What lands on the IT desk of a textile mill
Most of it starts outside IT: a planner’s Excel tracker, a buyer’s EDI change, a phishing email to a merchandiser. IT is asked to explain it, fix it, and then prove to a buyer that it is under control.
The ERP says one number, the shadow sheet says another
Planning, stores and merchandising keep Excel trackers because the ERP report lacks a field they need: a buyer reference, a shade group, a new ex-factory date. The sheet drifts, and a stock or WIP report sent to the buyer contradicts the ERP. IT is asked which is right. In a 2021 LogForum study of Bangladeshi apparel IT teams, a systems developer said wrong inventory, invoice and shipment data needed fixing often.
5.6%
Share of Vietnamese firms whose most-used tool for business administration is an ERP; 14% rely on handwriting. All sectors, surveyed before COVID-19; exporters use more advanced tools.
Source: World Bank (Policy Research Working Paper 9567), Firm-Level Technology Adoption in Vietnam (opens in a new tab), 2021.
One cyberattack can stop production for weeks
An attack locks the ERP, the planning files and the file server, and cutting, dispatch and invoicing stop with them. In January 2020, a ransomware attack hit Picanol’s weaving-machine sites in Ypres, Romania and China and halted production for a week, the company said. Afterwards the questions come at once: is the last backup restorable, who has been told, and what does the incident-response plan say to do first?
Nearly 6 weeks
Production lost at a German textile finisher after a cyberattack in March 2026. The company then filed for insolvency, blaming the attack’s financial fallout.
Source: The Register, German firm files for insolvency, blames cybercrims who shut down production for 6 weeks (opens in a new tab), 2026.
Patches, backups and access that are hard to prove
A buyer’s audit or a cyber-insurance renewal asks for evidence: critical patches applied within the policy’s window, the date of the last restore test, who still holds ERP access. The answers sit in the patch tool export, the backup log and the access-request sheet, and someone counts them by hand. Leavers can keep their logins, and a temporary exception can become permanent.
26%
Share of critical vulnerabilities on the CISA known-exploited list that organisations in Verizon’s dataset fully remediated in 2025, cross-industry. In manufacturing, 38% of breaches began with an exploited vulnerability.
Source: Verizon Business, 2026 Data Breach Investigations Report: Executive Summary (opens in a new tab), 2026.
An interface fails overnight, and nobody can say what it covers
The ERP feeds the buyer’s EDI, the bank file and the carrier booking. One fails at night, the message sits in the failed queue, and by morning an 856 ASN is missing or a goods receipt is unposted. The mapping lives in one consultant’s document or one engineer’s head, and a buyer can change its EDI mapping without notice. Shipping and stores wait while IT works out which fields the interface carries.
Policy quoted, failures counted, vendor emails drafted
It works from the files your IT team keeps, such as the integration map, logs, policy and registers, uploaded or in Google Sheets. Each answer names the file and rows used. No email, sheet row or Slack post it drafts runs until an admin approves it.
The security policy and incident plan, quoted to the clause
Reads your documents
Upload the security policy, the incident-response plan and the backup runbook. Ask what a contractor needs before getting ERP access, who must be told in the first hour of an incident, or the call-back rule before a supplier’s bank account is changed. The agent quotes the clause and names the document it came from. If the policy does not cover the case, it says so.
Ask“What does our security policy require before a contractor gets ERP access, and who approves the request?”
Failures and open incidents, counted from your rows
Answers with sources
Upload the month’s interface error log or incident log as an Excel or CSV file. Ask for a few counts per answer: the failures on each interface you name, on each of three error codes, or the open P1 and P2 incidents. It counts from the rows and lists them; for every interface, it works through a few at a time. When the log records resolution hours, it averages them per severity, so MTTR can be checked line by line.
Ask“From this month’s error log, how many times did the ASN interface fail, and which error code is recorded against each failure?”
The dashboard’s source, checked against an ERP export
Reads your documents
When the efficiency dashboard and the ERP disagree, the agent reads your BI dashboard catalog for the dashboard’s source, owner and refresh time. Upload the ERP’s production report for the day as a CSV or Excel export, and it sums the line’s output from those rows and lists them. It opens neither the dashboard nor the ERP, and changes nothing in either.
Ask“The sewing line efficiency dashboard and the ERP disagree on yesterday’s output for line 4. Which source does the dashboard draw from, how often is it refreshed, and what does the ERP export I uploaded show?”
