AI agents for spare parts availability requests

This work starts once the part has been identified. If the customer has supplied only an informal name or an incomplete description, someone first needs to establish which product they need. Our article on identifying spare parts from customer descriptions covers that earlier step. Here, the question is what a salesperson can say about a known part using current availability information.

Two on hand, one available to sell

Suppose the part number has been confirmed. The customer needs two units and wants to know when they will arrive. The warehouse being checked holds two, but one is committed to an earlier order. The remaining unit is available to sell. The information available to sales includes neither a confirmed replenishment date nor a ship date.

Replying with the full warehouse quantity would leave out the earlier commitment. Quoting a standard lead time as the arrival date for the full order would also go beyond what sales knows. The salesperson can say that one unit is available. They still need to find out whether the customer would accept a partial shipment and check when that unit could ship. The second unit needs a separate follow-up.

An agent should make this distinction visible beside the customer’s message: the quantity requested, the warehouse, the quantity on hand and the quantity available to sell. A retrieval time and a link to the inventory record help the salesperson check the source without reconstructing the answer across several screens.

Using only the facts in this example, a draft reply could read:

We currently have one unit available to sell. Would you accept a partial shipment? We still need to confirm when that unit can ship. We also do not yet have a confirmed availability date for the second unit.

The customer gets a useful question and an accurate account of what is known. Sales can add a ship date once it has been checked. Preparing the draft does not reserve the available unit. If someone else sells it in the meantime, availability needs to be checked again before the reply goes out.

What the date in your system actually means

Similar field names can answer different questions. A ship date says when the part leaves the warehouse. A receipt date says when it reaches the customer. Substituting one for the other in a short email can create a promise the company did not intend to make.

Microsoft’s documentation for Dynamics 365 Supply Chain Management illustrates why the underlying calculation matters. Its order-promising functionality calculates dates using the selected delivery date control method and transport days. The sales lead time method uses a default number of days without considering stock availability, known demand or planned supply. ATP, short for available-to-promise, calculates the quantity that can be promised for a particular date using uncommitted inventory, lead times, planned receipts and issues. These are product-specific behaviors described in Microsoft’s order-promising documentation.

Before introducing an agent, ask the person responsible for your system to explain the exact field the reply will use. Does it show physical stock, availability after commitments, or a calculation for a future date? Which warehouse does it cover? Different ERP systems and configurations should not be assumed to produce the same answer.

When an inventory lookup is enough

If requests consistently include the correct part number and quantity, and sales can see current availability on one screen, a better ERP view or a simple lookup may be sufficient. Adding an agent to a single lookup may do little to improve the job.

There is more work for an agent to help with when the details are scattered. A customer changes the quantity in a later message, asks about both pickup and shipping, or has not said whether a partial shipment would be useful. An agent can connect the latest request to the inventory data, identify the unanswered question and draft the next reply. Assess it by whether sales can finish those inquiries without rereading the whole thread.

An initial project can have the agent prepare replies while employees continue to reserve stock and confirm dates with customers. If the agent is to take on those tasks too, agree separately on when it can act and what information it must have. The same applies to substitutions: a shortage does not give the agent grounds to declare another part compatible. That decision belongs to the person responsible for product selection.

Checking whether it helps the sales team

For a trial, revisit a few completed inquiries with a salesperson and follow each one from the customer’s message to the reply. A case where the warehouse appeared to hold enough stock, but some was already committed, will be particularly useful. Did the draft preserve that distinction? Could the salesperson immediately see what still needed checking?

Then look at what happens when the information changes. The customer might correct the requested quantity, or the warehouse might correct its inventory record. The agent should use the new information and show the employee that the earlier draft needs revision. If it cannot retrieve current availability, the employee needs to see that problem before sending a reply. Also agree on who corrects an inaccurate inventory record. Editing the email alone leaves the same problem for the next salesperson.

Starting a project with Syntalith

Syntalith builds AI agents that can connect email handling with ERP data retrieval. For this use case, a project would address the work sales repeats with each inquiry: finding the customer’s current request, checking availability and preparing a reply. An initial scope could cover one warehouse and drafts reviewed by employees. That gives the team a limited setting in which to find out whether the difficulty lies in correspondence, access to information or the inventory records themselves.

For a first conversation, describe how your team currently answers a request for two units when only one is available to sell. Explain where you check inventory and who confirms the dates. You do not need a full specification or confidential customer emails to start. The system connection and employee responsibilities can be scoped once we understand that work. See Syntalith’s pricing page for information about service pricing.