Skip to content
Proton

Is your ERP data ready for AI? A distributor’s checklist

Start with the work you want AI to do. Then check whether the source data, the integration and the resulting outputs support that work.

Shelby KinneyShelby KinneyProton.ai

Published 10 min read

TL;DR

  • Readiness depends on the workflow. Pick the first AI use case before you judge the data.
  • A field existing in your ERP doesn’t make it ready. Check the source, the integration, and the output separately.
  • The ten checks below give you questions to ask about your ERP.
  • Record each result with an owner. A total score can’t tell you a workflow is safe to launch.
Sample extractcustomers.csv
✓BillToID✓AssignedSalesRepID✓ItemCategory✓InvoiceDate

4/4

fields reviewed

7 records to reviewNext: test outputs

An account list might group every delivery location separately even though a rep manages several of them as one customer. A price lookup might return a value without applying the rules your team expects when an order is saved.

Finding data mismatches is part of assessing your ERP readiness.

Here’s the short answer: your ERP data is ready for an AI workflow when the integration reaches the data that workflow needs, that data matches what your team intended, and the outputs hold up when the people who use them review them.

Does data quality affect AI results in distribution?

In one industry survey, distributors who rated their data highly were far more confident of getting a return from AI.

81%
of surveyed distributors who called their data strong were confident AI would pay off
18%
of those who called it poor or inconsistent said the same

That’s what respondents reported, not a measured return. Source: Distribution Strategy Group, 2026.

Talk through your ERP 

Which AI use case should you start with?

Pick one workflow you want to infuse with AI before you look at the data. Write down who will use it and what a good result looks like.

That tells you which data to check first.

What to check depends on the workflow:

  • Recommendations: purchase history, how accounts relate to each other, product information, and whether each account is allowed to buy what’s suggested.
  • Quote and order entry: how products are identified, units of measure, quantities, pricing, and what happens when the order is saved.
  • Product enrichment: which attributes you need, where the values come from, and who approves them.

Treat this as a starting list. Your implementation team should confirm the exact requirements for your setup. Keep in mind that a field existing in your ERP doesn’t mean its values are complete, current, or mapped correctly.

Which use cases should you assess first?

Tick the data you can review today. A use case becomes a candidate for validation when its starting data is in place.

Data you can review

  • Recommendations2 of 4
    • Invoice-line history
    • Bill-to / ship-to relationships
    • Product categories and vendors
    • Which products each account can buy
  • Quote and order entry0 of 2
    • Item master with units of measure
    • Pricing and order rules
  • Product enrichmentValidate first
    • Item IDs and descriptions

1 of 3 use cases to validate first

This identifies data to review, not a launch decision. Confirm requirements and test outputs for your implementation before launch.

The ERP data readiness checklist

Ten checks to prepare a conversation with your ERP admin, business owner and implementation team. Each one says what to look at and how to test it.

  1. Check 1

    Can the integration reach your ERP data?

    Ask your ERP admin which access methods your ERP license includes: APIs, a database replica or scheduled files. Confirm who can grant access and which permissions the integration needs. Some ERP subscriptions include a read-only database copy without licensed API access, so check what yours includes.

    Test it: Pull real records from the intended environment before you plan the rollout around it. A successful login or schema check isn’t enough on its own.

    AccessNot answered
    API · replica · files

    What your ERP license includes vs. the access the integration needs

    Your result

  2. Check 2

    Do your bill-to and ship-to records match how you sell?

    Compare bill-to and ship-to relationships with how your team sells and reports. Review representative customers with the reps who own them. Decide which locations should remain distinct and which views should roll them up.

    Test it: Before you use an account list to prioritize outreach, check whether its grouping matches how the rep sells.

    Customer relationshipsNot answered
    BillToID ↔ ShipToID

    The account list vs. how your reps sell

    Your result

  3. Check 3

    Does every account route to the right person?

    Check which record level each assignment lives on, and whether the integration reads it there. For example, a rep column on bill-to records might be empty while the assignments exist only on the ship-to records beneath them. Bill-to views would then show none of that rep’s accounts. Also handle house and service records deliberately, since they don’t always need a person.

    Test it: Confirm that each in-scope account routes to the intended rep, team or documented exception. Verify login identities for people who need access.

