How EMS Can Drive Profit With AI

How EMS Can Drive Profit With AI

By Everett Frank, DigiSource

How can an EMS company increase profits in the next 12 months?

New customers? Perhaps, but unlikely. Consider:

  • On average, EMS operations add one significant new customer per year (significant meaning >10% of revenue).
  • First year customers are rarely profitable due to the greater effort needed to bring on a new customer.

First year customers might increase revenue but they usually don’t increase profit.

Customers typically become profitable in the second year, and optimally profitable in the third year and beyond.

The best way EMS operations can grow profits is with existing customers. Sometimes by capturing market share from a competitor, but usually because their customers are growing.

So the key to growing profits is customer retention.

The Service-Profit Chain

Service-Profit Chain, HBR

This year Harvard Business Reviews Service-Profit Chain turns 30 years old. Here is a short version:

1    Profits are driven by loyal customers

2   Loyal customers are created by empowered employees

3  Empowered employees are enabled by excellent processes

EMS profits are driven by excellent processes focused on satisfying customer needs.

Excellent processes make it easy for employees to keep customers happy, so they like their jobs.

Inadequate processes are frustrating and make it hard to satisfy customers. Dealing with unhappy customers and bad processes all day long leads to burn out and turnover.

Program Managers are the most visible front line employees in EMS and must be supported and happy.

Program Managers are happy when excellent people and excellent processes behind them make it easy to satisfy customer needs.

What do EMS customers need?

In EMS, there are three primary drivers of customer satisfaction:

1     Quality

2     On time delivery (OTD)

3     Responsiveness

Quality and OTD are more like minimum requirements.

Product quality is almost a given. There are ways to differentiate on quality but a +99% final acceptance rate is all but required these days. To differentiate on quality alone is a sophisticated undertaking.

OTD is similar, +90% is expected.

Responsiveness is the key opportunity to differentiate you from your competitors.

By competitors we mean companies with comparable capabilities. If you are a small to mid-sized regional EMS then Flex is not your competitor.

If you’ve spoken with an unhappy EMS customer you will almost always hear one of these:

“I just can’t get answers from…”

“It takes forever to get…”

Responsiveness Simplified

Responsiveness usually boils down to Program Managers ability to answer just one customer question: When can I receive my assembly?

Let’s call this the Delivery Question.

Over simplification? Sure, but not too far off the mark.

Customer interactions with PMs 80-90% of the time fall into one of these scenarios:

👉   Request for Quote                          👉   Order Placement

👉   Order Placement                            👉    Expedite

Every one of these scenarios involves the customer needing to know delivery or the impact on delivery.

How long does it take your PMs to answer the Delivery Question in each of these scenarios?

If it takes them more than 30 seconds, read on.

AI Responsiveness at RFQ

Here’s how AI can answer the Delivery Question for RFQs in 30 seconds:

  1. AI can read almost any BOM in almost any format.
  2. AI can instantly enrich the BOM with real time price and distributor stock, lead times, stock the EMS has on hand for that customer, even equivalent stock on hand or stock earmarked for other customers.
  3. AI can instantly determine the mounting and case types for each component and use this to drive a labor estimate, which can then be used to calculate how much time the order would require at each work station.


All the PM has to do is drop the BOM:

Typical BOM drop window

Yes, there will be exceptions arising from incomplete documentation, fabricated items, or malformed component part numbers.

But the PM can be responsive to the Delivery Question almost instantly and with almost zero effort.

AI Responsiveness at order placement

Often many weeks can pass between RFQ and order placement, or from the last order for repeats.

And often customers will transmit a BOM with the order and leave it to the EMS to figure out if it matches the existing BOM. Sound familiar?

Here’s how AI can answer the Delivery Question at order placement in 30 seconds:

  1. Instantly execute a BOM compare (if necessary).
  2. If a first time build, instantly create the necessary templates to import Items, BOMs, and Routings into ERP.
  3. Instantly fetch real time distributor stock, EMS stock on hand, equivalent stock on hand, and lead times.
  4. With stock data the AI can instantly either confirm the customers request date and/or provide best possible delivery date.
  5. The labor and/or routing can be used by AI to check available capacity.

