01
Production error handling
If a failed run can lose money, delay a client or create duplicate records, the workflow needs retries, alerts and an owner.
Production n8n consulting, not drag-and-drop experiments.
n8n consultant in London, building AI workflow automation for UK businesses.
WiseSolutions designs, builds and operates n8n workflows for teams that need automation to survive real business mess: retries, credentials, queues, logs, branching and human review.
Section 01 · Plain terms
n8n is the control layer between the tools your team already uses.
In plain terms, n8n is a workflow automation platform. It listens for an event, pulls data from one system, applies rules, calls other systems and records what happened. A trigger can be a form submission, an email, a webhook, a schedule, a database change or a manual button.
The simple version is easy to understand. A new invoice lands in an inbox. The workflow reads the email, extracts the supplier, checks the Drive folder, updates a spreadsheet, tells the right person and logs the result. The hard version is what happens when the invoice is missing, the supplier name changed, the API times out, or the same email arrives twice.
That is where a consultant earns the fee. WiseSolutions uses n8n as an operating layer for AI workflow automation, routing and internal process work. We build for serious teams that need the workflow to keep running after the first demo.
Good n8n work is not just node wiring. It includes naming, credentials, test data, retries, alerting, error routes, access control, version notes and a runbook that a real team can read. If those pieces are absent, the workflow becomes another fragile thing everyone is scared to touch.
Where n8n fits
- Trigger: something happens in email, CRM, form, database or webhook.
- Decide: the workflow classifies, validates, routes or asks for approval.
- Act: it writes to the right systems and notifies the right people.
- Record: it logs the outcome so the business knows what happened.
The value is not the diagram. The value is a workflow that survives bad data and tired humans.
Section 02 · When to hire
DIY works until the workflow becomes part of production.
02
Branching that people can debug
A workflow with many IF and Switch paths can look fine in the editor and still be impossible for staff to reason about later.
03
Postgres, Redis and queue workers
Self-hosted production setups need a real database, queue mode and worker processes so larger jobs do not block the whole instance.
04
Self-hosting decisions
Backups, credentials, domains, SSL, upgrades and restore plans matter when automation is touching client data.
05
Silent failures
The worst workflow failure is the one nobody sees. We design status checks so missing runs are noticed, not discovered by a client.
DIY is fine for a personal notification or a one-step handoff. Hire WiseSolutions when the automation has business risk, private data, multiple branches or enough volume that manual checking becomes its own job.
Section 03 · Production model
How we work with n8n in production.
Layer 01
Managed UK infrastructure
For serious work we use self-hosted n8n with Postgres as the database, Redis for queue mode and separate workers for long-running jobs.
Layer 02
Observability
Every important workflow has logs, alerts, execution status, error routes and a place where the team can see what happened without reading raw node output.
Layer 03
Version control
We keep workflow exports, notes and configuration changes under control so a working automation can be restored after a bad edit or platform update.
Layer 04
Human review
AI steps that classify, extract or draft need a review path when confidence is low. We design the handoff instead of pretending every record can be processed blindly.
Section 04 · What we build
Practical workflows where the numbers are visible.
Invoice triage
Inbox to files, audit log and approval route.
Emails arrive, attachments are extracted, suppliers are classified, files move to the right folder and exceptions go to a person.
Chatbot routing
Website and WhatsApp into one backend.
The workflow normalises messages, identifies intent, routes to the right team and keeps a record of each conversation.
Content pipelines
Research, draft, quality check and publish queue.
We use n8n to coordinate research steps, approvals, image tasks, CMS publishing and status reporting.
Lead routing
Form, CRM, notification and follow-up.
New enquiries are scored, enriched, assigned and sent to the right person with the context needed for a fast reply.
Section 05 · Real cases
Production automations, anonymised.
Healthcare services group, UK, anonymised
Invoice triage that replaced a weekly manual finance loop.
A team was spending around four hours a week moving supplier invoices from email to folders and tracking what had been checked. We built an n8n workflow that polls the inbox, classifies invoices, saves attachments, updates the audit log and sends exceptions to a human reviewer.
- Source: shared finance inbox, Drive folders and audit table.
- Controls: retry route, duplicate detection and exception queue.
- Outcome: fewer missed invoices and a visible record for the finance lead.
Professional services firm, UK, anonymised
Multi-channel chatbot routing across website and WhatsApp.
The business had enquiries arriving from a web widget and WhatsApp with no shared routing logic. We built a workflow that normalises both channels, identifies the service line, stores the transcript and sends the handoff to the right person with context.
- Source: website widget, WhatsApp provider and CRM records.
- Controls: consent marker, fallback path and staff notification.
- Outcome: one routing layer instead of two disconnected inboxes.
Content business, UK, anonymised
Content production pipeline with approvals and publish status.
A small team needed a repeatable way to move topics through research, draft, review, image preparation and CMS publishing. n8n became the orchestration layer, with each status visible and failures routed back to the person who could fix them.
- Source: topic database, documents, CMS and image assets.
- Controls: approval gates, run logs and failed-publish alerts.
- Outcome: a calmer publishing rhythm with less manual chasing.
Names redacted out of respect for our clients. We can share detail on a call.
Section 06 · n8n vs alternatives
n8n is not always the right answer. That is why we compare it first.
n8n
Best for production workflows with logic.
Strong when you need code, branching, self-hosting, database access, private credentials, queue workers and serious error handling.
Zapier
Best for quick SaaS handoffs.
