Services
Six things businesses ask me to automate
I researched hundreds of real automation job posts before building my toolkit — these six services cover the overwhelming majority of what companies actually need. Every one is backed by working, open-source code you can inspect before we ever talk.
01 — the most requested
Dashboards, reporting & monitoring
One live dashboard and one morning digest that replace the weekly spreadsheet ritual. KPIs, trends, and — more valuable — alerts when something moves: a revenue dip, a cost spike, a metric drifting out of range.
- Live KPI dashboards fed by your actual systems (orders, calls, CRM, ads)
- Morning digest to Slack or email — the 30-second read that replaces five tabs
- Anomaly detection tuned to stay quiet unless it matters — alert fatigue kills dashboards
smb-ops-dashboard · ecom-ops-radar
A zero-dependency ops dashboard with Slack digest, and an e-commerce radar that catches revenue anomalies, return spikes, and stockouts before they hurt. Both run in one command.
02 — the silent revenue leak
Booking, reminders & no-show recovery
Most reminder tools stop at the reminder. The money is in what happens after the miss: an automatic day-1 rebook text and a day-3 offer, with a measured recovery rate.
- Confirmation → T-24h and T-2h reminders → no-show detection → win-back sequence
- Crash-safe and idempotent: the system can restart without ever double-texting a customer
- Sequences stop instantly when the customer replies — no tone-deaf follow-ups
appointment-noshow-recovery
Full state machine with a 10-day simulation you can run in one second — watch bookings confirm, miss, and get recovered, with the recovery rate printed at the end.
03 — where leads go to die
CRM & lead operations
Respond to new leads in seconds, score them with reasons a sales team can argue with, route them to the right person, and keep the data clean enough to trust.
- Speed-to-lead: instant first-touch the moment an inquiry lands
- Explainable lead scoring — every point has a stated reason, no black boxes
- Deduplication across systems (CRM vs billing vs marketing) with a human-review queue for risky merges
- Follow-up cadences that know when to stop
realestate-lead-engine · crm-dedupe-kit · lead-scoring-pipeline
Lead cadence engine, a dedupe tool that collapses 30 messy rows into exactly the 15 real people, and a scoring pipeline where reasons always sum to the score (tested).
04 — done right, not done loudly
AI agents & LLM integration
I use Claude and GPT in production daily — and the most important thing I know about them is where they don't belong. My AI layers handle the ambiguous 20% while deterministic rules handle the boring 80%, and every AI failure degrades to a safe fallback instead of a crash.
- Email/ticket triage, document intelligence, summaries, drafting — with human review where it counts
- MCP servers: give Claude safe, read-only tools over your business data, so you can ask questions in plain English and get grounded answers
- RAG assistants that cite sources and refuse to guess
smb-ops-mcp · inbox-triage-bot
A working MCP server exposing an ops database as Claude tools, and a triage bot whose LLM layer is an enhancement with a rules fallback — never a dependency.
05 — the plumbing
API integrations & data pipelines
Webhooks, syncs, scrapers, ETL — the connective tissue between the tools you already pay for. Built to survive the real world: retries, idempotency keys, dead-letter logging, and alerts when an upstream API silently changes.
- Two-way syncs between CRM, billing, calendars, phone systems, spreadsheets
- Webhook receivers that never lose an event, even mid-crash
- Deployed on AWS infrastructure you own — with systemd, cron, monitoring, and a runbook
War story: I once found a phone API whose call index silently omitted forwarded calls — the exact calls a client's ad tracking needed. Polling looked fine and was quietly losing data. The fix (webhooks + reconciliation) is why their attribution numbers are trustworthy today.
Call attribution pipeline
Google Ads → call forwarding → phone API → CRM, matching inbound calls to ad campaigns in real time. Attributes $100K+/year in ad spend; removed 5+ hours/week of manual matching.
06 — paperwork on rails
Documents, billing & back-office ops
Quotes, proposals, invoices, onboarding packets — generated from your data, escalated on a sensible ladder, and always drafted for human review before anything customer-facing goes out.
- Invoice follow-up ladders: friendly at 7 days, firm at 30, final at 60 — and at 90+ it stops emailing and tells the owner to call
- Client onboarding as a versioned pipeline: emails, tasks, and checklists generated per client
- Document intelligence: extract, route, and act on PDFs and contracts
invoice-chaser · client-onboarding-engine
The escalation ladder and the onboarding pipeline, both runnable in one command. In production I've built document-intelligence systems that route contracts unattended.
Not sure which one you need?
Most projects start as "we waste hours every week on X." Describe the X — I'll tell you honestly whether it's worth automating, what it would take, and what I'd build first.