Who this is for
A business where someone spends part of every day moving information from one place to another: enquiries into a sheet, orders into an invoice, a report into a slide, a file into a folder, a reminder into a chat. The tasks are too varied for a single-purpose page like lead generation automation or CRM integration, but each follows the same shape: a trigger, a few steps, a result, and a person who no longer has to do it.
Workflows I have built or build regularly
| Trigger | Steps | Result |
|---|---|---|
| New client signs a proposal | Create folder structure in Drive, copy the onboarding checklist, create the CRM deal, send the welcome email and WhatsApp, book the kickoff, post to the team channel | Onboarding done in seconds, the same way every time |
| Month end | Pull Google Ads, Meta, GA4 and Search Console data, write to a sheet, refresh Looker Studio, email the link with an AI-drafted summary for review | Monthly report ready on the 1st; see automated marketing reports |
| Invoice PDF arrives by email | Extract supplier, amount and date, append to the expenses sheet, file the PDF, flag anything above a threshold for approval | No manual expense entry; see document and data automation |
| Google review posted | Alert the owner, draft a reply for approval, log the rating in the sheet, trigger a thank-you if 5 stars | Every review answered within a day |
| Job application received | Parse the CV, score against the role’s criteria, update the tracker, send the acknowledgement, schedule shortlisted candidates | Screening time cut; see HR and recruitment automation |
| Website goes down or a form stops firing | Uptime and heartbeat checks, alert by WhatsApp and email with the failing component | Problems found by the workflow, not by a customer |
| Daily at 9 am | Compile yesterday’s leads, unanswered conversations and stale deals into one message to the manager | A daily operations brief with no one compiling it |
Choosing the platform
| Platform | Choose it when | Trade-off |
|---|---|---|
| Zapier | Non-technical team, standard apps, few steps, fastest to build | Cost per task grows quickly at volume; limited branching |
| Make | Visual scenarios with routers, iterators and error handling at a lower price | Steeper learning curve for your team to edit |
| n8n | You want self-hosting, unlimited runs, code steps and control over data location | Needs a server and someone to update it, which I can provide |
| Google Apps Script | Everything lives in Sheets, Gmail, Drive and Calendar | Custom code; harder for a non-developer to change |
I pick with you based on the tools involved, run volume, who will maintain it and where the data may live. Every account is opened in your name.
What the service includes
- Process capture: I watch the task done by hand and write down every step, exception and “sometimes we also”.
- One-page architecture: trigger, steps, branches, failure path, alerts. Approved before build.
- Build in a test workspace with your real tools and sample data.
- Error handling: retries for transient failures, parking of unexpected data, alerts to a named person.
- Run log and monitoring, including a heartbeat for scheduled workflows.
- Documentation in plain language: what it does, which credentials it uses, how to pause it, what to check when it fails.
- Handover training and an optional maintenance plan for API changes and new steps.
How I scope a workflow
- Count it. How often, how long, by whom. A task done twice a month rarely justifies automation; a task done ten times a day always does.
- Find the record. Which tool owns the data the workflow moves. If two tools claim it, that is resolved before building.
- Draw it on one page and walk it through with the person who does the task today; they know the exceptions.
- Build and test with duplicates, empty fields, Nepali text, large attachments and two events at once.
- Shadow run for a week alongside the manual process.
- Go live, monitor, document.
What to decide before automating a workflow
- Whether the manual process is actually consistent. If three people do it three ways, we agree one way first; automation encodes whatever it is given.
- Who is alerted when it fails, and who can pause it.
- Which data may pass through a third-party platform, and whether self-hosting (n8n) is required.
- Whether a person should approve any step, at least at first.
- What the workflow must not do: send to customers, delete records, move money, unless explicitly designed and tested for it.
What breaks, and how I prevent it
- Expired credentials (OAuth tokens, API keys rotated): dedicated service accounts and an alert on the first auth failure.
- Rate limits and timeouts: batching, backoff and queueing rather than hammering an API.
- Schema changes (a renamed sheet column, a new form field): validation at the top of every run; unknown shapes are parked, not dropped.
- Duplicate triggers (a webhook fired twice): idempotency keys so the same event is processed once.
- Silent failures: run logs, heartbeats and a weekly summary of failed runs.
- Tribal knowledge: documentation that lets another developer take over.
Pricing pointer
A single workflow with a handful of steps sits at the low end of the setup range, which is quoted on request; a set of connected workflows across five or more tools sits at the top. Maintenance is quoted on request, and platform subscriptions or self-hosting costs are yours. Workflows that need a model to make a judgement are scoped under AI automation services; ones that need a screen for your team become internal tools.
Next step
Describe one task you do by hand, including how often, and I will reply with a one-page flow and a price. Everything else I automate is listed on the automation hub.
