Agentic supply chain planning for Business Central

Pre-launch - join the waitlist for founding pricing.

Last reviewed: 2026-05-30

Agentic supply chain planning is the next step after dashboards and forecasting tools. Instead of only showing a planner what might happen, an agent forecasts demand, decides what to do, queues the action for review, and writes approved changes back to the ERP. For Microsoft Dynamics 365 Business Central distributors, that means recommendations for demand forecasts, safety stock, reorder points, purchase orders, transfer orders, and exception handling inside the planning workflow your team already uses.

Isovel is built for the Business Central mid-market version of that category: one-click connect, first actionable recommendations in under ten minutes, 30-day read-only shadow mode, approval-gated write-back, idempotency, rollback, and audit logs on every action.

49%

of expedited orders are caused by inaccurate demand forecasts.

APQC via Supply & Demand Chain Executive [1]

55%

of SMBs report holding at least 20% excess stock.

Netstock 2025 State of Supply Chain Planning [2]

<10 min

from one-click connect to your first actionable Business Central recommendation.

Isovel product commitment [3]

Phase 1   Pre-launch waitlist. $999/month or $10,000/year. One tier. 30-day read-only shadow trial before write-back is enabled.

What is agentic supply chain planning?

Agentic supply chain planning uses AI agents to sense planning problems, decide on recommended actions, and execute approved changes across forecasting, inventory, replenishment, and exception workflows.

The important word is agentic. A dashboard shows information. A predictive model estimates what might happen. A workflow automation routes a task someone already created. An agent closes the loop:

  1. It observes the ERP state.
  2. It forecasts what is likely to happen.
  3. It recommends what to do.
  4. It asks for human approval where risk requires it.
  5. It writes the approved action back.
  6. It logs the reason and learns from overrides.

For a Business Central distributor, that loop should touch the real planning surfaces: Demand Forecast Entries, Item and SKU planning parameters, Purchase Orders, Transfer Orders, and approval workflows. If a product stops at a dashboard export, it is not a supply chain planning agent. It is an analytics tool.

Why this matters for Business Central distributors

Business Central has the operational primitives distributors need: items, locations, vendors, sales orders, purchase orders, transfer orders, SKUs, planning worksheets, requisition worksheets, and planning parameters. The challenge is that the planning decisions often live outside the ERP.

Planners export data, rebuild forecasts, calculate safety stock, compare warehouses, decide what to buy, and then copy the decision back into Business Central. That is why Excel persists in supply-chain work. [4] It is flexible enough to patch the gaps, but fragile enough to become a second system of record.

Agentic planning moves the decision loop back toward Business Central. The agent does the repetitive math and exception sorting. The planner keeps judgment, override authority, supplier context, and the trust boundary.

Agentic vs. predictive vs. Copilot vs. automation

The language around AI planning is noisy. These four categories are not the same.

CategoryWhat it doesWhere it stops
Predictive analyticsForecasts demand, risk, or stock position.The planner still decides and acts.
Workflow automationRoutes a document, approval, or task.It automates a process someone already defined.
CopilotHelps a user understand data, draft work, or complete a task faster.The user initiates and steers the work.
Agentic planningForecasts, decides, queues, writes back, audits, and learns from overrides.It stops at the trust boundary the business sets.

Microsoft uses the same broad distinction inside Business Central: Copilot works with the user, while agents automate specific processes with work that stays transparent and reviewable. [5] Isovel follows that pattern for supply chain planning. Copilot can help a user understand Business Central data. Isovel decides and acts on replenishment recommendations, with approval gates and rollback.

Where Microsoft’s Business Central agents are today

Microsoft is actively bringing AI and agents into Business Central, and that is good for the category. Microsoft documents Business Central AI capabilities across Copilot, finance, sales, setup, and other AI or machine-learning features. [5] Sales Order Agent is a clear example of an integrated Business Central agent: it can handle sales quote and sales order requests from customer email, involve users for review, and keep a task timeline. [6]

That validates the agent model inside Business Central. It also clarifies the gap Isovel is built for. Microsoft’s published Business Central AI surface includes inventory forecasting, but the Copilot FAQ states that sales and inventory forecasting rely on specific Azure AI models and are not Copilot features. [7] Forecasting exists; closed-loop supply chain planning is the missing layer for mid-market distributors.

Isovel is Microsoft-complementary by design. It does not replace Business Central. It reads Business Central, plans on top of Business Central, and writes approved recommendations back to Business Central.

How Isovel’s planning agent works

The Isovel loop is built around six steps.

1. Connect. Sign in with Office 365 and authorize the Business Central tenant. Isovel reads item, location, sales, inventory, vendor, purchase, transfer, and lead-time history through documented BC surfaces.

2. Forecast. The agent generates demand forecasts at SKU-location level with confidence bands and rationale. Fast movers, slow movers, intermittent items, new items, and noisy items are handled differently.

3. Optimize. The agent recalculates planning parameters: safety stock, reorder points, reorder quantities, maximum inventory, and Item or SKU policy choices. Business Central documents these parameters as the inputs planning uses to calculate order suggestions. [8]

4. Decide. The agent decides whether the right action is buy, transfer, wait, reduce, or review. If one warehouse has excess stock while another is short, a transfer can beat a purchase order.

5. Queue. Every recommendation lands in an approval queue with the rationale visible. Low-confidence or high-impact decisions are routed to the planner.

6. Write back. Approved actions write back to Business Central. Every write-back is approval-gated by default, idempotent on re-run, reversible inside a 24-hour rollback window, and audit-logged.

