Loading…
Long-running, transactional, with a custom UI a non-technical operator can actually use. Idempotent by default, retried with backoff, audited per attempt.
What you get
Explicit transitions, not implicit chains. You can see exactly what state a run is in.
Every step retried with backoff and an idempotency key — a replay never double-charges or double-emails.
Logs, traces and durations per step — not just per run. You know exactly where it slowed down.
A real admin tool for the human-in-the-loop step — a queue for the reviewer, a dashboard for the manager.
How it runs
The same stages run every engagement. The service decides the scale, not the method.
01Blueprint
We map the workflow as it exists — the Zap, the spreadsheet, the shared inbox. Failure modes named.
02Blueprint
Data model, queue topology, retry policy and idempotency strategy on the page. You sign off.
03Forge
Four to eight weeks. Weekly demo on a real preview URL. Every external call audited from day one.
04Forge
Deploy to your cloud, cut over from the legacy Zap, watch the first week of real traffic together.
Pricing
Single workflow
One workflow with two integrations and an admin surface.
From £18,000
Workflow engine
Queues, retries and dead-letter handling built as a reusable engine.
Example: £30,000
Multi-tenant & custom UI
Multi-tenant runs, custom step UI, a year of headroom for new flows.
Example: £50,000
Proof in production
AI Automation · Case study
Categorisation, summarisation, and triage automated across an operations team's daily workload. Graceful degradation when the model is…
how-to11 min
Should you fix a messy process first, or automate it as-is? The honest answer, a 5-minute consistency audit,…
Read guideexplainer11 min
Signs you've over-automated a process: how to tell when automation costs more than it saves, and the honest…
Read guidehow-to13 min
When a small business actually needs a custom internal tool instead of a spreadsheet or email chain — the…
Read guideQuestions
When steps get conditional, when one failure needs surgical retry rather than the whole run, when an auditor will want a replay log. Beyond simple two-step flows, off-the-shelf tools refuse and you write more glue around them than the workflow itself.
Background jobs, not serverless timeouts. A Postgres-backed queue runs for hours if needed. Heartbeats, progress updates visible in the admin, hard cancel from the UI. Survives a deploy mid-run.
Yes. Workflow state is persisted at every checkpoint — pause for human approval, resume from the same step. The state machine is in the database, not in memory, so a restart doesn't lose progress.
Retry with exponential backoff for transient failures. Idempotency keys on outbound calls so a replay doesn't double-charge. Dead-letter queue for steps that exhaust retries, surfaced in the admin.
Your call — take it in-house with plain TypeScript and a runbook, or System membership (£500/month) for hosting, security patches, upgrades, backups and four hours of changes a month. Same engineer who built it.
Talk to us
Scoped and priced before anything is built.