Workflow Engineering
We build the automations that should already exist, on the platform that fits, with git behind every flow the platform allows. Our own estate runs 14 production workflows on self-hosted n8n for EUR 20.49 a month of infrastructure. When a flow breaks, monitoring says so before your customer does.
n8n / Power Automate / Make / Zapier
git + dev / staging / prod
an SLO per flow
14 flows in production
Runs on your stack
All four numbers are from ITSailor's own estate: the workflow exports ship in the delivery kit and all 14 are committed to our git repo, the 21-second figure is our tracer order of 2026-07-06, and the EUR 20.49 is the invoice for the shared Hetzner box that also runs the rest of the platform. They describe our estate, not a promise about yours.
Automation fails in four familiar ways.
The zap nobody owns
Built by a citizen developer who left in March. It silently stopped in May. A customer found out in June. No owner, no monitoring, no version history to roll back to.
The copy-paste department
Sales pastes into Finance, HR pastes into IT. The integration is a person, and the person is on leave. The highest-volume manual loop is invisible until you count what it costs.
The task-bill cliff
A Zapier flow priced per task looks free at 300 runs a month and becomes a four-figure line at 30,000. Nobody re-platformed, because nobody was watching the executions curve.
The un-auditable flow
A Power Automate flow holding production credentials, no environment separation, no DLP policy, and no export anyone can read. It works, until the day it needs to be explained to an auditor.
One real order, the platform maths, and the kit.
Three views of the practice: a real order workflow from our own estate played back node by node, an honest cost simulator across the four platforms, and the delivery kit folder by folder, runbooks included and inspectable before you sign.
Three ways to buy automation here.
This service: the plumbing
Flows between systems you already run, with or without a model in a step. We build and instrument the automation; you own it in your git.
HOIST: a finished product
A Tier-0 support agent that deflects tickets. If your need is "answer employees automatically", buy that, do not commission it.
OpenCustom AI Integration: the model layer
Model choice, evals and prompts inside your own app. When the AI is the product, not a step in a workflow, that is the engagement.
OpenFour artifacts, each one yours to keep.
A ranked automation backlog
Every candidate scored by volume and cost per loop, so the first wave is the highest-return work, not the loudest request. The backlog is yours whether or not you build with us.
Flows in your git, with promotion history
dev / staging / prod for the platforms that support it (git-native on n8n, ALM on Power Platform; export-based on Zapier and Make, and we say so). Every change is a commit you can read.
An SLO sheet per flow
Volume, success-rate and latency targets, an alert channel and a named owner column, from the observability/slo-sheet.md template in the kit. Monitoring says a flow broke before your customer does.
The transfer-out runbook, from week one
RUNBOOK-10-transfer-out exists before flow one ships: the per-platform export reality, credentials, and the SLO sheet, so leaving is a handover, not a hostage negotiation.
In scope
- Process mapping + automation backlog
- Platform selection (n8n / Power Automate / Make / Zapier) + licence procurement
- Build with version control + dev / staging / prod environments
- AI-assisted steps where high-value
- Power Platform CoE (where in scope)
- Monitoring dashboard + an SLO per flow
- Optional operate-tier handover
Deliberately out of scope
- Custom RPA (UI-scraping bots) - we recommend API-first instead
- Replacement of full ERP / CRM workflow modules (case-by-case)
- Citizen-developer training programme (separate Adoption engagement)
- Off-hours emergency response without a retainer (case-by-case)
Two real n8n workflows and the broken-flow triage runbook.
ZIP · 2 n8n workflow exports + a triage runbook
Two sanitised exports from our own production estate (n8n.itsailor.io): node names and structure verbatim, node code summarised. Paired with RUNBOOK-06, the broken-flow triage procedure we run ourselves. Yours to read before any contact.
- csp_order_provisioning.sanitised.json: the Stripe order to Pax8 tenant flow, node by node.
- pax8_catalog_sync.sanitised.json: the catalogue-sync flow that keeps our marketplace rates live.
- RUNBOOK-06-broken-flow-triage: trigger, triage, escalation, and the honest response window.
- The flow naming convention we use across the estate.
Why this is real, not a demo
We run our own estate
14 production workflows on self-hosted n8n at n8n.itsailor.io, on a EUR 20.49/mo Hetzner box. The Stripe-to-Pax8 path provisioned a real tracer order in 21 seconds.