Within minutes of receiving a PO the PM can answer the Delivery Question and confirm the order.


AI Responsiveness at expedite

Customer expedites come in several forms. Might be a routine status check on a single assembly, or a request to confirm the entire backlog.

Or if the order is late.

Let’s focus on late orders.

Here are the three top reasons orders are late;

  1. Materials
  2. Materials
  3. Materials

The key to resolving late orders is almost always material, i.e. the dreaded Shortage List.

AI can present solutions to the Delivery Question for late orders in 30 seconds by:

  1. Instantly generating a relevant Shortage List.
  2. Enrich the Shortage List with status, either the material order was not placed or the material order is late. Late material orders may or may not have an expected delivery date.
  3. Enrich the Shortage Report with current distributor inventory, including inventory for approved alternates.
  4. Enrich the Shortage Report with suggested alternates including available stock.

There are countless scenarios for shortages, but most can be converted into expected impacts.

If the material is available in distribution you might expect to start the work order in 3-4 days.

If the material is late from a supplier you might expect to start the work order within 5 days.

In late delivery situations communication is the key.

And remember, providing the current status and expected impact does not mean you won’t try to improve it.

The important thing is the PM is once again responsive to the Delivery Question almost instantly and with almost zero effort.

AI Responsiveness to ECOs

We will focus on ECOs to the BOM.

BOM ECOs are always one of three actions:

  1. Add to the BOM
  2. Substract from the BOM
  3. Delete from the BOM

To respond to customer ECO requests the PM must assess these impacts:

  1. Impact on cost
  2. Impact on delivery/lead time
  3. Impact on WIP

Since this article is focused on the Delivery Question, we only need to assess adds to the BOM and ECOs with immediate effectivity.

AI can assess the Delivery Question for ECOs in 30 seconds by:

  1. Instantly conducting a BOM compare and identify adds/subtracts/deletes on the new BOM.
  2. For adds (either quantity increase or new item) can instantly provide distribution stock and lead times.
  3. For immediate effectivity ECOs, can instantly assess material availability to replace material issued to WIP.

Formal responses to ECO requests often require more review, but the key here is the PM can instantly provide feedback on the Delivery Question.

How does AI accomplish all this?

To accomplish all this (and much more) you will need an AI application that uses APIs.

We describe how you can create your own AI app in our article Here’s What AI Can Actually Do Today.

Or you can use SnapChip.

Then you just need to write a prompt. Let’s look at some prompts for a new BOM.


“Please review my BOM and tell me about any issues that may cause problems in volume production, and please create a table including IPN, Part Number, Manufacturer, Leadtime, Lifecycle, and RoHS.”

Maybe your PM is dealing with an automotive customer, they can just change the prompt:


“Please review my BOM and tell me about any issues that may cause problems in volume production, and please create a table including IPN, Part Number, Manufacturer, Leadtime, Lifecycle, RoHS, and Automotive.”

Or maybe the PM wants to price the BOM:


“Please help price my BOM. Please find the lowest prices to purchase 100 sets of my BOM and return a table with these columns: IPN, Part Number, Qty Per, Description, Extended Qty Per, and Best Price.”

Hopefully this illustrates how you can use prompts to accomplish all the use cases we described above.

How much does AI cost?

If you decide to create your own AI app, figure 2-3 months of one person full time.

With the assistance of ChatGPT and some supervision you can use an enthusiastic and tech savvy junior employee.

A solution like SnapChip will cost approximately $99/month/user, less per user for larger companies.

What training is required?

Maybe an hour, probably less. Any employee who has familiarized themselves with ChatGPT will be productive immediately.


We have used the Service-Profit Chain to show the best way to drive profit is by empowering employees to satisfy customer needs.

By investing in AI targeted directly at improving customer satisfaction, an EMS company will increase profits by creating delighted long-term customers.

Whether you create your own AI app or use a SaaS like SnapChip, the potential impact on profit far exceeds the cost.

TrustedParts x A

About The Author