A systems status summary, built from the map and the logs
Creates files
Ask for the status summary and the agent builds it as a document from the integration map and the logs you uploaded: each interface marked healthy, degraded or failing by what the logs record, with the rows behind each mark. It is a summary of your files, not a live view. Ask for a change to one section and it edits that part; every revision is kept.
Ask“Make this week’s systems status summary from the integration map and the error logs I uploaded.”
The email to the ERP vendor, drafted from the error rows
Acts in your tools
When an interface keeps failing, the agent prepares the email for your ERP vendor or the buyer’s EDI team, to go out through Gmail or Outlook. It puts in the error codes, message IDs and times from the log rows. It can also prepare the matching row for your incident sheet in Google Sheets. Each one waits for an admin to approve or decline it, and it runs only if approved.
Ask“Draft an email to our ERP vendor about the 856 ASN failures, with the error codes and times from today’s log.”
An IT team’s week, question by question
A fabric stock report to the buyer contradicts the ERP
ERP support engineer
The fabric stock report sent to the buyer, built from the store’s tracker, shows more of a fabric than the ERP’s stock report. The owner asks which is right.
Asks
“The ERP stock export and the store’s tracker disagree on fabric FB-2210. Show both figures, where each comes from, and what the integration map says about that field.”
The agent
Reads the item’s row in the stock report exported from the ERP and its row in the store’s tracker, both uploaded to IT. It shows the two figures side by side with the unit each file records, such as kilograms in the export and metres in the tracker, and quotes the integration map’s line for the stock field. Asked which figure the business should use, it says that belongs to Stores & Inventory.
Result
The engineer sees where the two figures part, each with its source and unit, and can fix the source as the playbook says, then retire the sheet.
The ASN feed to a buyer failed overnight
Integration owner
Morning. The buyer’s EDI team reports no 856 ASN for two shipments, and last night’s interface error log is uploaded as a CSV export.
Asks
“What does the ASN interface to this buyer cover, and which error is recorded against each failed message in last night’s log? Then draft an email to their EDI team.”
The agent
Reads the integration map’s entry for the interface (source, target, cadence, owner, retry policy) and the buyer’s mapping spec. Counts the failed messages on that interface in the uploaded log and lists each row with its message ID, time and error code. It then prepares an email to the buyer’s EDI team, to send through Outlook, with those rows and the fields the map says the ASN carries.
Approval
The email waits for the IT head, as department admin, who approves or declines it. It is sent only if approved, and the decision, either way, is kept in the audit trail.
Result
The integration owner has an email for the buyer’s EDI team that lists each failed message and its error, taken from the log rows rather than from memory.
A buyer’s security questionnaire asks for evidence
IT head
A buyer’s annual audit includes a security questionnaire that asks for patch status and the date and result of the last restore test.
Asks
“Count the open critical and the open high patches the patch register marks overdue, give me the last restore test and its result, and make the evidence table.”
The agent
Counts the open critical and the open high patches whose overdue column says Yes in the uploaded patch register, and lists the rows for each. Reads the restore-test log for the latest test date and result. Quotes your security policy’s patch windows beside the playbook’s: critical under 14 days, high under 30. Builds the evidence table as a spreadsheet, each line citing its file.
Result
An evidence table for the questionnaire with every line traced to a file and each overdue critical and high patch listed by name, so the answer to the buyer matches your own records, line by line.
Sizing the risk: published figures on data, attacks and fraud
These are published figures on spreadsheet errors, incident readiness and email fraud, each cited at the source. None is a Mavick result. Before a pilot, baseline your own IT KPIs below from your logs and registers, so the pilot is judged on your numbers.
Shadow sheets set beside the ERP export, row by row
86%
Share of 50 operational spreadsheets, audited cell by cell, that held at least one error giving a wrong result. Workbooks from consulting firms, a bank, an energy company, a state agency and others; not textile; 2009.
How the agent helps
The agent does not fix the ERP or retire a sheet. Upload the ERP report beside a shadow sheet and it shows the matching rows from each file. List the Excel trackers each department keeps (sheet, owner, the field the ERP lacks) and upload the list: it counts the sheets kept for each missing field you name, such as the buyer reference, the playbook’s gaps to close.
Track: Shadow sheets in use, by process and missing ERP field
Source: Journal of Organizational and End User Computing (IGI Global); Powell, Baker and Lawson, Errors in Operational Spreadsheets (opens in a new tab), 2009.
The incident plan, answerable when the IT head is away
25%
Share of UK businesses with a formal incident response plan; 57% of medium and 76% of large businesses have one. Official UK survey, all sectors, fieldwork late 2025; not textile.