The agent’s job is not to remove planners. It is to remove the work that keeps planners from planning.

What Isovel agents handle

Isovel starts with the supply chain decisions that hurt mid-market distributors most.

AI demand forecasting

The demand agent forecasts at SKU-location level, returns a confidence band, and explains why the forecast moved. It degrades gracefully on sparse or volatile data instead of returning a hard no-forecast wall. Read the deeper page at AI demand forecasting software for Business Central.

AI inventory optimization

The inventory agent updates dynamic safety stock, reorder points, reorder quantities, maximum inventory, ABC and forecastability classes, and slow-moving or dead-stock exceptions. It writes approved planning-parameter changes back to the Item or SKU card. Read AI inventory optimization for Business Central.

Multi-location rebalancing

The location agent detects when stock sits in one location while another location is short. It recommends transfer orders before purchase orders when moving stock is the better decision. Read multi-location inventory rebalancing.

Purchase order automation

The replenishment agent decides what to buy, from whom, in what quantity, with MOQ, order multiple, case pack, pallet, and lead-time constraints applied. It queues draft purchase recommendations for approval and writes approved actions back. Read purchase order automation for Business Central.

Exception triage

The exception agent keeps humans focused on the few decisions where judgment matters: high-value stockouts, uncertain forecasts, unusual supplier behavior, low-confidence recommendations, dead-stock risk, or planner overrides that keep recurring.

Trust controls are the product

An agent that writes to an ERP of record needs a stronger control model than a dashboard.

Isovel’s trust model has five parts:

  1. Shadow mode first. The 30-day trial is read-only. Isovel pulls Business Central data and surfaces recommendations, but it does not write during the trial.
  2. Approval-gated write-back. Every write-back requires explicit approval at the start.
  3. Idempotency. Re-running the same recommendation does not duplicate records in Business Central.
  4. Rollback. Approved writes can be reversed inside a 24-hour window.
  5. Audit logs. Every recommendation records the data, rationale, user, approval state, and write-back result.

Autonomy is graduated. Trust primitives are permanent. A planner can allow a narrow class of low-risk actions to auto-execute after the agent earns trust, then restore manual approval at any time.

Why dashboards are not enough

Many planning tools show stockout risk, excess inventory, and forecast charts. That helps, but it leaves the planner with the hardest part: deciding and acting.

A dashboard says: “This SKU may stock out.”

An agent says: “Raise the reorder point from 42 to 57, buy 120 units from Vendor A, or transfer 38 units from Dallas to Houston. Here is the forecast, the lead-time assumption, the confidence range, the MOQ constraint, and the reason this recommendation changed.”

The second answer is operational. It can be approved, written back, audited, rolled back, and improved from the planner’s override.

When agentic supply chain planning is a fit

Isovel is a good fit when:

  1. You run Microsoft Dynamics 365 Business Central.
  2. You manage roughly 100 to 5,000 SKUs.
  3. You have multiple locations, warehouses, branches, or service territories.
  4. Your planning process depends on Excel exports.
  5. Your team is carrying excess stock and still expediting orders.
  6. You want BC-native write-back, not another standalone planning dashboard.
  7. Your IT team needs approval gates, rollback, audit logs, and a read-only trial.

It is not a fit if you need a full enterprise S&OP suite, custom network optimization, production scheduling at enterprise depth, or a consulting-led implementation with heavy business-process redesign.

What to ask an agentic supply chain vendor

Ask these before trusting any system with ERP write-back.

  1. What exact Business Central records does the agent read?
  2. What exact Business Central records can it write?
  3. Can the agent run in read-only shadow mode?
  4. Are write-backs approval-gated by default?
  5. Are writes idempotent on re-run?
  6. Can writes be rolled back?
  7. Does every recommendation include a rationale?
  8. Can planners override recommendations and preserve the reason?
  9. Which workflows are agentic, and which are only dashboards?
  10. How long until the first useful recommendation appears?

Weak answers stay abstract. Strong answers name the table, document, field, approval state, rollback path, and failure mode.

Try agentic planning in shadow mode first.

Connect Business Central, compare Isovel’s recommendations against your current planning process for 30 days, and keep write-back disabled until your team is ready.

Get early access

FAQ

What is agentic supply chain planning? Agentic supply chain planning uses AI agents to forecast, decide, queue, and execute approved planning actions across demand forecasting, inventory optimization, replenishment, transfers, and exceptions.

How is agentic supply chain planning different from supply chain analytics? Analytics shows what is happening or what might happen. Agentic planning recommends what to do, asks for approval where needed, writes approved actions back to the ERP, and logs the result.

Does agentic supply chain planning replace planners? No. It shifts planners away from spreadsheet upkeep and repetitive exception sorting. Planners still own judgment, supplier context, overrides, approval policies, and risk tradeoffs.

Can agentic AI work with Business Central? Yes, when it respects Business Central’s existing item, SKU, purchase, transfer, approval, and planning workflows. Isovel reads BC data and writes approved planning actions back to BC.

How does Isovel work with Microsoft Copilot? Microsoft Copilot helps users work inside Business Central. Isovel is a supply chain planning agent that decides and writes approved replenishment actions back to Business Central. The two are complementary.

Is agentic supply chain planning fully autonomous? Not at the start, and not without controls. Isovel starts in read-only shadow mode, then approval-gated write-back, then graduated autonomy only for action classes the planner explicitly trusts.