Quality-Issue-to-ECO & Quote Agent
Built on requestEvery change starts somewhere. It should end in a clean ECO and a quote.
Pick where your change starts
Agent 5
Agent 5
Quality-Issue-to-ECO & Quote Agent
- AI reads the input, then coded rules (fixed logic, not AI) check what it found, with sources.
- Plans what each change needs, checks your files for it, drafts the ECO and prices the quote.Read-only: it never changes your files, and works within set limits.Cited: every finding names the file and row it came from.Reviewed: it goes over its own work before handing it to you.
- People review the draft and make and record every decision. Routed to Quality, Engineering, and others as appropriate.
- Switch on only the follow-up actions you need: emails, tasks, and updates in your other systems such as SAP, Oracle, Jira or Teams. Not in this demo: we’ll discuss your follow-up actions in a discovery call. Once switched on, nothing changes in your systems without your approval.
The steps above combine:AUTOMATIONAIHUMAN-IN-THE-LOOP
Automate any of 30+ follow-up actions
switch on what you needNot in this demo: we’ll discuss your follow-up actions in a discovery call.
ECO initiation
Design engineering
Document control
Manufacturing eng.
Purchasing & supply
QA/RA
Project leadership
Agent 5Built on request
Turn CAPAs and NCRs into ECO drafts and quotes, then automate 30+ actions
Let’s build this agent. Tell us about your current process and requirements.
About the Quality-Issue-to-ECO & Quote Agent
What a custom AI agent could do with a CAPA or a nonconformance report: turn the fix into a draft ECO and a quote, for your team to review.
We build custom AI agents around each client's own process. This page shows one we would build for medical device makers and contract manufacturers: an agent that starts from a quality record, a CAPA or a nonconformance report (NCR), and prepares the change that fixes the problem. It is part of our Change Control Agent Suite. Its first agent, the BOM-to-ECO & Quote Agent, is live, and you can try it with sample files or your own.
See the approach working: try the live BOM-to-ECO & Quote Agent →
The manual work today
Once an investigation points to a fix, such as a different material, a tighter tolerance, a new inspection step or a process change, the fix itself has to become a controlled change. Someone has to carry the CAPA's findings into an ECO, work out which parts, assemblies and devices are affected, decide what happens to stock and to units already built, and make sure the change doesn't create a new problem while fixing the old one. The CAPA and the ECO often live in different systems, so the link between them is kept by hand.
Why it matters under the QMSR
ISO 13485:2016 §8.5.2 asks you to verify that a corrective action does not adversely affect your ability to meet regulatory requirements or the safety and performance of the device. Under §8.3, nonconforming product needs a documented disposition: rework, concession or otherwise. Since FDA's Quality Management System Regulation took effect on 2 February 2026, 21 CFR Part 820 incorporates ISO 13485 by reference.
Verifying the fix is where it often falls short. In a 2018 warning letter, FDA quoted a manufacturer's own procedure, "verification will be made that the corrective and preventative action is effective and will not adversely affect the manufactured product", and found that corrective actions for one issue "had 0 days to determine if these actions were effective."
What the agent would do
The same four steps as every agent in the suite, applied to a quality issue:
AI reads · coded rules check
Read the quality record
AI reads the CAPA or NCR: the problem, the affected part numbers and lots, the root cause and the proposed fix. Coded rules (fixed logic, not AI) then check what it found, and each item is designed to point back to where it came from in the record. Anything the record doesn't settle, such as which lots are affected, is designed to be listed as open rather than filled in.
AI agent · plans, checks, drafts, prices
Draft the ECO and the quote
The AI agent plans what the fix needs, checks your files and drafts:
- Where it's usedthe BOMs, assemblies and finished devices in your files that the fix touches.
- Stock, lots and units builton hand, work in process and finished units in the affected lots, with a suggested disposition for each (rework, scrap, use as is under concession, hold) and the quantity it rests on.
- The link backthe ECO carries the CAPA or NCR number, so the change and its reason stay connected.
