Sefism Automation for Clients Cheat Sheet

What this cheat sheet is

The Sefism Automation for Clients cheat sheet is a quick-revision book for people building workflow automations that businesses pay for. Each concept is a short card: what it is, why it matters and, for most, the mistake people make with it. It is the companion to the Automation for Clients roadmap and its chapters follow the roadmap's stages, from the HTTP and JSON basics every tool hides to scoping, pricing and handing over the work.

Three editions: n8n, Make and Zapier

The sheet comes in three editions that match the tool you chose on the roadmap. Seven chapters are shared; one chapter is specific to your tool, and the error handling, AI and security chapters include tool-specific cards where the tools genuinely differ.

• n8n edition (72 concepts): nodes and items, expressions such as $json, credentials and the encryption key, the Code node, self-hosting versus n8n Cloud, the Sustainable Use License, and how executions are counted. Its error chapter covers Retry On Fail and error workflows, and its security chapter adds hardening a self-hosted instance.
• Make edition (71 concepts): scenarios, modules and bundles, mapping pills and functions such as ifempty and formatDate, routers and filters, iterators and aggregators, data stores, connections and the HTTP app, and how credits are billed. Its error chapter covers error handler routes.
• Zapier edition (71 concepts): Zaps, field mapping and line items, Filters and Paths, Formatter, Code by Zapier and Webhooks by Zapier, Tables and Forms, and how tasks are counted. Its error chapter covers Autoreplay and error handler paths.

Where the sheet quotes pricing units or plan limits, it states the date it read them, so check the tool's own pricing page before you quote a client.

Chapter by chapter

1. How automation works

Trigger, action, JSON, data mapping, HTTP requests and methods, REST APIs, webhooks, API keys, OAuth 2.0 and rate limits (including the 429 Too Many Requests response and the Retry-After header). These ten cards are what you fall back on when a connector is missing or a step fails with a status code you do not recognise.

2. Your tool in depth

The tool chapter described above: eight cards on how your chosen tool builds, runs and bills workflows, so you can estimate what a workflow costs a client each month before you quote.

3. Shaping data and logic

Lists and loops, merging data on a shared key, dates and time zones (convert once, near the end, into a named zone), text cleaning and regex, search before create, deduplication on a genuinely unique key, pagination and reusable sub-workflows.

4. Errors, retries and monitoring

Types of failure (temporary, permanent and silent), retries and backoff, your tool's error handling, idempotency, alerts that include the workflow, error, record and a run link, run history and logs, test data and staging, and versioning workflows so a bad change can be rolled back.

5. AI steps in workflows

The LLM step, choosing a model by task, prompts for workflows, structured output and validation, AI agents in each tool, human in the loop, tokens and AI cost, hallucination and prompt injection. The pitfalls are concrete: using AI for maths or lookups, and giving an agent tools that let injected text in an email act on the client's systems.

6. Common client workflows

Lead capture, CRM sync, email parsing, document parsing, invoicing automation, scheduled reports, team notifications, Sheets sync and onboarding workflows. Each card names what makes that build reliable, such as one-direction syncs from a clear source of truth.

7. Data privacy and security

Least privilege, client-owned accounts, secrets handling, personal data, GDPR basics, data minimisation and retention (including in AI prompts, logs and run history), webhook signature verification and access review and offboarding.

8. Finding clients and selling the work

The largest chapter, with eleven cards: niche and offer, finding clients, the discovery call, the process map, scope and proposal, change requests, value pricing, maintenance retainers, the handover document, contracts and getting paid, and case studies shared with permission and without client data.

Rules of thumb

The sheet ends with ten rules worth memorising, among them:

• Map the process on paper before you open the tool; a messy process automated is still messy.
• Search before you create, and key every record on a unique ID, so duplicates never start.
• Retry temporary errors with backoff, alert on permanent ones, and never retry a step that already succeeded.
• Every workflow you hand over should tell someone when it fails.
• Keep AI steps narrow, request structured output, validate it and put a person before anything irreversible.
• Build in accounts the client owns, with the smallest scopes that work.

How to use it with the roadmap

• After each stage, read the matching chapter and explain each card aloud without looking.
• While building, keep chapters 3 and 4 open: most client complaints come from duplicates, time zones and silent failures.
• Before a discovery call or proposal, revise chapter 8.
• Before every handover, run through chapters 4 and 7 as a checklist: alerts, logs, credentials, scopes and offboarding.

How it works on Sefism

The public cheat sheet page lists every chapter, with the first chapters open to read. Keeping the sheet in your member area uses credits, and following the Automation for Clients roadmap unlocks it with the roadmap. Everything on Sefism is free to learn from, and new members get starting credits when they join. If a client situation does not fit any card, send Tauseef a query.

Related on Sefism

• Roadmaps: Automation for Clients, Freelancing, AI agents and MCP
• Cheat sheets: Freelancing, Model Context Protocol
• FYP idea: Lead Capture to CRM Automation with AI Qualification
• Resources: n8n, Make, Zapier, Webhook.site

FAQ

What is the difference between a webhook and polling? A webhook is an HTTP request an app sends to your URL the moment an event happens, which powers instant triggers. Polling means your tool checks every few minutes for anything new.

What does idempotency mean in automation? Designing a step so that running it twice has the same effect as running it once, usually through a unique source key and update-or-create actions. It is what stops a retry from creating a duplicate invoice.

How do n8n, Make and Zapier bill? n8n Cloud plans are priced by executions (one full run of a workflow), Make by credits and Zapier by tasks (successful action steps). The tool chapter in each edition explains what counts, so you can estimate a client's monthly cost.