on a single invoice (standard rate for services, reduced rate for certain products).
• Withholding tax on invoices from professionals (common in many jurisdictions).
• Equivalence surcharges in operations with certain retail merchants.
• Simplified invoices (receipts) with highly varied formats.
• Intra-community invoices with reverse charge notation.

AI tools for accounting

Integrated solutions

• Xero with AI: accounting software with integrated invoice OCR and automatic categorization.
• Dext (ex-Receipt Bank): expense capture and categorization with AI. Integration with major ERPs. From 24 EUR/month.
• Yokoy: expense management with AI for mid-size and large companies. Duplicate and policy violation detection.
• Tipalti: end-to-end AP automation with AI for matching and fraud detection.

For teams using classic ERPs

• SAP with AI: accounting automation modules within SAP S/4HANA.
• Sage with AI: automatic categorization of bank transactions in Sage.
• QuickBooks with AI: automated categorization and reconciliation features for SMBs.

Ejercicio practico

Puntos clave