I find the recurring work that quietly eats a week — in admin, finance, HR, operations, marketing and sales — and automate it off people's desks. Then I build the operating system that keeps it running, because you cannot automate a process nobody has written down.
Work with me
01 — What I do
Most AI projects fail one step earlier than people think: the process was never written down, so there was nothing to automate. I work across all three layers, which is why the automation survives after I leave the room.
AI implementation
Not a pilot. The version people open on a Tuesday morning without being asked.
Automation
Integrations that run unattended and are documented well enough for someone else to fix.
Operations
The unglamorous half. Without it, every automation is built on sand.
02 — Automation
I do not automate for the sake of it. I go looking for the task that repeats, costs a person their morning and produces nothing anyone reads — then take it off the desk for good.
Operations
Notion and Asana rebuilt as the company operating system, migrated off ClickUp with zero data loss: projects, client portals, wikis, 1:1s and an OKR cycle with KPI coverage for every role. Nobody asks where something lives any more.
Administration
A new hire is productive in ten days instead of four weeks, off written SOPs and 11 reusable internal AI skills anyone on the team can run without asking me first.
Finance
Invoicing and payment tracking through Qonto and Holded, payroll coordination, vendor negotiation, and cost audits that surface what the company is quietly paying for twice.
HR & hiring
Pipeline and ATS in HomeRun, onboarding rebuilt end to end, and the Spain-to-UK employment transition with Deel, Remote, Oyster and Multiplier compared on real numbers rather than sales decks.
Content generation
Research, drafting and publishing chained into one workflow. It goes out every Monday and nobody opens a document to make it happen.
Video
Research, scripting and publishing as a single workflow, built alongside the editors so it survived the handover — plus the AI media stack (Descript, ElevenLabs, Runway) in daily production use.
Marketing
The short-form pipeline feeds a planned publishing calendar instead of last-minute posts. I also designed and delivered a public masterclass on AI in creative operations for 20+ external professionals.
Sales & CRM
A bidirectional HubSpot ↔ Notion sync so deals and delivery share one source of truth, outbound sourcing and sequencing through Sales Navigator, Clay and Lemlist, and a CTR-based success-fee model on the largest account.
03 — Consulting & freelance
A small number of engagements alongside my day job — the kind where the answer is a working system, not a slide deck. Scope and fee agreed up front, in writing, before anything starts.
You are paying for more software than anyone can name, and half of it overlaps. I map what you have, what it costs, what it duplicates and what is missing — then tell you what to buy, what to cancel, and what to connect.
You getA stack map, a keep / cancel / buy list with real costs, and a 90-day rollout order.
A specific workflow, built and shipped: n8n automations, Claude agents, MCP integrations, CRM syncs, reporting that assembles itself. Handed over documented, so it is yours and not mine.
You getWorking automations in your accounts, a runbook, and a walkthrough with whoever owns them next.
The part everyone skips. I train non-technical teams to use AI on their own work — with their files, their clients, their annoying Tuesday task — and leave behind prompts and SOPs they will actually reuse.
You getA live session, a shared skills library, SOPs, and an adoption check-in a month later.
For founders past the point where everything fits in their head. Operating system, OKRs, onboarding, hiring, finance and vendor ops — set up properly, then handed to the person you hire for it.
You getA documented operating system in Notion or Asana, an OKR cycle running, and an ops hire briefed to take over.
04 — The stack
65Vendors chosen,
wired and paid for
This is the vendor list of a company I run operations for. It is also why my advice on what to buy tends to be short and specific.
Claude is the spine — Claude Code and MCP servers connect it to everything below. The rest is media generation the production team uses daily.
Where the hours come back. n8n runs the recurring work, Apify scrapes what no API exposes, Workers host what needed a backend.
I have migrated between most of these on live projects, which is the only way to learn what a migration actually costs.
Outbound as a system rather than a mood: sourcing, enrichment, sequencing, and a CRM that syncs both ways with the work tools.
Invoicing, payroll coordination, contracts and hiring — including a Spain-to-UK employment move with all four major EOR vendors compared on real numbers.
Years inside video production, first as a co-founder and now running its operations — so I know which of these is a real bottleneck and which is a preference someone will defend loudly.
I wrote the company Data Protection & Security Playbook, so this list is a policy as much as a shopping basket.
Vendor choice is rarely the hard part. Knowing which three you need, which two you already own without realising, and what the integration really costs in someone's week — that is what I get hired for.
05 — Built
I do not only advise on this work. I build with the same tools on my own time, which is where most of what I recommend gets tested first.
Solo product · Mac app
Keyboard shortcuts learned the Duolingo way. Designed, built and shipped alone with Claude Code — spaced-repetition engine, a companion Chrome extension, and a landing page that converts.
Tester
An AI operations app. I run it against real operations work and report what breaks — some of that feedback has made it into the product.
Teaching
A public masterclass for 20+ external professionals on putting AI into creative and operational work — designed and delivered end to end. Half of this job is explaining it to people who are not sold yet.
06 — Track record
Current
Camping TV — B2B video & strategy agency for SaaS scaleups, Barcelona
Joined as employee #8. Built the operating system that carried the company to 25 people on a four-day week, then led AI adoption from zero to the entire team. Own finance and admin operations with two direct reports, and client operations for the largest account. Led the Spain-to-UK employment transition, hiring for every role, and the GDPR playbook.
Earlier
Film production house — 7 people
Grew it to €300K in revenue. Learned what a badly scoped project costs you at 2 a.m., which is why my scoping documents are boring on purpose.
Before that
Samsung
Where the habit of arguing with numbers instead of opinions comes from.
Contact
Hiring for an operations, AI implementation or solutions role? Or sitting on a stack that does not talk to itself? Send one paragraph about the situation — I will tell you straight whether I am the right person for it.
filgawel@gmail.com