Workflow automation and internal tools
The repetitive work between your tools, automated with code you own: CRMs, email, spreadsheets, WhatsApp and AI steps, with logs and retries.
Automations and small internal tools that remove manual steps: n8n workflows or custom services that move data between your CRM, spreadsheets, email and messaging, with AI steps (classifying, extracting, drafting) where they help and a person approving where they should. I have written on this site about running a self-hosted WhatsApp gateway with n8n.
- 01
Automations in n8n or custom code, running in your cloud or on your server
- 02
An internal admin screen where people review or approve
- 03
AI steps with prompts, test examples and cost limits
- 04
Logs, retries and alerts for every workflow
- 05
A one-page runbook per workflow
Who this is for
- Operations teams copying data between tools every day
- Businesses that talk to customers on WhatsApp and want it connected to their systems
- Teams that outgrew no-code automations, or their monthly bills
Who this is not for
- Bulk or unsolicited messaging
- Automations that remove a human check the business or the law requires
What this service is
Workflow automation and internal tools is removing the repetitive manual steps between your tools with code you own: moving data between your CRM, spreadsheets, email and messaging, with AI steps for classifying, extracting or drafting where they help, and a person approving where they should. I build in n8n, an open-source workflow tool you can host yourself, or as custom services when volume, tests or complex logic demand it. It is for operations teams copying data between tools every day, and teams that outgrew no-code automations or their monthly bills.
Scope
One workflow at a time. I start by shadowing the manual process, timing it and noting where mistakes happen. The first workflow is the one that saves the most time at the lowest risk; it is quoted on its own, and the next one is chosen once the first has run in production for a while.
What is built and checked
Each workflow runs in your cloud account or on a server you own, so the data stays with you and there is no per-task pricing. Every step logs what it did, retries when a service is briefly down, and alerts the right person when it keeps failing. Anything unusual goes to a person with the context attached, through a small internal screen for review and approval.
AI steps get a narrow job, a prompt with test examples it is checked against, a confidence threshold and a cost limit. Below the threshold, a person decides. Each workflow ends with a one-page runbook: what it does, what can go wrong and who to call.
The pattern is the one behind Exhibit Social's automated data collection, where scheduled pipelines replaced manual entry, and I've written on this site about pairing a self-hosted WhatsApp gateway with n8n.
What it is not
It is not bulk or unsolicited messaging, and it is not an automation that removes a human check the business or the law requires. For customer messaging on WhatsApp, I build on the official WhatsApp Business Platform; WhatsApp's terms prohibit automated and non-personal use of its regular app, so I don't build business processes on unofficial gateways.
How the engagement runs
Shadow the work
Watch the manual process, time it, and note where mistakes happen.
Pick the first workflow
The one that saves the most time at the lowest risk.
Automate with a person in the loop
Automation does the routine; anything unusual goes to a person with the context attached.
Hand over
Runbooks, alerts to the right people, and a walkthrough.
Technologies I use for this
- n8n
- Node.js
- Python
- Laravel
- WhatsApp Business Platform
- Docker
- Redis
- Anthropic API
- OpenAI API
How this works in your market
How this works in the United Arab Emirates
For teams in the UAE, our working days overlap almost completely, so I can shadow a process live with your team. Automations run in the region you choose, and where a workflow sends personal data to a third-party service, the mapping shows which fields leave and to whom, for your counsel to confirm under the regime that applies.
How this works in the United Kingdom
For UK teams, automations that touch customer data are designed with UK GDPR's data minimization in mind: each step gets only the fields it needs, logs avoid storing personal data where they can, and retention is enforced by the workflow itself. The engagement includes a data processing agreement where I handle personal data.
Questions about Workflow automation and internal tools
n8n, Zapier or custom code?
n8n when the workflow is simple and your team wants to edit it; custom code when it needs volume, tests or complex logic. Often both: n8n orchestrates and code does the heavy steps.
Where do the automations run?
In your cloud account or on a server you own, so the data stays with you and there is no per-task pricing.
Do you use the official WhatsApp API?
For anything a business relies on, yes: the WhatsApp Business Platform. WhatsApp's terms prohibit automated and non-personal use of its regular app, so I don't build business processes on unofficial gateways.
How do AI steps stay reliable?
Each one has a narrow job, examples it is tested against, a confidence threshold, and a person who reviews anything below it.
What happens when an automation fails?
It retries, then alerts a named person with the failed item and the error, and the item can be replayed once the cause is fixed. Nothing disappears silently.
Can our team change the workflows later?
Yes. n8n workflows are visual and editable, and the runbook explains each step. Custom-code steps live in your repository with tests.
Case studies behind this service
- Case study: Exhibit Social · B2B SaaS platform
- influencers indexed
- 10,000+
- Case study: Digital Wardrobe · AI pipeline
- average time per image, upload to result
- 3–8 s
Related writing
- Article: OpenWA: the free WhatsApp API gateway I wish I'd found sooner · 22 June 2026
A hands-on look at a self-hosted, open-source WhatsApp API gateway: setup, architecture, rough edges and who it's for.
- Article: RAG evaluation sets: what goes in and how to score them · 10 October 2026
How to build the set of questions that tells you whether a retrieval-augmented AI feature is getting better or worse, and how to score retrieval and answers separately.
Where this work happens
- Working with teams in United Arab Emirates · Dubai
Digital teams building platforms that must meet the UAE's data-protection law, with nearly the full working day in common.
- Working with teams in United Kingdom · London
Scale-ups and agencies that want a senior engineer for platforms, performance and security, with most of the working morning in common.