Order IntakeERP integrationsPricingFree pilotPartnersBook a demo

Change language

English (US)English (Europe)EspañolCatalàPortuguêsFrançaisItalianoMagyarRomânăDeutschDutchPolskiΕλληνικά
Illustrative example

Customer request

2 cases of tomatoes and 6 bottles of olive oil

Structured result to verify

  • Customer: verify the account
  • Product: match the exact catalogue SKU
  • Quantity: 2 cases and 6 bottles
  • Pack size: confirm from the catalogue

Unresolved details need clarification

Examples illustrate the process. They are not customer results or promises of unconfirmed functions.

How AI Matches Customer Descriptions to Product SKUs

AI product matching connects the words a customer uses with the product codes in your catalogue. It can help interpret abbreviations, informal names and customer references, but a reliable order also needs the correct variant, packaging and unit. Uncertain matches need clarification or review.

Test matching against your catalogueSee all questions

Why extracting a product name is not enough

A phrase such as ‘tomato sauce’ may correspond to several brands, sizes or pack formats. Reading the phrase correctly does not identify the SKU by itself. The matching process needs catalogue information and any relevant customer references configured for the business.

Preserve product identity and units

Illustrative example: ‘3 cases of 500 ml water’ identifies a size and a quantity of cases, but it does not establish the number of bottles in each case. The appropriate product and pack definition must come from the catalogue. A conversion should follow an agreed rule, not an assumption based on another customer or product.

Use reference data carefully

A current product catalogue is the starting point. Customer-specific product references and historical documents may help configure matching, but old descriptions, discontinued products and previous prices can be misleading. Confirm which sources are used and which system remains authoritative.

Measure matching separately from complete order accuracy

A pilot should show where the customer was identified correctly, where product and quantity interpretation succeeded, and where a person had to intervene. Orders containing several lines may expose different problems from single-item examples. Include similar products and incomplete descriptions so the evaluation reflects the real workload.

A matching check you can repeat

Illustrative example. The catalogue contains a 500 ml bottle and a case of 12 bottles. A request for three cases must retain that unit. It does not mean three individual bottles.

Compare the selected SKU, packaging and quantity with the original request. A similar product name does not establish the correct variant. Where the catalogue lacks a pack definition, record the missing information instead of inventing a conversion.

Test ambiguous descriptions alongside explicit product codes. Record successful matches, corrections and unresolved lines separately. Historical invoices can help configuration, but the current catalogue remains the authority.

Related questions

How does AI match customer descriptions to the correct product SKU?

Product matching compares the customer's wording with your catalogue and the references configured for your business. Customers may use abbreviations, brand names, informal descriptions or their own product codes. evolio.ai identifies products from incoming orders, but reliable matching depends on the quality of the catalogue and onboarding. Similar products and uncertain matches need explicit handling rather than a silent guess.

Direct link to this answer

How are cases, packs, pieces and kilograms handled in an order?

The order must preserve both the requested quantity and its unit. “Three cases” and “three pieces” are different requests, even when the product name is the same. Configure the relevant units, pack sizes and conversion rules from your catalogue, then test representative orders. Where a unit is missing or a conversion is uncertain, the workflow should request clarification or review.

Direct link to this answer

What happens when several products match the same description?

When a description could refer to several catalogue items, the process needs more information or a human decision. Brand, size, packaging and customer-specific references can help distinguish products. AI should not silently select a plausible item when the order remains ambiguous. Include these cases in your pilot and agree how unresolved matches are shown to the team.

Direct link to this answer

Can customers order without knowing our product codes?

Customers do not always need to provide an internal SKU if their description contains enough information to identify the product. AI order intake can interpret familiar names and natural language against your catalogue. Clear customer references reduce ambiguity, while incomplete descriptions may still need clarification. The goal is to accommodate real ordering habits while preserving the correct product and unit.

Direct link to this answer

Does AI order intake automatically determine prices and stock availability?

Interpreting an order is different from confirming price or stock. Customer-specific pricing, inventory availability, discounts and commercial conditions should come from the authoritative business system or an agreed validation step. Do not assume that extracting a product request confirms its availability or final price. Ask which checks are included in your particular integration and which remain in the ERP.

Direct link to this answer

Can historical invoices help configure product and customer matching?

Historical documents can provide useful examples of customers, product descriptions and previous order patterns. They can support configuration, but they may contain discontinued products, old prices or outdated references. The current catalogue and business rules remain the authority. Agree how historical data is used during onboarding, and do not assume that history alone can resolve every ambiguous future request.

Direct link to this answer

How should we measure AI order entry accuracy?

Define the metric before testing. Correctly identifying a customer, matching each product and quantity, and completing an entire order are different measures. Evaluate both the interpretation result and the data delivered through your ERP workflow. Report exceptions and required corrections as well as successful cases. A headline accuracy percentage is meaningful only when its definition, sample and review policy are clear.

Direct link to this answer
See all questions

Practical guides for your order workflow

AI Order Entry for Distributors: From Email and PDF to ERP

Turn customer emails and PDF orders into structured order data. Learn how product matching, exception handling and ERP delivery fit together.

Automate WhatsApp Orders Without Changing How Customers Buy

Process customer WhatsApp orders into structured data while keeping familiar ordering habits. Test messages, images and voice notes with your team.

The Real Cost of Manual Order Entry

Estimate the time and cost of reading, interpreting and entering orders. Compare recoverable effort with automation costs using your own inputs.

AI Order Entry vs OCR, EDI and B2B Portals

Compare ways to automate distributor orders by customer behavior, input structure, product matching and operational requirements.

How to Run an AI Order Entry Pilot on Your Own Orders

Test AI order intake with your customers' messages and documents. Define the sample, accuracy measures, approval rules and ERP validation.

Test the workflow on your real orders

Bring representative messages and documents. Check product matching, exceptions and the agreed ERP output before expanding.

Test matching against your catalogue