# Pick the right automation pattern for the work: a six-pattern map Source: https://getstartedwithai.net/tutorials/ai-workflow-automation-patterns Site: Get Started With AI Tagline: Build AI automations that actually run your business. Audience: solo professionals, freelancers, consultants and AI implementers in training Contact: hello@getstartedwithai.co.uk This Markdown companion is generated from the prerendered page for AI crawlers. # Pick the right automation pattern for the work: a six-pattern map Naming the pattern before you open the builder saves the hour you would otherwise lose backtracking from a tool that does not fit. It also gives you a price you can quote a client without rebuilding the calculation each time. Six patterns cover almost everything: linear, branch, aggregator, enricher, two-way sync and agent. This guide names each one, gives a real example, lists the right tool, and flags the common mistake that breaks each pattern. After about ten implementations you spot the pattern in the first five minutes of a discovery call. ## Pattern 1: Trigger and act (linear) A single event fires a fixed sequence of actions. About 60 percent of solo automation work fits this shape. - Example: form submission becomes CRM record, then Slack ping, then calendar invite. - Right tool: Zapier. Linear builder, fastest setup. - Common mistake: trying to add branching here. If you need branching, you are in Pattern 2. ## Pattern 2: Branching (router) A single event splits down two or more paths based on conditions. - Example: new lead becomes hot path (email plus Slack alert) if score above 3, cold path (auto-reply only) if below. - Right tool: Make (visual canvas with native Router) or Zapier Paths. - Common mistake: too many paths. Cap at three or four. Past five, the canvas becomes unreadable and the maintenance breaks. ## Pattern 3: Aggregator (many to one) Multiple events feed into a single batched output, usually on a schedule. - Example: weekly digest of all CRM activity, sent on Friday at 16:00. - Right tool: Make (native iterators and aggregators) or n8n. - Common mistake: real-time triggering instead of scheduled. Aggregators run on a clock, not on each event. ## Pattern 4: Enricher (input plus AI) A simple input is processed by an LLM and the enriched output is written back somewhere. - Example: meeting transcript becomes structured action items, written into Notion. - Right tool: Zapier with AI Steps, or Make with the OpenAI or Anthropic modules. - Common mistake: skipping the structured-output prompt. Always tell the LLM exactly what fields to return, in what format. ## Pattern 5: Two-way sync Two systems must mirror data in both directions. The hardest pattern. Only build it when nothing else works. - Example: HubSpot deal stage syncs to Notion project status, and Notion status changes update HubSpot. - Right tool: n8n (granular control) or a dedicated tool such as Whalesync. - Common mistake: infinite loops. A change in System A triggers a write to B, which triggers a write back to A. Always add a last-modified-by guard. ## Pattern 6: Agent (multi-turn dialogue) Conversational, multi-turn, takes actions across several steps inside one session. - Example: a website chat that asks two qualifying questions, books a calendar slot, creates a CRM lead. - Right tool: Zapier Agents, or a dedicated platform (Vapi or Retell for voice). - Common mistake: using an agent when a Zap with AI Steps would do. Agents are 5 to 10 times more expensive and slower to debug. ## Walk down the list to pick the pattern Start at the top, stop at the first match. - One trigger, one fixed sequence of actions: Pattern 1. - One trigger, multiple paths based on conditions: Pattern 2. - Many events, batched into one output: Pattern 3. - Input that needs LLM judgement before output: Pattern 4. - Two systems where edits in either side must mirror: Pattern 5. - A real conversation with someone who expects multi-turn answers: Pattern 6. ## What you ship A pattern-led discovery script you can run on the next prospect call. Open every implementation engagement by naming the pattern. The client may not know the language but they recognise the shape. Indicative ranges by pattern: 1 sells for £900 to £2,500; 2 for £1,500 to £4,500; 3 for £1,800 to £5,000; 4 with AI for £2,500 to £6,500; 5 for £4,000 to £15,000 (rare); 6 for £5,000 to £25,000 depending on integration depth. ## Common mistakes - Treating every automation as a unique snowflake. Six patterns cover almost everything solo and small-firm clients ask for. - Picking the tool before naming the pattern. Reverse the order. The pattern picks the tool. - Building Pattern 6 when Pattern 4 fits. Agents are seductive and expensive. - Skipping the pattern check on a build that has been running quietly. Most broken automations are actually pattern mismatches. ## What to build next Pair this with the WIRE Method one-pager and the first Zapier or Make tutorial for your chosen pattern. The pattern names the shape, the WIRE Method names the work to put inside it. ## Related builds More tutorials in Automation. Each one runs in 60 to 90 minutes and ships a finished asset. ### Capture every lead automatically: a 30-minute Zapier build 30 to 60 minutes. Solo Pro. [Open tutorial](/tutorials/first-zapier-automation-30-min) ### Decide between native AI Steps and a Custom GPT in 10 minutes 60 to 90 minutes. Solo Pro. [Open tutorial](/tutorials/zapier-ai-features-vs-custom-gpt) ### Run automations Zapier cannot: the first Make scenario 60 to 90 minutes. Solo Pro. [Open tutorial](/tutorials/first-make-scenario-visual) ### Self-host the automation stack: when n8n is worth the effort 60 to 90 minutes. Solo Pro / Implementer. [Open tutorial](/tutorials/n8n-self-hosted-when-worth-it) ## Questions people ask before they start ### Where does AI fit in? AI is most useful inside Pattern 4 (the enricher). It also lives inside Patterns 2 (for routing decisions) and 6 (for agent dialogue). Patterns 1, 3 and 5 rarely need AI. ### Are these patterns universal? For solo professional and small firm work, yes. Enterprise automation has more patterns (orchestration, saga, transactional outbox). Below 200 staff, the six covers it. ### Can a build mix patterns? Yes, that is the norm rather than the exception. A real client onboarding automation usually combines Pattern 1 (linear sequence) with Pattern 2 (branch by client type) and Pattern 4 (AI to draft the welcome email). ### How long does pattern recognition take? About 10 implementations. After that, you see the pattern in the first 5 minutes of the discovery call. ### Which pattern should I learn first? Pattern 1. Build three linear Zaps before touching anything else. Most patterns share Pattern 1 as a sub-shape.