How the agent helps
The agent does not detect or stop an attack. It keeps the incident-response plan, the backup runbook and the escalation contacts answerable to whoever is on duty in the IT department, quoting the step and its source. Afterwards, ask and it builds the incident timeline from the incident log rows as a document.
Track: Incident MTTR by severity (P0, P1, P2)
Source: UK Department for Science, Innovation and Technology and Home Office, Cyber security breaches survey 2025/2026 (opens in a new tab), 2026.
The call-back rule on file, and phishing clicks counted
$3.05 billion
Losses to business email compromise reported to the FBI’s Internet Crime Complaint Center in 2025, from 24,768 complaints, mostly from the US. All sectors, not textile; under-reported.
How the agent helps
The agent does not filter email or spot a fake invoice. It quotes your policy’s call-back rule for a changed supplier bank account when the IT team asks; share the policy across the factory and each department’s own agent can quote it too. From the uploaded phishing-test results, it counts the clicks for the departments you name, a few per answer.
Track: Phishing-simulation click rate, by department
Source: FBI Internet Crime Complaint Center (IC3), 2025 IC3 Annual Report (opens in a new tab), 2026.
What to baseline before a pilot
Measure each one before the pilot starts, then again at its end.
| KPI to baseline | Why it matters |
|---|---|
| ERP uptime % | The playbook targets 99.5% or more. When the ERP is down, planning, stores and dispatch stop together, so total the downtime from your outage log, not from memory. |
| Integration success rate, per interface | The playbook sets 99.9% or more for each interface. An average across all interfaces can hide one failing feed, such as the ASN to one buyer. |
| Patch compliance % within the policy’s window | The playbook wants 100% within SLA: critical under 14 days, high under 30, medium each quarter. Count critical, high and medium from the patch register in one answer, so each overdue patch shows by name. |
| MES data coverage %, per shift and machine | The playbook targets 95% or more per shift, per machine. Mavick does not connect to the MES; count the gaps from an MES export you upload, if your MES gives one. |
| Phishing-simulation click rate | The playbook targets under 5%, tested each quarter. It reads a rising rate as a failure of training, not of users, and calls for retraining. |
| Incident MTTR, by severity | Read it against the playbook’s incident SLA: P0, a production halt, under one hour; P1, a broken workflow, under four hours; P2, cosmetic, in the next planned release. |
What the agent works from
Records it reads
- ERP integration map
- MES specification
- BI dashboard catalog
- Security policy
- Interface error log
- Incident log
- Licence register
- Backup and DR runbook
In your ERP
The factory’s ERP connection gives the IT agent no records, and the agent writes nothing to the ERP. Its settings, user roles, batch jobs, queues, stock and orders reach IT as Excel or CSV exports you upload; incidents and access requests from a connected Google Sheet; vendor and EDI emails from Gmail or Outlook.
One ERP connection serves the whole factory: SAP Business One, SAP S/4HANA, Odoo, Oracle NetSuite, Microsoft Dynamics 365, WFX Cloud ERP or Datatex NOW. This agent reads only the records named above; we check yours on the call.
In your tools
Reads a vendor’s or buyer EDI team’s emails in Gmail or Outlook, and your incident or access-request sheet in Google Sheets, when you ask. Prepares an email for the ERP vendor, a new row for the access-request sheet, or a Slack post to the IT channel. An admin approves or declines each one; it runs only if approved.
Tools named today: Gmail, Outlook, Google Sheets, Slack and WhatsApp Business.
Never made up
- System configurations, access rights or credentials
- Integration mappings, field definitions or data flows
- Incident causes or resolution status
- Licence counts, costs or vendor commitments
If it is not in your connected data, the agent says so.
Belongs to another department
- Buying hardware, licences or servicesSourcing & Procurement
- Data-protection contracts and regulatory filingsLegal & Regulatory
- Physical site securityAdmin & Security
- The business meaning of a stock figure in a systemStores & Inventory
- The business meaning of an invoice or ledger figure in a systemFinance & Accounts
What the IT & Systems playbook covers
SOP: the ERP support day
The playbook’s daily ERP health check covers the nightly batch runs, the integration queues and the error logs. If the ERP goes down, the playbook has the IT head, the COO and the plant head told within 15 minutes.
Decision rule: a new buyer asks for an EDI feed
For an 850 PO or 856 ASN feed, take the mapping spec from the buyer, build it on your integration platform and test it in the buyer’s test environment. Agree the cutover window. The playbook’s line is firm: do not go live without signed UAT.