Git-native, not a black box
n8n and Power Platform flows are version-controlled (git / ALM). Zapier and Make are export-based, not natively versioned, and we say so before you build on them.
Sanitised, not fabricated
The exports keep every node name and the connection graph verbatim; only the node code is summarised. What you download is the shape of a flow we actually operate.
Build the first wave. Then decide who keeps it run.
Each tier carries its delivery arc inside it: read-only process mapping first, the first build wave next, then an operated estate or a governed programme. The scope sheet gets signed during discovery; the fee is fixed per tier.
Essential
Your first 3-5 automations, built and handed back.
- Process-mapping workshop on your top manual loops
- 3-5 automations built on the platform that fits each one
- Source-control integration where the platform allows
- Basic monitoring + error alerting
- The delivery kit + 30-day post-deploy support
Outcome
Working flows in your git, with the runbooks to keep them alive.
Operate
An operated estate with named owners and an SLO per flow.
- Everything in Essential
- Up to 20 flows operated, next-business-day triage target
- AI-assisted steps on your top high-volume flows
- A per-flow SLO dashboard + monthly flow-health review
- The founder as your named contact (business hours, not a 24/7 desk)
Outcome
Flows that are watched, so a break is a ticket, not a customer complaint.
Sovereign
Governed automation for a wider estate.
- Everything in Operate
- Power Platform CoE baseline + ongoing governance
- Citizen-developer guardrails with an approved-connector list
- Execution logs and change history, exportable for your auditors
- Up to 50 flows + migration of legacy ad-hoc automations
Outcome
A governed estate an auditor can walk, with the citizen-dev sprawl contained.
Honest answers to the questions buyers actually ask.
How is this different from a one-off Power Automate engagement?
+
We instrument every flow, version it, and assign a named owner. The optional operate tier keeps them alive - a flow that breaks gets triaged the next business day, not discovered months later by a customer complaint.
Do you cover Power Platform governance (CoE)?
+
Yes. Center of Excellence baseline, environment strategy, DLP policies, and ALM patterns are included when Power Platform is in scope, deployed from the Microsoft CoE Starter Kit with a documented baseline checklist.
n8n self-hosted or cloud?
+
Self-hosted by default for sovereignty and cost. n8n Cloud only when the operational overhead of self-hosting is not justified by the use case. We run n8n in production ourselves, so we know where the trade-offs are.
Who actually does the work?
+
The founder, directly. One engineer, accountable end to end, who also holds a full-time engineering role, so scope is set against real hours, not agency capacity. No account manager, no offshore hand-off. Response on the operate tier is next business day, business hours; this is not a 24/7 desk, and the tiers say so.
When should we buy HOIST instead of commissioning this?
+
If the need is "deflect and answer support tickets automatically", HOIST is a finished Tier-0 product and you should buy it, not pay to have it rebuilt. This service is the plumbing between systems. If you are unsure, the discovery call sorts it in ten minutes.
How do we leave, and what about our Zapier flows?
+
The transfer-out runbook is written on day one. n8n and Power Platform flows are git-native and ALM-exportable, so they leave cleanly. Zapier and Make are export-based rather than version-controlled, so the honest answer is you leave with an export and the runbook that documents each flow, not a full git history the platform never kept. We tell you that before you build on them.
n8n or Zapier at volume?
+
Zapier is priced per task and is the fast path at low volume. Self-hosted n8n is a flat infrastructure cost, so somewhere between a few thousand and a few tens of thousands of executions a month the lines cross and keep diverging. The platform-maths tab above lets you find your own crossover; we run our own estate on self-hosted n8n for that reason.
The plumbing, the product, and the model layer.
HOIST
Deflect tickets instead of routing them: the finished Tier-0 support agent, when the need is a product not plumbing.
OpenCustom AI Integration
The model layer itself: evals, prompts and model choice inside your own app, when the AI is the product.
OpenFinOps & Cost Management
The same git-and-ownership discipline applied to cloud and SaaS spend, with a ranked savings backlog.
OpenPick the platform once, on the numbers, before anything is built.
n8n, Power Automate, Make and Zapier price the same workflow very differently at your volume. The workshop runs that comparison against your real run count.