ERP migration

Migrate from AI Works to Epicor Prophet 21

Field-level mapping, validation, and rollback between AI Works and Epicor Prophet 21. We move data and schema; workflows are rebuilt natively in Epicor Prophet 21.

AI Works logo

AI Works

Source

Epicor Prophet 21

Destination

Epicor Prophet 21 logo

Compatibility

83%

10 of 12

objects map 1:1 between AI Works and Epicor Prophet 21.

Complexity

BStandard

Timeline

6-9 weeks

Rollback included Accuracy guarantee Field-level validation

Overview

What this migration involves

Moving from AI Works to Epicor ERP is a manufacturing-data migration that touches the full production stack: part master, bill of materials, routing, customer and supplier records, open work orders, inventory positions, and financial history. Epicor Kinetic (the cloud-native manufacturing variant) stores these in a strongly-typed relational schema with required field constraints and site-specific cost layers that AI Works does not enforce in the same way. We profile the AI Works data model during discovery, build a custom field equivalency matrix, resolve multi-level BOM dependencies before import, and sequence the migration so that parent records (Sites, Part Cross-References, UOM classes) land before dependent child records (Parts, BOMs, Suppliers, Work Orders). Approval chains and workflow configurations do not migrate; we deliver a written inventory of every active approval chain and workflow rule for the customer's Epicor administrator to rebuild post-migration.

Field-level fidelity

Every standard and custom field arrives verified.

Schema-aware mapping

AI proposes the map; you confirm before any record moves.

Relationships preserved

Parent–child, lookups, and ownership stay linked.

Full activity history

Calls, emails, meetings — with original timestamps.

Attachments & notes

Documents, uploads, and inline notes move with the record.

Why teams make this switch

Two sides of the same decision

Leaving

AI Works logo

AI Works

What's pushing teams away

  • No public product documentation or schema reference — customers needing to migrate data cannot self-serve and must rely on direct vendor engagement.
  • Not a packaged ERP in the conventional sense — companies needing a system of record with published modules and roadmap typically move to NetSuite, Acumatica, or Microsoft Dynamics 365.
  • No published pricing, no integrations directory, and no public API surface available on the corporate website.
  • Small-vendor risk for finance-critical workflows; customers outgrowing the team typically migrate to better-supported platforms.
  • Limited third-party reviewer coverage — there is no G2/Capterra/Software Advice presence to validate fit or compare against alternatives.

Choosing

Epicor Prophet 21 logo

Epicor Prophet 21

What's pulling them in

  • Industry-specific design for wholesale distributors, not a general-purpose ERP repurposed for distribution — distributors choose P21 because it matches their replenishment, kitting, and counter-sale workflows out of the box.
  • Strong inventory control with automated replenishment, lot and serial tracking, and multi-warehouse management appeals to distributors with complex stock requirements and tight margin pressure.
  • Responsive customer support cited across G2 and Gartner reviews, with Epicor's 90% retention rate reflecting long-term customer satisfaction in a market where switching costs are high.
  • Cloud deployment on Microsoft Azure provides the flexibility to scale user counts and warehouse locations without on-premise infrastructure investment.
  • The Software Development Kit lets distributors personalize P21 to their specific business processes without modifying the application source code, preserving upgrade paths.

Object mapping

How AI Works objects map to Epicor Prophet 21

Each row shows how a AI Works object lands in Epicor Prophet 21, including any object-level transformations, lookup resolution, or schema-design dependencies.

Typical mapping — final map is confirmed during the sample migration step.

AI Works

Users

maps to

Epicor Prophet 21

User

1:1
Fully supported

AI Works User records map to Epicor Kinetic User. We resolve by email address as the dedupe key. Any AI Works user without a matching Epicor User is held in a reconciliation queue for the customer's Epicor administrator to provision before record import. Epicor User licensing (which tier of Kinetic) is determined by the customer's licensing decision and is outside migration scope.

