Process automation
Orders, invoices, onboarding and reports move between your systems on their own. No re-keying, no "did anyone send that?"
Microsoft 365 · Google Workspace · Shopify · Amazon · NetSuite · Salesforce · n8n
Automation + AI for small and mid-sized businesses
The repetitive work your team does every week, done by systems that don't get tired.
I connect the tools you already pay for, add AI where judgement helps, and keep it all monitored and backed up. You get hours back and fewer things falling through the cracks.
Email me
No pitch deck. We look at one process and see what it would take.
amazon-settlement-to-ledger · example run
3 for review
- 07:00Amazon settlement report in: 412 orders, 38 refunds, £61,204 deposited
- 07:01Each line matched to a NetSuite invoice, fees and FX split out
- 07:02409 matched. 3 unmatched: 2 chargebacks, 1 refund with no return
- 07:02Journal for fees, refunds and the deposit drafted, not posted
- 07:03Finance gets a workbook with the 3 exceptions and the reasons
412 lines · 3 minutes · a day of spreadsheets, gone
supplier-invoice-three-way-match · example run
1 flagged
- 11:20Invoice PDF arrives in accounts@, supplier and PO number read off it
- 11:20Matched to PO-0917 and the goods receipt in NetSuite
- 11:21Quantity short: 180 received, 200 invoiced. Unit price matches.
- 11:21Sent to the buyer with the numbers side by side
- 13:05Buyer approves 180, credit note requested from the supplier, bill posted
caught a £62 overbill · one approval · no re-typing
leaver-offboarding · example run
verified
- 17:30Leaver date reached in the HR sheet
- 17:30Mailbox and OneDrive backed up, restore tested
- 17:32Sign-in blocked, sessions revoked, mailbox shared with the manager
- 17:33Laptop locked, 4 SaaS seats and 2 Microsoft licences released
- 17:34Every step re-checked, report sent to IT and HR
6 systems · 4 minutes · nothing forgotten on a Friday
support-inbox-triage · example run
2 for approval
- 09:0014 emails overnight in support@
- 09:019 "where is my order": tracking looked up, replies sent
- 09:013 invoice copies: PDFs pulled from NetSuite, replies sent
- 09:022 return requests drafted, held for a person
- 09:02Support opens 2 emails instead of 14
14 emails · 12 answered with real data · 2 for a human
Three things I do for SMEs
AI where it earns its keep
Agents that read documents, reconcile numbers, summarise meetings and triage requests. Running on your data, with a human on the last step.
Agents · MCP · RAG · Claude · OpenAI
Systems that stay up
Self-hosted or cloud, with single sign-on, backups that are tested, monitoring that reaches a person, and nothing exposed to the internet.
OIDC/SAML · WireGuard · restic · Zabbix · Proxmox · Cloudflare
Recent work
- Reconciliation
- Settlement reports matched against payments by an agent. A day of spreadsheets became a form upload and an emailed workbook.
- Employee onboarding/offboarding
- Account, MFA, groups, device and licence created in one run. Offboarding backs everything up and verifies it before anything is removed.
- Supplier invoice matching
- Invoice PDFs read from the accounts inbox and matched against the purchase order and goods receipt in NetSuite. Short deliveries and price changes go to a buyer with the numbers; everything else posts itself.
How it works
Map one process
We pick the one that hurts most and write down every step.
Automate the boring 80%
Your existing tools, connected. Usually live within weeks.
Add AI where judgement helps
Only after the plumbing works, and always with a person on the last step.
Monitor, back up, hand over
Documented, alerting, and yours. No lock-in to me.
I'm Yash.
I run IT operations and automation for a manufacturing and e-commerce business, and build the same for small and mid-sized companies that don't have an IT department to spare.
Email me
Bring one process that wastes your team's time. You'll leave with a plan either way.
Email: yash@cyberblaze.dev