Map
Current trigger, tools, fields, owners, status points, and failure cases.
Make, n8n, Airtable, CRM, APIs
I help owner-led teams debug, simplify, and rebuild brittle automations around leads, bookings, invoices, PDFs, CRM updates, and human review.
Built by Duncan Anderson, an AI engineer and data scientist who ships review-first automation systems for real operating work.
Rescue map
Form, email, call, webhook, spreadsheet row
Fields, IDs, dedupe key, source record
Owner, approval, exception, missing data
CRM update, draft, message, report, notification
Rule
The first repair should make one workflow trustworthy before it makes the system bigger.
Current trigger, tools, fields, owners, status points, and failure cases.
The smallest reliable repair or rebuild path for the workflow that matters first.
A human approval step where AI or automation should not act silently.
Notes, test cases, and next-step recommendations so the system can be maintained.
Best fit
This is not a broad AI strategy package. It is for a workflow that already exists, already matters, and needs a practical repair path.
A Make, n8n, Zapier, Airtable, or CRM workflow works sometimes but breaks when the data shape changes.
Leads, bookings, invoices, PDFs, or client requests enter one tool and get lost before a person owns the next step.
An AI step is producing useful drafts or summaries, but there is no clean review queue or approval boundary.
A few tools are connected, but nobody trusts the automation enough to rely on it during real work.
Working stack
The offer is intentionally tool-agnostic, but it fits best where the workflow depends on practical glue between apps, data, and people.

Builder, not agency layer
You work with the person mapping the workflow, inspecting the failure points, and building the first reliable version.
Proof
These projects show the kind of work behind the offer: private operating portals, review queues, content and receipt extraction, data pipelines, and production automation.
Sports analytics
Customers can compare a projection with current prices, make their own decision, and return later to see the recorded result.
Read case studyProduction operations systemBusiness monitoring
The recurring checks now happen on a schedule, issues are kept in one review queue, and production changes cannot happen without explicit approval.
Read case studyClient deliveryHealthcare content operations
The team can move approved copy into a structured WordPress draft without rebuilding the article by hand.
Read case studyStart here
Include the tools involved, the trigger, what should happen, what actually happens, and whether you need a repair, rebuild, or implementation partner.