- Open changesother open ECOs on the same parts in your ECO log, to help keep two changes from colliding.
- Regulatory questionsthe same seven checks as our live agent (significance, FDA filing, risk, testing, process / labeling / device ID, documents and training), plus whether units in the field need action.
- Cost and quotematerial, rework or scrap, verification and any field action, rolled into a cost impact and, where a customer pays for the change, a quote.
People · review, decide, send
People review, decide and send
The draft goes to Quality, Engineering and Manufacturing. Your team makes and records every decision, including whether the fix is effective; the agent prepares the work and decides nothing.
Automation · optional
Follow-up actions, if you want them
Emails, tasks and updates in systems such as SAP, Oracle, Jira or Teams, each switched on separately and designed so nothing changes in your systems without your approval.
Each finding is designed to name the file and row it came from, and a check whose file wasn't provided shows as Not checked, not as a clean result. The output is only as complete as the records and files it is given, and AI can misread a record, which is why people review every draft.
The checks, and where they come from
| Check | Requirement |
|---|---|
| Link the change to its CAPA; check the fix for adverse effects | ISO 13485 §8.5.2 (via 21 CFR 820.10) |
| Disposition of nonconforming product (rework, concession) | ISO 13485 §8.3 |
| Does the fix change the design? | ISO 13485 §7.3.9 |
| Does the change need a new 510(k)? PMA supplement? | 21 CFR 807.81(a)(3); 21 CFR 814.39 |
| Labeling and packaging affected? New UDI device identifier needed? | 21 CFR 820.45; 21 CFR 830.50 |
| Units already in the field: correction or removal? | 21 CFR Part 806 |
Each requirement, with its source, is on our QMSR mapping page. The agent raises these questions and shows its working. Whether the fix is effective, and whether a field action or filing is needed, stays your team's call.
What your team would get
- A draft ECO linked to its CAPA or NCR: the affected parts, assemblies, lots and devices found in your files, the findings and the regulatory questions answered or left open, for your team to check and complete.
- Disposition suggestions for stock, work in process and units built, each with the quantity it rests on.
- A cost impact of the fix, and a quote where a customer pays for it.
- A record that connects the problem, the fix and the decisions.
How we'd build yours
Every company runs CAPA and nonconformance its own way: the quality system that holds the records, how investigations are written, how ECOs are numbered and routed. So we don't sell this agent off the shelf. We start from your process:
- 01
Discovery
we map how quality issues become changes today, with the people who handle them.
- 02
Proof of concept
a working version on a few of your real CAPAs or NCRs, so you can judge the output for yourself.
- 03
Build and integration
connected to the systems you choose, such as your quality management system, ERP or PLM, and tested against your cases.
- 04
Go-live and tuning
your team trained, results monitored, rules adjusted as your process changes.
The live BOM-to-ECO & Quote Agent shows the approach working today, on the same engine this agent would use.
Questions buyers ask
Is this a product we can buy?
Not off the shelf. We build it for you, around your process, your customers' requirements and your systems (ERP, PLM, quality). The best way to start is with a 2-week pilot around your needs. Book a Discovery Call →
Does it do the investigation or root cause analysis?
No. It starts once your team has a proposed fix, and prepares the change that carries it out. The investigation stays with your people.
Can it judge whether a fix worked?
No. It is designed to prepare the change and record the evidence your team gathers. Deciding effectiveness stays with your team.
Is it validated?
Not out of the box. If your team uses an agent like this as part of your quality system, ISO 13485 §4.1.6 asks you to validate that software for its intended use, in proportion to the risk. We design our builds with that in mind and can talk through what your validation would need.
Talk to us about an agent built to your process.
Book a Discovery CallDisclaimer: This page describes an example of an agent we would build to a client's process: how it is designed to work, not a product or a guaranteed result. Any agent we build is advisory and not a validated system. AI can make mistakes, results depend on the data it is given, and the regulatory checks raise questions without deciding them. Your team reviews the output and makes and records every decision. Time figures are estimates, not measurements. Nothing on this page is legal or regulatory advice.