Good for simple marketing and operations automations where convenience matters more than control or cost at scale.
Make.com
Best for visual business logic.
Useful for teams that want a strong visual builder and can accept the platform boundaries around hosting and deep operations.
Custom code
Best when the workflow is the product.
Better when latency, security model, UI or scale needs are beyond what a workflow tool should own.
Section 07 · Pricing
Clear starting points, scoped after we see the process.
Proof of concept
from £4,000
- One workflow or process lane.
- Two to four weeks.
- Working demo with real test data.
- Build notes and next-step estimate.
Production build
from £12,000
- Multiple integrations and branches.
- Error handling, alerts and documentation.
- Self-hosted or managed deployment.
- Six to ten weeks for most projects.
Managed support
from £350/month
- Monitoring and monthly review.
- Minor fixes and workflow changes.
- Incident notes for failed runs.
- One named point of contact.
Hosting and third-party usage are quoted separately. We keep the cost model visible before production.
Section 08 · Implementation detail
What we check before calling an n8n workflow production-ready.
The first check is trigger reliability. A scheduled workflow that polls every five minutes needs a different failure model from a webhook that must respond immediately. We document what starts the run, what counts as a duplicate, what happens when the source API is slow and how far back the workflow can safely replay.
The second check is data shape. Most automation bugs come from fields that are empty, renamed, nested differently or formatted by a human who was in a hurry. We validate the fields that matter before the workflow writes to another system. Bad records go to review instead of poisoning the output.
The third check is credentials. Production n8n work should not depend on one staff member's personal login or a token pasted into a note. We prefer named service accounts, scoped credentials, clear ownership and a renewal plan. Any business that handles client data should be strict about this from day one.
The fourth check is change control. A working workflow can break because a node changed, a SaaS API changed, a team renamed a field or someone edited the wrong branch. WiseSolutions keeps workflow exports, notes and recovery paths so the team is not guessing during an incident.
Production checklist
- Trigger contract: what starts the workflow, how duplicates are handled and how missed events are replayed.
- Data validation: required fields, accepted formats, fallback values and review queues.
- Error route: retries, alerts, failed-record storage and the person who owns the fix.
- Credential owner: service account, renewal process and documented permission scope.
- Runbook: plain English notes for support, handover and future edits.
This is why consultancy work costs more than a quick node diagram. The hidden work is what keeps the automation alive.
Handover means someone else can operate it
- Workflow purpose: what business process the automation owns and what it deliberately does not own.
- Inputs and outputs: which fields arrive, which fields are created and which systems are updated.
- Known failure modes: what breaks most often and the first place to look when it happens.
- Escalation path: who gets told when a run fails, when a record is ambiguous or when a source system is down.
- Change notes: how future edits should be tested before they touch the production route.
We write this for the person who will be on call after the exciting part of the build is over.
The goal is simple: a staff member should be able to see the last run, understand the next action, and know who owns the fix without asking the original builder. That is the bar. We also make the commercial trade-off clear. Some workflows should stay small and manual because the volume is not there yet. Others should move into n8n because the same checks happen every day, the same data is copied between systems, and the same people are interrupted to confirm what a well-designed workflow could record automatically. That is where automation pays.
Section 09 · FAQ
Questions we get about n8n consulting.
What does an n8n consultant actually do?
We map the process, build the workflow, connect the systems, add error handling, document the handover and monitor the production run. The work is part automation design, part integration engineering and part operations.
Can you work with our existing n8n instance?
Yes. We can audit an existing instance, fix broken workflows, add queue workers, improve credential handling, document production routes and move fragile workflows into a cleaner structure.
Do you self-host n8n?
Yes. For production work we normally prefer self-hosted n8n with Postgres, Redis and queue workers, running on managed UK infrastructure. That gives better control over cost, logs, credentials and deployment.
When is Zapier or Make.com a better choice?
They are often better for small personal automations, quick marketing handoffs and simple workflows owned by non-technical teams. n8n is better when the workflow has branching, code, credentials, private data or serious error handling.
How long does a proof of concept take?
A focused proof of concept normally takes two to four weeks. A production automation with monitoring, retries, documentation and a review loop usually takes six to ten weeks.
Can you integrate email, CRM, forms and spreadsheets?
Yes. Those are common sources. We also work with webhooks, databases, file storage, team chat, WhatsApp providers, accounting tools and custom APIs.
How do you stop silent failures?
We design explicit error routes, retry logic, alerts, dead-letter queues where needed, and regular status checks. If a workflow fails at 2am, someone should know what failed and which records were affected.
Can you migrate workflows from Zapier or Make.com?
Yes. We rebuild the logic in n8n rather than copying blindly. A migration is a good moment to remove duplicated steps, add logging and make the workflow easier to operate.
Do you provide documentation?
Yes. Each production workflow gets a plain English runbook covering trigger, credentials, expected output, failure modes, alert route and owner. That matters when staff change.
Can n8n handle AI workflow automation?
Yes, if it is designed carefully. We use n8n for routing, approvals, logging, retrieval, transformation and human review around AI steps so the automation is useful in real operations.
What ongoing support do you offer?
Managed support starts from GBP 350 per month. It covers monitoring, minor fixes, monthly review and practical advice when workflows need to change.
[ MODULE · 06 ] · n8n-discovery
Have a workflow that is too important to keep duct-taping?
Tell us what the workflow is supposed to do, where it breaks today and what is on fire when it goes wrong. We will reply within one working day with first questions or a discovery call slot.