KPI: integration success rate, interface by interface
The playbook targets 99.9% or more per interface and has the failed-message queue reviewed daily. If a production-impacting feed has failed for over two hours, it goes to the IT head and the business owner within the same hour.
Heuristic: a backup nobody has restored
The playbook sets daily incremental and weekly full backups, an off-site copy, and a DR drill once a year. Its rule of thumb: a backup that has never been restored is not a backup, so test a restore every quarter.
Written as general textile practice and reviewed by four independent AI models. Not yet signed off by a human subject-matter expert. Upload your own SOPs and standards, and every answer shows which source it used.
Ten questions it ships with
The template’s starter questions. Your team can edit them, or ask in its own words.
What an integration covers
What does the integration between these two systems cover, which fields does it map, and where is it currently failing?
ERP modules live and their data
Which ERP or MES modules are live, and what data does each hold? Show the integration map.
What a dashboard reports
What does this BI dashboard report, which source does it draw from, and how often is it refreshed?
Access policy for a system
What does the security policy require for access to this system, and what does the approval process involve?
Where a data discrepancy starts
These two systems disagree on this figure. Trace the field mapping and sync history and tell me where the discrepancy originates.
Systems status summary
Prepare a systems status summary from the integration map and logs: what is healthy, what is degraded, and what is failing.
Open incidents and their age
Which IT incidents are open, what is the impact of each, and how long has each been open?
Integrations failing most often
Which integrations have failed most often this month, and what error is recorded against each?
Licences and renewal dates
Which software licences do we hold, how many seats, and when does each come up for renewal?
Shadow spreadsheets in use
Which business processes are still running on spreadsheets outside the ERP, according to our records, and which system should own each?
IT & Systems: questions before a demo
Can the agent read our ERP’s error logs, user roles or batch-job status?
No. The factory ERP connection gives the IT department no records: not the ERP’s settings, user accounts, roles, batch jobs or integration queues, and not its orders or stock. Export what you need, such as the error log, the user list or a stock report, as a CSV or Excel file and upload it to the IT department; the agent then counts from its rows and lists the rows it used. A plain log file is read as text: it can be searched and quoted, not counted.
Will it watch our interfaces and tell us when one fails?
No. The agent answers when someone asks. Nothing checks the failed-message queue, the network or an inbox between questions, so keep your own monitoring tool for that. Upload the latest error log as a CSV or Excel file and ask: the agent counts the failures on the interfaces or error codes you name, a few per answer, from the rows and lists them. It does not retry a message or restart a job.
Can it reset a password or remove a leaver’s ERP access?
No. It does not change access rights, passwords or settings in any system. It can quote what your security policy requires when someone leaves and list the open requests in your access-request sheet. It can also draft a new row in that Google Sheet or an email to the system owner. The row and the email wait for an admin, who approves or declines each; nothing runs unless approved. It states who holds access only when that is on record.
Can it read data from our MES or the edge devices on the shop floor?
No. Mavick does not connect to the MES, PLCs, edge devices, RFID or barcode systems. If the MES exports a per-machine downtime or data-coverage table as a CSV or Excel file, upload it. The agent then counts from its rows, such as the gaps for one machine on one shift, when the table records them. The MES specification itself can be uploaded and quoted like any other document.
What does Mavick keep about our ERP login, and who can see IT’s files?
Credentials for your connected systems are encrypted. The IT agent sees only what is connected to the IT department, plus files shared across the factory. That boundary is built into how Mavick stores and retrieves everything, so other departments’ agents cannot read IT’s runbooks or logs. Removing someone’s access to Mavick takes effect at once, everywhere. Two-factor sign-in by email is optional, every action is kept in an audit trail nobody can alter, and Mavick holds no compliance certifications yet.
Our incidents live in a helpdesk tool. Can the agent read our tickets or raise new ones?
Not directly. Your helpdesk or ticketing tool is not one of Mavick’s connectors, so the agent cannot read a ticket there or open one. Export the tickets as a CSV or Excel file and upload it, with the table on the first sheet, up to 5,000 rows: the agent counts the open incidents for each severity you name and, when the export records resolution hours, averages them for the resolved incidents of each. If your team logs incidents in a Google Sheet, it can read it and draft a new row, which an admin approves or declines; it is added only if approved.
Bring IT & Systems. We’ll build it on the call.
Book a 30-minute demo (opens in a new tab)- support@mavick.si
- Demo calendar
- calendly.com/safock/30min (opens in a new tab)