AI Works

Accounts

maps to

Epicor Prophet 21

Customer and Supplier

1:many
Fully supported

AI Works Account records must be split into Epicor Customer (sold-to and ship-to) and Supplier records. We determine split direction using an AI Works account type or classification field if present. Customer records in Epicor require a Territory, Terms Code, and Credit Limit; we map these from AI Works fields where available and flag missing required fields for the customer's admin to configure before import.

AI Works

Part / Product

maps to

Epicor Prophet 21

Partmaster (Part table)

1:1
Fully supported

AI Works part records map to Epicor Partmaster. The part number maps to PartNum, description to Description, and standard cost to the Plant-specific standard cost layer. Epicor requires a UOM Class assignment; we resolve the AI Works unit-of-measure to an equivalent Epicor UOM Class or flag it for admin configuration. Part Type (stock, make, buy, configure) maps to TypeCode.

AI Works

Bill of Materials

maps to

Epicor Prophet 21

PartMtl (Part BOM table)

1:1
Fully supported

AI Works BOM records map to Epicor PartMtl. We resolve the MtlPartNum reference (the component part) by looking up the part number in the migrated Partmaster. Multi-level BOMs are resolved recursively: bottom-level components import first, then sub-assemblies, then top-level BOMs. Epicor's Revision and Approved flag on PartMtl are set based on AI Works BOM version status. Phantom assembly flags map from AI Works BOM type if available.

AI Works

Routing / Operations

maps to

Epicor Prophet 21

PartOpr (Part Routing table)

1:1
Fully supported

AI Works routing or operation records map to Epicor PartOpr. We map Work Center to Epicor Resource Group, operation sequence to OprSeq, and standard labor hours to EstLaborHours. The ResourceID reference is resolved against migrated Work Center records. If AI Works routing has no equivalent, we create a default Work Center and flag it for Epicor admin to reassign post-migration.

AI Works

Purchase Orders

maps to

Epicor Prophet 21

PoHeader and PoDetail

1:1
Fully supported

Open AI Works Purchase Orders map to Epicor PoHeader and PoDetail. The Supplier reference resolves to the migrated Supplier record. Line items map with POLine, PartNum, OrderQty, and UnitCost. Closed or completed POs are archived rather than migrated per Epicor performance guidance (decades of closed POs load the new system unnecessarily). We recommend archiving closed POs per the Archon Data Store approach.

AI Works

Sales Orders

maps to

Epicor Prophet 21

OrderHed and OrderDtl

1:1
Fully supported

Open AI Works Sales Orders map to Epicor OrderHed and OrderDtl. Customer reference resolves to the migrated Customer record. Line items map with OrderLine, PartNum, SellingQuantity, and UnitPrice. Linked Parts must exist in Partmaster before OrderDtl insert to satisfy the FK constraint. Quote history does not migrate as it is not required for day-one operations.

AI Works

Work Orders / Jobs

maps to

Epicor Prophet 21

JobHead and JobMtl

1:1
Fully supported

Open AI Works Work Orders map to Epicor JobHead and JobMtl. The JobNum is assigned from Epicor's number sequence (not preserved from AI Works). PartNum links to the migrated Partmaster. JobMtl lines resolve component PartNum references from the migrated BOM. If AI Works stores WIP labor and machine hours, these are mapped to JobOper if the operation sequence exists or flagged as notes for the customer's production manager to enter manually.

AI Works

Inventory / Stock

maps to

Epicor Prophet 21

PartWhse (Part-Plant-Warehouse)

1:1
Fully supported

AI Works inventory positions map to Epicor PartWhse by PartNum, Site (Plant), and Warehouse. OnHandQty, DemandQty, and SafetyStockQty migrate per site. Inventory value maps to the Part-plant cost layer as a cost layer adjustment rather than a transaction. We flag any inventory records where the Site does not exist in Epicor for admin to create before import.

