[object Object]

A small distributor handles 200 inbound purchase orders a week. Three people spend half their day rekeying supplier emails into Business Central. The Wave 1 agentic ERP capability turns most of that into a review queue, and the people get their afternoons back. The promise is real; the rollout discipline decides whether you keep it.

The Direction

BC’s 2026 Wave 1 shifts toward agentic ERP — AI agents automating sales order processing, purchase requisitions, approval flows. For SMB and mid-market, ERP-level automation that previously required ERP engineers becomes citizen-buildable. The agents are pre-built for common scenarios and configurable for the variations every business has.

This matters most for organizations without a full ERP team. The agents replace the consulting engagements that used to be the only path to automation.

Sales Order Automation

Inbound orders parsed, validated against stock and pricing, processed with fulfillment kickoff. Human review for exceptions. Order-to-cash velocity improves meaningfully. The parser handles structured EDI, semi-structured emails, and PDF order forms with comparable accuracy.

Inbound: supplier email with PO attached
Step 1: extract line items, quantities, pricing
Step 2: validate against item master and price agreements
Step 3: check inventory availability
Step 4: if all pass, create sales order; else queue for review

The validation step is where most automation projects fail in older tools. The agentic version handles unit conversions and product alias matches that hard-coded parsers miss.

Purchase Scenarios

Requisition to approval to PO to receipt. Routine items flow automatically within policy. Purchasing team handles exceptions and strategic buys. Define the policy explicitly; the agent will not invent a budget threshold for you.

purchase_policy:
  auto_approve_below: 500
  auto_approve_strategic_vendors: true
  require_approval_above: 5000
  require_dual_approval_above: 25000
  budget_check: required

The policy is the contract between the agent and the finance controls. Review it quarterly; budget thresholds drift faster than people remember.

Developer Productivity

Extensibility investments make extending BC faster. Agents assist developers with AL code. Custom functionality that used to require specialist developers becomes accessible to in-house IT. The AL agent suggests refactors, generates test scenarios, and explains existing extensions in plain language.

// Agent-suggested refactor
procedure ValidateOrder(SalesHeader: Record "Sales Header"): Boolean
begin
    if SalesHeader."Sell-to Customer No." = '' then
        exit(false);
    if SalesHeader."Document Date" = 0D then
        exit(false);
    exit(CheckLines(SalesHeader));
end;

The agent does not replace the AL developer; it shortens the time between “I have a request” and “I have a deployable change.”

Adoption Pattern That Works

Start with one scenario, measure for a quarter, then expand. The most common starting scenario is sales order automation because the value is immediate and the failure mode is visible (an order in the review queue is obvious). Purchase requisitions come second because the policy work is heavier.

Quarter 1: sales order automation, 60% auto-process target
Quarter 2: tune to 80%, add purchase requisition pilot
Quarter 3: expand purchase to 50% auto-process
Quarter 4: add approval orchestration

Audit Considerations

Every agent action should be logged with the agent identifier, the input, the decision, and the resulting record. Auditors want to know who created the PO, and “Copilot” is not an acceptable answer without traceability. Build the audit fields into the schema before going live.

What to do this week

Pick one scenario, define the policy in writing, and stand up a sandbox with the agentic order processing enabled. Run a week of real inbound orders through the sandbox and triage the review queue with the team that will operate the production version.

[object Object]
Share