    Work routingNot answered
    SalesRepID

    Where assignments live vs. where the integration reads them

    Your result

  4. Check 4

    Are your contacts attached to the right accounts?

    Review contact records with the reps who use them. A contact can legitimately belong to several locations, for example where purchasing is centralized. A shared phone number or address does not prove two records are the same person. If you run more than one ERP instance, make sure contact IDs can’t overlap.

    Test it: Confirm account and location associations, stable identifiers and available contact details for representative accounts. Investigate suspected duplicates before merging them.

    ContactsNot answered
    ContactID

    Each contact’s accounts vs. what reps expect

    Your result

  5. Check 5

    Do sales figures match the report finance trusts?

    This matters for workflows using sales or margin. Use the same branches, dates and transaction scope in both systems. Agree how tax, freight, credits, cancellations and replacement invoices are handled.

    Test it: Reconcile a closed period against the report finance trusts. If totals differ, trace the discrepancy through the source, the export and the application before changing any calculation.

    Financial outputsNot answered
    InvoiceDate · InvoiceNo

    A closed period vs. the report finance trusts

    Your result

  6. Check 6

    Are your products classified the way the use case needs?

    Missing or inconsistent product classifications can make category-level analysis and search less reliable. Review categories, vendors and identifiers for the products relevant to your first use case. Confirm what category codes mean and whether product IDs are unique across companies or ERP instances.

    Test it: Validate both the category structure and its coverage of the products that matter. Separate values missing in the source from incomplete exports or syncs. If you use AI product data enrichment, review enriched attributes before relying on them.

    ClassificationsNot answered
    ItemCategory · VendorID

    Coverage of the products your use case depends on

    Your result

  7. Check 7

    Where do cost and margin live, and what’s in them?

    Confirm where cost and margin live in your ERP and which fields the integration receives. Validate item-level unit cost and transaction-level margin separately. Agree which fees or adjustments belong in each calculation.

    Test it: Compare sample records with the source ERP before relying on margin reports.

    Cost and marginNot answered
    UnitCost · LineMargin

    Sample records vs. the source ERP

    Your result

  8. Check 8

    Does the feed show what’s happening now?

    Agree whether status lives on the order line or the order header, how partial shipments and cancellations are represented, and how often updates arrive. A field that was populated once is not the same as a live feed. Confirm how you will detect failed updates and recover missing records.

    Test it: Follow a known open order and a known backorder through the ERP, the export and the application.

    Live statusNot answered
    LineStatus · QtyOnHand

    A known open order across ERP, export and app

    Your result

  9. Check 9

    Do pricing and order rules hold end to end?

    Contract pricing, customer-specific pricing and unit-of-measure logic can depend on several inputs at once. Check the rules your reps rely on and how the workflow handles a missing or unexpected price. Ask your ERP admin whether rules that run when an order is saved also run for orders created through an API or import.

    Test it: Test representative quotes and orders from lookup through creation or saving. Include the customer, item, quantity, unit of measure and applicable contract.

    Pricing and order rulesNot answered
    ContractPrice · UOM

    A test quote vs. the price your ERP applies on save

    Your result

  10. Check 10

    Is there enough history, and will your IDs hold up?

    Agree on the historical coverage your use case needs. A routine update window can be shorter than the history your ERP keeps, so confirm how older records can be re-exported or backfilled. For example, a quotes file and a customer file might write the same account ID with different capitalization, such as acct-1001 and ACCT-1001. If the integration matches IDs exactly, those quotes won’t appear for anyone.

    Test it: Confirm the regular update window and the backfill route. If a migration is planned, document old-to-new identifiers and changes to fields, definitions and routing, then retest affected outputs after cutover.

    History and IDsNot answered
    acct-1001 ≠ ACCT-1001

    The history your use case needs vs. what the feed sends

    Your result

How should you read your results?

As you work through the checklist, mark each check one of three ways:

  • Checked: you tested it and the result held up.
  • Gap: you tested it and found something to fix. Give it an owner.
  • Unknown: you haven’t tested it yet, or nobody can answer it yet.

Then read the pattern of results instead of adding them up. Here’s what the common patterns mean.

A gap in ERP access or pricing

Resolve it first. A gap in check 1 or check 9 can hold up any workflow that depends on it, even when every other check is marked Checked.