AI Works

GL Accounts / Chart of Accounts

maps to

Epicor Prophet 21

GlbAcct (Global Account)

1:1
Fully supported

AI Works GL accounts map to Epicor Global Account. Account code and description migrate. Account Type (Asset, Liability, Expense, Revenue) maps to the Epicor GLAcctType or Segment values. If AI Works uses a segmented chart (e.g., Division-Dept-Account), we preserve the full segment string and configure Epicor's corresponding account segment structure before import.

AI Works

Invoices

maps to

Epicor Prophet 21

InvcHead and InvcDtl

1:1
Fully supported

AI Works AP and AR invoices map to Epicor InvcHead and InvcDtl. We migrate open (unpaid) invoices only; historical paid invoices are archived. Invoice header maps to Customer or Supplier reference, invoice date, and total. Line items map PartNum or GlbAcct, quantity, and amount. Epicor Invoice matched to a PO carries a reference to the PoNum; we resolve this by looking up the migrated PO number.

AI Works

Custom Fields

maps to

Epicor Prophet 21

UD columns + UD01-UD30

lossy
Fully supported

AI Works custom fields on any entity map to Epicor User-Defined columns (UD columns) on the equivalent table or to the UD01-UD30 cross-reference tables. We create the UD column schema in Epicor Application Studio before any data import. Field type mapping: AI Works text maps to character, number to decimal or integer, date to date, checkbox to logical. Epicor UD30 is used for cross-entity key-value pairs if AI Works stores freeform custom data on records.

Gotchas + challenges

What specifically takes care here

Platform-specific issues from each side, plus the pair-specific challenges that don't show up on either platform's page on its own.

AI Works logo

AI Works gotchas

High

Vendor is not positioned as a conventional packaged ERP

High

No publicly documented API or schema

Epicor Prophet 21 logo

Epicor Prophet 21 gotchas

High

Third-party bolt-on integrations complicate migration scope

High

Dirty data without standardized processes compounds migration risk

Medium

SDK customizations and BPMs may not survive platform upgrades

Medium

Report-based export only for non-technical users

Low

Per-user pricing model requires accurate user count before migration planning

Pair-specific challenges

  • BOM and routing dependencies must resolve before parts import

    Epicor enforces foreign-key constraints between Partmaster, PartMtl (BOM), and PartOpr (routing). The component part must exist in Partmaster before a BOM line referencing it can insert. We resolve multi-level BOMs bottom-up (leaf components first, then sub-assemblies, then top-level), and we validate the complete BOM tree in a staging environment before production import. Migrations that attempt to load BOMs in source-system order routinely fail at the FK constraint because parent parts are not yet created, causing a cascade of silent skips or hard errors that require a re-run.

  • Approval chains and workflow rules do not migrate as code

    AI Works approval chains and workflow configurations are platform-specific rules that do not have a data-level representation in Epicor ERP. Epicor uses BPM (Business Process Management) and Kinetic REST API customizations for approval routing, which are code artifacts rather than records. We deliver a written inventory of every active AI Works approval chain and workflow with its trigger, approver logic, and conditions, so the customer's Epicor administrator can rebuild them as Epicor BPMs or Kinetic customization post-migration. This handoff is the end of migration scope for workflow rebuild.

  • Epicor requires site-level schema before inventory and cost records insert

    Epicor Kinetic stores inventory positions, cost layers, and GL references at the Site (Plant) level. Every PartWhse record requires a valid Plant and Warehouse combination that must exist in Epicor before any inventory data can insert. We audit the AI Works multi-site configuration during discovery, create the corresponding Plant and Warehouse records in Epicor as a prerequisite step, and map AI Works site identifiers to the Epicor Plant codes. Sites without a mapping are flagged for the customer to resolve before inventory migration begins.

  • Historical closed transactions bloat the Epicor environment

    Epicor migration specialists including Archon Data Store and Baker Tilly consistently advise against migrating all historical closed transactions. AI Works systems with years of closed POs, closed work orders, and historical invoices add unnecessary load and cost to the Epicor environment without supporting day-one operations. We migrate open and current-year financial records and archive historical closed transactions to an external data store. The customer retains audit access to historical records without the performance and licensing cost of keeping them live in Epicor.

