AI Copilot for Custom or Homegrown ERPs — Bring Your Own Data, Get an AI Operating Layer

Even without an API, the copilot can run in CSV or Google Sheets mode. Your ERP or accountant exports the invoice register, customer master, and vendor master to a shared folder or Google Sheet on a schedule you control. The Perimattic ingest job picks up files, deduplicates rows against the previous snapshot, and runs the same query, anomaly, and forecasting workloads against the delta. This mode works for FoxPro-based systems, MS Access databases, AS/400 green-screen ERPs, and any Windows tool that produces a scheduled export. Data volumes up to about 5,000 rows per day are supported without extra infrastructure on your side.

That is the fastest path to go live. The copilot ships a pre-built template for each of the three platforms — you import it into your workspace, wire your ERP's existing event source to it, and the template routes events into the Perimattic ingest layer. Customers with an existing n8n workflow are usually running their first live copilot query within 24 hours of signing up. The middleware handles the transport; the copilot handles the AI workload.

The copilot only receives the fields you explicitly map at scoping — customer name, invoice number, invoice amount, due date, and optionally an invoice PDF URL. Your general ledger, purchase register, journal entries, and payroll data never leave your systems unless you deliberately include them in the mapping. Data transits over TLS 1.3, processing happens in your chosen region (default India), and retention is configurable — default 90 days, configurable up to 7 years for tax-audit compliance. A DPDP Act 2023 aligned Data Processing Agreement is signed before ingestion begins.

Custom ERPs evolve — new columns appear, invoice numbering formats change, GST rate columns get renamed. The copilot uses a version-locked field mapping, so a schema change does not silently break the AI workload. If a mapped field is missing or renamed, the copilot pauses new ingestions for that data type, sends a schema-drift alert to your registered ops contact, and shows the offending rows in the ERPPlugs dashboard for manual acknowledgement. A remap is done in the dashboard without a code change. This is the same safety net used for our TallyPrime and Zoho customers, adapted for custom sources.

A typical custom-ERP onboarding runs 3–7 working days. Day 1–2 is a discovery call and data-mapping workshop with the Perimattic solutions team to understand your data model, ingestion mode preference, and priority use cases. Day 2–4 is the ingestion setup itself — REST endpoint deploy, CSV pull configuration, or middleware template import. Day 4–5 is verification against a controlled query set. Day 5–7 is user-acceptance testing before going live. Complex ERPs or non-standard invoice formats may extend this by 3–5 days.

Every custom-ERP onboarding ships with a runbook covering: the exact field mapping your ERP uses; how the ingestion mode works (REST vs CSV vs middleware) including the credentials and endpoints in play; how to add or rename a field without breaking the copilot; how to trigger a manual reprocess if an ingestion misses a batch; how to interpret the schema-drift alerts; and how to escalate to Perimattic if something outside the runbook happens. The goal is that your ops team can run the copilot without our involvement — the retainer is for improvement, not for keeping the lights on.

Early access with hands-on guidance. Because every custom ERP is different, we cap the onboarding rate to roughly two new custom-ERP customers per week — that is what our solutions team can support properly given the variability. If your ERP fits one of the three ingestion modes (REST + webhooks, CSV, or middleware), contact us to join the pilot. If it does not fit any of the three, we will tell you honestly at scoping instead of forcing an integration that will not hold up.

The Perimattic AI Suite ERP AI Integration service is a bespoke build — we design and deploy custom ML modules inside your ERP under a fixed-scope engagement. AI Copilot for Custom ERPs is a subscription product — the same off-the-shelf query, anomaly, and forecasting capabilities used by our Tally and Zoho customers, brought to your ERP through one of the three ingestion modes. If you want a specific ML model (say, industry-specific demand forecasting), the AI Suite is the right path. If you want the general operating-layer capabilities that most of our customers use daily, the Copilot is faster to deploy and cheaper to run.