Automation Health Check.

A diagnostic, not a sales document.

Most automation stacks accumulate three problems that the team running them does not see, because the workflows are still firing and the dashboard still says green. A Health Check finds them.

The patterns we look for

Ghost automations. Workflows still running months, sometimes years, after the person who built them left. They duplicate work that has since shifted elsewhere or feed data into systems nobody uses any more. They are quietly expensive: platform credits, API calls against tools the business is paying for, and stored personal data that under GDPR the business is responsible for retaining or deleting whether anyone remembers the workflow exists or not.

Silent API drift. Vendors change contracts. Xero rotated its invoice-scope permissions in March 2026, requiring granular OAuth scopes for new apps and breaking older ones. Gmail tightened its OAuth posture the year before. Any automation built against an old contract is now failing or returning unexpected data, and on most setups nobody notices because the team that built it never wired in failure alerts. A workflow that ran cleanly twelve months ago can be dropping every third record today while the dashboard reports “completed”.

Cost creep no-one signed off. A workflow built to handle 500 runs a month is now firing 12,000 because an upstream filter was removed in a different change, or because someone plugged a new lead source into the same trigger without telling the automation owner. The platform bill quietly doubles, and on the tools commonly in use (n8n cloud, Make, Zapier, Power Automate) the cost shift does not surface unless you go looking for it.

How the work runs

Two to three weeks, three phases.

Week one is access and inventory. We need read access to your automation platform, the logs, and ideally the credentials store. We catalogue every live workflow, every disabled-but-not-deleted workflow, every scheduled job, and every webhook endpoint. Where the automation handles personal data (customer records, payroll, invoice contents, anything that identifies a real person) we flag the processor relationship, the data retention behaviour, and any cross-border flow that GDPR Article 28 would require you to document. That part has become material since Anthropic launched Claude for Small Business in May 2026 and packaged AI suites started routing client financial data through new third-party processors before practice owners had read the data processing agreement.

Week two is analysis. Every workflow gets a health rating (healthy, attention, failing) with evidence from execution logs and error rates. Findings are sorted by impact: what is costing money, what creates regulatory exposure, what wastes time, what is at risk of breaking under a known upcoming vendor change. Quick fixes (a credential rotation, a filter restored, a webhook re-registered) are usually resolved during this phase rather than waiting for the report. We log what we changed.

Week three is the report and walkthrough. The deliverable is a ranked list, not a sales document. Every finding has a fix, an estimate of time to remediate, and a cost-or-time-saved figure where one can be honestly calculated. We walk through it on a video call so you can interrogate the reasoning rather than read a PDF cold. You can act on the report yourself, hand the brief to your existing partner, or commission us to deliver the larger fixes through an Automation Build or an ongoing Managed Automation engagement.

Where this fits

The Health Check reviews what you already have. The AI and Automation Audit is the bigger-picture cousin that maps where automation and AI could help you next. Different starting point, both end with a prioritised action list.

The methodology aligns to internal audit and process optimisation discipline. The questions are the same as those an auditor asks of any operational process: what is the design intent, what is happening in practice, where is the gap, and what is that gap costing in money, time, or risk. The automation layer is the part of the business those questions have not historically been asked about.

The reason an independent health check matters is the same reason an independent stock-take of any kind matters. Anthropic will not tell you to switch off Claude for Small Business. Microsoft will not tell you Copilot is the wrong fit. The platform you built on will not tell you you have outgrown it. We are independent AI and automation specialists. No vendor pays us for the recommendation, so the report says what we actually find.