  • Epicor UD field schema must be created before custom data import

    AI Works custom fields on parts, customers, and suppliers do not have automatic Epicor equivalents. We create the destination UD column schema (UD01-UD30 or table-specific UD columns) in Epicor Application Studio as a pre-import step. Epicor enforces that UD columns exist on the table before any data can insert into them, and the column type must match the incoming data format. A mismatch between AI Works text-stored-as-number and Epicor decimal columns will silently truncate or reject records unless validated during staging.

Migration approach

Six steps for a successful AI Works to Epicor Prophet 21 data migration

  1. Discovery and data profiling

    We audit the AI Works environment: record counts for parts, BOMs, routings, customers, suppliers, open POs, open work orders, inventory positions, GL accounts, and custom fields. We profile data quality (null rates, duplicate part numbers, missing GL codes, inconsistent UOMs) and produce a Data Quality Report. We identify the Epicor Kinetic product line (Kinetic for manufacturing, Prophet 21 for distribution, or Eclipse for electrical/HVAC) and the specific Epicor edition and modules being provisioned, which determines which tables and fields are available at licensing tier.

  2. Schema design and Epicor pre-configuration

    We design the destination Epicor schema: Plant and Warehouse creation, UOM class mapping, GL account segment structure, and PartMtl/Purchasing term defaults. We create UD column schema in Epicor Application Studio for every mapped AI Works custom field. We configure the Epicor number sequences that will replace AI Works native IDs (PartNum, JobNum, PONum, OrderNum) and document the ID mapping for reconciliation. Schema is deployed to a Epicor Sandbox or test company for validation before any production migration begins.

  3. BOM and routing tree resolution

    We resolve the full BOM and routing dependency tree before any import begins. Bottom-level component parts import first, then sub-assemblies, then top-level parts with BOMs, then routing operations. We validate that every PartMtl MtlPartNum and every PartOpr ResourceID has a matching migrated record. Multi-level BOMs are validated for circular references (which Epicor rejects on insert) and for orphan lines where the component part has no master record. Any orphan lines are flagged for the customer's engineering team to resolve.

  4. Sandbox migration and reconciliation

    We run a full migration into an Epicor test company using production-like data volume. The customer's Epicor administrator reconciles record counts (Parts in, BOMs in, Customers in, Suppliers in, POs in, Work Orders in, Inventory in), spot-checks 25-50 random records against the AI Works source, and signs off the schema, mapping, and UD field results before production migration begins. Any BOM orphans, missing site mappings, or UD type mismatches are corrected in the staging pass, not in production.

  5. Production migration in dependency order

    We run production migration in dependency order: Plants and Warehouses (admin-created, validated), UOM Classes, GL Accounts, Partmaster (with Part-Plant cost layers), PartMtl BOMs, PartOpr routings, Customers and Suppliers (with UD fields), open Purchase Orders, open Sales Orders, open Work Orders, PartWhse inventory positions, and open Invoices. Each phase emits a row-count reconciliation report. We migrate open and current-year financial records; historical closed transactions are archived rather than imported.

  6. Cutover, final validation, and approval chain handoff

    We freeze AI Works writes during cutover, run a final delta migration of any records modified during the migration window, then enable Epicor ERP as the system of record. We deliver the Approval Chain and Workflow Inventory document to the customer's Epicor administrator. We support a one-week hypercare window where we resolve any reconciliation issues raised by the customer's team. We do not rebuild AI Works approval chains as Epicor BPMs or workflow rules inside the migration scope; that is a separate engagement or an internal Epicor administrator task.

Platform deep dives

Context on both ends of the pair

AI Works logo

AI Works

Source

Strengths

  • Flexible custom-build engagement model suits organizations needing bespoke AI overlays.
  • Direct, small-team vendor relationship.
  • AI talent resourcing can complement internal teams.
  • No prescriptive product lock-in for those starting from a custom data model.

