Why your OTIF doesn't move

Many VPs of Supply Chain may know their supplier OTIF, but almost none have visibility into the root causes driving their cost of expedites, excess inventory, or production disruptions. OTIF is an outcome measure. It records that the plan missed, without recording which conversation the miss started in.

The plan got smarter

Manufacturers have spent years upgrading their S&OP processes, forecasting, and MRP. The plan keeps getting smarter, but their supply chain resilience barely improves.

Most of the challenges that break the plan start outside the system, with partners.

• A customer cannot accurately forecast demand for its products.
• A supplier fails to flag a shortage forming on their end.
• A forwarder misses notifying about a prolonged customs hold.

Let's take suppliers for instance. When the communication and coordination are done on email, your system of record will always lag behind reality. POs get issued from the ERP, but confirmations arrive as replies that never make it back to the system.

Planners end up estimating based on stale information, and the delay surfaces only after it is too late. No plan can absorb such unexpected delays. That is why execution starts to matter more than planning.

Resilient supply chain execution resolves risks of delays as they surface. It reduces the need for expedites.

Why the MRP looks healthy

Most manufacturers I talk to name exception management as their main execution pain. Besides the major global disruptions, it's the steady stream of small supplier deviations every week that breaks the production schedule.

Here is an example. When a supplier moves a delivery date or cuts the quantity, the change either never gets communicated or arrives too late. It lands in an inbox and stays there until a planner reads it and manually updates the ERP.

Planners continue working from the original commitment in the ERP or the planning software until someone manually updates it, so the MRP looks healthy while the real schedule has already slipped. By the time the deviation surfaces at assembly, a material shortage occurs and planners have to reschedule the production line.

Expedite logs and batch open-order reviews only surface PO changes days later, after someone reads the reply and manually updates SAP. By then the window to act has closed.

What never reaches the ERP

The information never makes it into the ERP: a revised promised date lost in the inbox, a wrong PO line quantity, or an early delay warning from the supplier that the buyer failed to enter. By the time the production team notices the missing material, they are forced to go into firefighting mode.

This is how most manufacturers operate today, despite having best-of-breed ERPs and planning software. The real reason for this is that collaboration and communication with suppliers remain in emails; unless manually captured and entered by buyers, this information evades the organization's systems.

Take, for example, a manufacturer whose supply chain team manages thousands of suppliers globally, each confirming POs in different formats (PDF, Excel, free-text emails). The buyer team handles each one individually, validates the information, cross-references it with master data, and enters it into the ERP. With just a small human error in a date or elsewhere, the team faces an inevitable disruption, starting with a document.

An AI agent handles that work instead. Every format and language is validated against the original PO. If there is a discrepancy, the AI replies to the supplier, just like a buyer, and upon clarification automatically updates the ERP with a clean, formatted record.

This is done without requiring the buyer team or supplier to adopt a new tool or process. The entire workflow operates out of the buyer team's inbox.

What the buyers do now

The buyer team in that example isn't processing confirmations anymore. They're managing supplier relationships, identifying strategic opportunities to lower costs, and occasionally making decisions that actually require human intervention.

If you are working on the same problem, I'm happy to compare notes and show you how other manufacturers have approached it. There is a write-up on automating exception management here if you would rather read first.

First published on LinkedIn on 14 July 2026.