Why does a timeline matter?
When businesses consider AI automation, the first two questions are always "How much does it cost?" and "How long does it take?" We have covered cost in detail. This post answers the second question with a concrete, week-by-week breakdown of what happens from kickoff to go-live.
Understanding the timeline serves two purposes: it helps you plan internal resources (who needs to be available and when), and it gives you a realistic benchmark to evaluate proposals from automation providers. If someone promises to deliver a complex multi-system integration in three days, that is a red flag.
Week 1 — Discovery and process mapping
The first week is entirely about understanding the current state before building anything.
What happens:
- Kickoff session (60-90 minutes) — The automation team meets with the process owner and key stakeholders. The goal is to understand the end-to-end workflow as it exists today: triggers, steps, decisions, systems involved, and edge cases.
- Process documentation — The workflow is mapped visually, showing every step from trigger to completion. Each decision point is documented with the rules that govern it.
- System audit — Every software tool the workflow touches is reviewed for API availability, authentication requirements, rate limits, and data formats.
- Edge case collection — The team specifically asks: "What are the exceptions? What happens when something goes wrong? What do you do when the normal process does not apply?" These edge cases are often where 80% of the implementation complexity lives.
What you deliver:
- Availability for the kickoff session.
- Access to the relevant team members who actually perform the process daily.
- Honest descriptions of how the process really works, not how the manual says it should work.
What you receive:
- A process map documenting the current workflow.
- An architecture proposal showing how the automated version will work.
- A confirmed scope with clear boundaries of what is included and excluded.
Week 2 — Architecture and build
With the process mapped and architecture approved, the build begins.
What happens:
- Platform setup — The orchestration platform (typically n8n or Make) is configured in your own cloud account.
- API connections — Authenticated connections are established to each system the workflow touches: CRM, email, calendar, accounting, communication channels.
- Core workflow build — The main automation pipeline is constructed: triggers configured, data transformations built, AI processing nodes integrated, output actions connected.
- AI prompt engineering — If the workflow includes AI decision-making (classification, scoring, drafting), the system prompts are crafted, tested, and refined for accuracy and consistency.
What you deliver:
- API credentials and access to the systems being integrated.
- Feedback on the architecture proposal if any business rules need adjustment.
What you receive:
- A working prototype of the core workflow running in a development environment.
- A mid-build review session to confirm the automation is behaving as expected.
Week 3 — Testing and refinement
The prototype is stressed with real-world scenarios to identify failures before they reach customers.
What happens:
- Functional testing — Every path through the workflow is tested with realistic data: normal cases, edge cases, error cases, and boundary conditions.
- Load testing — If the workflow handles high volumes, it is tested at expected peak loads to verify performance.
- Error handling — Failure scenarios are deliberately triggered: API timeouts, malformed data, missing fields, rate limit hits. Each failure must be handled gracefully with appropriate fallbacks.
- User acceptance testing — Your team runs the automation with their own test cases and provides feedback on accuracy, speed, and output quality.
- Refinement — Based on testing results, the automation is adjusted: prompts refined, edge cases handled, error messages improved.
What you deliver:
- Test scenarios from your real operational experience.
- Feedback on output quality and accuracy during user acceptance testing.
What you receive:
- A fully tested automation ready for production deployment.
- Documentation of all test results and known limitations.
Week 4 — Go-live and stabilisation
The automation goes live with real data and real users, under close monitoring.
What happens:
- Production deployment — The tested workflow is activated in the live environment.
- Parallel running (optional) — For high-stakes workflows, the automation runs alongside the manual process for the first few days. Results are compared to verify accuracy before the manual process is retired.
- Active monitoring — The automation partner monitors error rates, processing times, and output quality in real time during the first week of production.
- Edge case resolution — Real-world data inevitably surfaces edge cases that testing did not anticipate. These are addressed as they emerge.
- Handover documentation — Complete documentation is delivered: how the automation works, how to modify business rules, how to troubleshoot common issues, and who to contact for support.
What you deliver:
- Prompt reporting of any issues or unexpected behaviour during the first week.
What you receive:
- A fully operational, documented automation running in production.
- A monitoring dashboard showing key health metrics.
- A support agreement covering post-launch maintenance and adjustments.
What happens after week four?
The automation runs autonomously. Your team monitors the health dashboard periodically and reports issues if they arise. Business rules change over time — new CRM fields, updated pricing, revised qualification criteria — and the automation is updated accordingly, either by your team (with documentation) or by the automation partner under an ongoing support agreement.
This four-week blueprint is how we deliver every project. Explore our Workflow Automation service to start the process.