A gap with no owner

Name who will fix it, and agree whether the fix belongs to your team or the integration provider.

Mostly unknowns

That’s a normal place to start. Test the checks your first use case depends on before the rest. Your unknowns are the agenda for your next conversation with your ERP admin.

Gaps outside your first use case

Keep validating your first use case while those gaps are worked in parallel. For example, a gap in cost and margin (check 7) doesn’t stop you from testing product enrichment.

Every check marked Checked

Every check passed, but that still isn’t a launch decision. Test real outputs for your first workflow with the people who will use them, then plan the rollout.

Your results · your next step

Work through the checks above

Mark each one Unknown, Gap or Checked as you go. Your results build here.

10 not answered0 unknown0 gap0 checked

Gaps to resolve

  • No gaps recorded

Still to test

  • 1. AccessNot answered yet
  • 2. Customer relationshipsNot answered yet
  • 3. Work routingNot answered yet
  • 4. ContactsNot answered yet
  • 5. Financial outputsNot answered yet
  • 6. ClassificationsNot answered yet
  • 7. Cost and marginNot answered yet
  • 8. Live statusNot answered yet
  • 9. Pricing and order rulesNot answered yet
  • 10. History and IDsNot answered yet

Checked

  • Nothing checked yet

What this means for each use case

  • RecommendationsChecks 1, 2, 6, 104 of 4 to test
  • Quote and order entryChecks 1, 6, 8, 94 of 4 to test
  • Product enrichmentChecks 1, 62 of 2 to test

Each use case is matched to the checks most related to it. Confirm the exact requirements with your implementation team.

What should you ask your ERP admin?

Every ERP and every deployment stores the business differently, and many distributors run more than one after acquisitions or regional expansion. These are questions to ask about your own version, access rights, identifiers and business rules. They are not statements about what any connector supports.

  • Epicor Prophet 21 (P21): Which API views or database access does our subscription include? Do business rules that run on order save also run for API or imported orders?
  • Epicor Eclipse: Does our export include invoice dates and line-level status? How are cancelled and replacement invoices represented?
  • NetSuite: Which ID does each export use, internal or external? How are contacts linked across customers, ship-to addresses or sub-customers, and subsidiaries?
  • Infor CSD and SX.e: Which customer, ship-to and contact files does our export include, and how often do they refresh? Which item types, such as non-stock items, are included in our exports?
  • DDI Inform: Which records can we read through the API, and which need a report export?
  • Microsoft Dynamics 365: How many instances do we run, and which ones hold sales, inventory and pricing?
  • Multiple ERPs: Which system is the source of truth for each customer, item and transaction? Do customer, item and contact IDs overlap or differ in format between systems and files, and how will they be kept apart? Do terms like “active customer”, “sales” and “margin” mean the same thing in each system?

What if you’re planning an ERP migration?

Assess what you can validate before cutover and what will change afterward. An ID crosswalk helps, but also review field meanings, transaction history, permissions and account routing. For financial outputs, reconcile a full closed period after the change.

Who should own ERP data readiness?

Name a business owner who can define a useful result and an ERP admin who can confirm access and source behavior. Bring finance into reconciliation, reps into account and recommendation review, and product specialists into attribute validation.

Ask reps to review specific recommendations, including whether the account is allowed to buy the recommended products. Agree which fixes belong to your team and which belong to the integration provider. Our guide to software implementation in distribution covers rollout ownership in more depth.

What can you do this week?

  1. Choose one workflow.
  2. Name its business owner and ERP admin.
  3. Confirm access with the questions in check 1.
  4. Collect representative records for that workflow and test the relevant checks.
  5. Record the gaps and agree on the next validation step.

Want to talk through your ERP and first AI use case? Use this checklist to prepare questions about access, mappings and validation, then book a call with our team.

Sources

Shelby Kinney

Written by

Shelby Kinney

Shelby Kinney is the senior manager of editorial, brand, and content at Proton.ai.

Keep up with distribution in five minutes a month

Latest news, from tariff changes to acquisitions, and what they mean for you, plus unique insights from leading executives.

Once a month. Unsubscribe anytime.

Curious what’s trapped in your systems?

A 30-minute call with someone who speaks fluent distribution. No scripts, no pressure.