Weaknesses

  • No public ERP product documentation, schema, or pricing.
  • Absent from major reviewer aggregators (G2, Capterra, Software Advice).
  • Migration scoping requires the vendor's direct cooperation; self-service is not viable.
  • Roadmap and product versioning are not publicly visible.
  • Small-vendor delivery risk for finance-critical systems.
Epicor Prophet 21 logo

Epicor Prophet 21

Destination

Strengths

  • Purpose-built for wholesale distribution with industry-specific replenishment, kitting, and counter-sale workflows out of the box.
  • Multi-warehouse management with bin locations, cross-docking, and real-time inventory visibility across all warehouse locations.
  • Automated replenishment engine with demand-based and min-max planning reduces stockouts and overstock carrying costs.
  • AI-infused reporting via Epicor Prism provides Gen AI-driven insights into ERP data without requiring a BI team.
  • Strong customer retention at 90% and a 50-year track record in the distribution vertical provides long-term vendor stability.

Weaknesses

  • High total cost of ownership — per-user pricing of $150-200/month plus $10K-$500K implementation creates significant budget commitment for small and mid-market distributors.
  • Customization via SDK requires technical expertise and introduces upgrade risk when custom code conflicts with new P21 releases.
  • Report generation performance is a known pain point — multiple users report system freezes during large or complex report exports.
  • Third-party bolt-on reliance for functionality that competitors include natively increases integration complexity and total solution cost.
  • Limited public API documentation — developers building custom integrations report difficulty finding P21 API authentication methods and endpoint specifications.

Complexity grading

How hard is this migration?

Standard ERP migration. 2 of 8 objects need a mapping; the rest are 1:1.

B

Overall complexity

Standard migration

Derived from compatibility, mapping clarity, API constraints, and data volume across AI Works and Epicor Prophet 21.

  • Object compatibility

    B

    2 of 8 objects need a mapping; the rest are 1:1.

  • Field mapping clarity

    C

    Field mapping is derived from defaults — final spec confirmed during the sample migration.

  • Timeline complexity

    B

    8-object category — typical timelines run 2–7 days end-to-end.

  • API constraints

    B

    AI Works: Not publicly documented — typical SaaS limits assumed and confirmed during scoping.

  • Data volume sensitivity

    B

    AI Works doesn't expose a bulk API — REST + parallelization used for high-volume runs.

Estimator

Estimate your AI Works to Epicor Prophet 21 migration cost

Rule-based pricing — no per-record fees, no manual quotes. Migrations over 2M records are scoped individually.

Step 1

What are you migrating?

Pick a category, then your source and destination platforms.

Category

FAQ

Frequently asked questions about AI Works to Epicor Prophet 21 data migrations

Answers to the questions buyers ask most during AI Works to Epicor Prophet 21 migration scoping. Not seeing yours? Book a call.

Can't find your answer?

Walk through your AI Works to Epicor Prophet 21 migration with a real engineer — 30 minutes, free, written quote within 24 hours.

Book a free 30 minute consultation

Most migrations land between six and nine weeks for accounts under 10,000 parts, single-site, and no historical transactional data. Migrations with multi-level BOMs (over 5 BOM levels), multi-site inventory (over 3 plants), open work order volumes (over 2,000 jobs), or extensive UD field customizations move to twelve to twenty weeks because of BOM recursion resolution, Plant and Warehouse pre-configuration, and GL account segment mapping. Epicor Kinetic cloud onboarding timeline (separate from FlitStack AI scope) typically adds 2-4 weeks before migration begins.

Adjacent paths

Related migrations to explore

Ready when you are

Move from AI Works.
Land in Epicor Prophet 21, intact.

Tell us record counts and timeline. We'll come back with a written quote inside 1 business day — no commitment, no sales pitch.

Accuracy guarantee Rollback included Quote in 1 business day