Human-led automation
You cannot automate a process you haven’t measured first.
Fix the process first. Then automate it.
I redesign the workflow, then build and run the automation for you.
For logistics, freight, pharma and manufacturing teams where the operations load is heavy. I sit with the people doing the work, map the process as it actually runs, then build what takes the manual handling out of it. EMEA first, US second.
The first conversation is about your operation, not a slide deck. Fixed fee, quoted on request.
01 / Why I say no
I say no.
I get inbound requests constantly: “We need you to automate our email intake.” I listen. I take notes. Then I say no. Not because email intake is not automatable — it usually is. I say no because I will not know whether it is the right thing to automate until I have watched it happen.
Plenty of consultants take the other route: tell me what you need, and I will build it. It is a fast way to end up with beautiful automation nobody uses. The build does exactly what the specification said and changes nothing, because the specification was written from how the team thinks the work flows rather than how it flows.
The most expensive automation is the one nobody uses.
So Assess always comes first. You show me how the work actually flows — not how you think it flows. I check that flow against how your business makes money, then design the standard version of it. Then you and I decide what to automate, and in what order.
My stance
The audit is not overhead.
It is the difference between solving the real problem and building something impressive that fails. I will not start a build without one — the order is the whole point.
If what you want is a tool bolted onto the process as it stands today, I am the wrong person for the job, and I would rather say so on the first call.
Yellow flag
If the audit's cost or timeline is the first thing that gets pushed back on, that is a yellow flag. It usually means a quick fix is wanted, not something that sticks. Better to find that out on the first call than halfway through a build.
02 / What the audit found
They asked me to automate their email order processing. The audit found the constraint was never the processing.
One anonymised engagement, and what the audit found before anything was built.
Anonymised case record
Freight forwarder · ≈40 people · Benelux
01What they asked for
Their words
“Automate our email order processing.”
Their stated goal: handle more orders without more headcount.
02What the audit measured
In their numbers
| Finding | Figure |
|---|---|
| Orders arriving incomplete | 23% |
| Time one person spent chasing the missing information | ≈6 h / week |
| That same person, chasing across a year | 300+ h / year |
| Manual labour in that chasing | €12–15k / year |
| Orders delayed each week, by one to three days | 15–20 |
The audit’s baseline for this client, and the numbers the recommendation was built from.
Automating the processing would have made incomplete orders arrive faster. The constraint was never the processing speed — it was the 23% arriving with something missing.
03What I recommended instead
Missing detail treated as the normal case
01
Auto-parse what arrives.
02
Automatically chase the customer for whatever is missing.
03
Create the order only once the details are complete and the credit check passes.
04
Escalate to the owner on Slack when it does not.
05
Email the shipping label.
Why this approach
Hiring someone to build what you ask for would have delivered it — faster processing, for orders that were already arriving incomplete. The audit found what the brief could not. That is why it comes first: I would rather spend the time finding that out than hand you a system that speeds up the wrong thing.
03 / The workflow
One order, end to end — seven steps, four components, three systems.
Workflow · order intake, freight booking
07 steps · 04 roles · 03 systems
01The workflow, as it runs
From the first enquiry to the saved order
01
New enquiry arrives
Trigger
A form submission, an inbox, or a new row in the TMS — whatever already happens in your operation.
02
Clean up and check the data
n8n step
The fields, the references and the attachments, checked before anything gets written anywhere.
03
Read the documents and pull out the details
Claude API
A purchase order as a PDF, a supplier confirmation buried in an email thread, a rate sheet in three different formats — all read and turned into clean data the rest of the workflow can use.
04
Match it to the customer, order and rate
n8n step
Looked up against the records you already hold, not a copy held somewhere else.
05
Turn supplier rate sheets into one format
Custom code
Three supplier formats into one — the step ready-made blocks cannot do on their own.
06
Save it into your systems
n8n step
Order, quote and status created where your team already looks for them.
Saved to → TMS · order management
07
Confirm and notify
n8n step
Supplier confirmation out, customer status updated, nobody re-keying any of it.
Saved to → Customer database
04 / Assess · Automate · Support
Three kinds of work, in this order
Assess how the work runs and how your business makes money at the same time, automate what matters, then support what is live. Start with the automation and you get something impressive that nobody uses.
01
Assess / Fixed scope · 4 weeks
One standard process, built around how you make money
I do not automate a process until I have watched how the work really runs, checked it against how your business makes money, and worked out the one standard way it should run. That is the discipline I brought to Phoenix Contact and Merck KGaA: looking at the process, the team and how the business makes money as one picture, before anything gets automated. The expensive mistake is automating a process that should have been redesigned first.
What happens in this phase
- The process as it runs today, mapped end to end
- What that manual work costs you, in your own numbers
- What it means for how you make money
- The standard way it should run, with the bottlenecks marked and an owner named for each step
- The changes worth making, ranked by what each one is worth
02
Automate / Planned after the Assess phase · 8 weeks
Workflow automation in n8n
Then the build: n8n workflows I set up, run and maintain for you, fully documented. Ready-made blocks where those are enough, hand-written code where they are not, and a Claude API step for the work that means reading and understanding documents.
You are not handed a system and left with it: the build goes live with ongoing support already in place.
What happens in this phase
- n8n workflows, fully documented
- Built, launched and maintained by me
- A Claude API step for reading documents
03
Support / Keeping it running · fixing it · improving it
Monthly retainer and ongoing improvement
Automation is not a one-off install. I keep it running and keep improving it as your volume and process change.
A live system needs someone watching it, fixing it when it hits a case nobody planned for, improving it as volumes grow, and changing it as the business grows — the difference between a system that works for six months and one that works for years. I own the technical layer. Your team owns the outcomes.
What happens in this phase
- Live workflows watched and kept running
- Fixes and improvements over time
- New workflows as your volume grows
One phase leads to the next. Nothing gets built until the Assess phase has shown what is worth building.
05 / Human-led, in practice
AI does the building. I decide what gets built, and I answer for it running.
AI is doing more of the building than any one person could do alone. What it does not do is decide. What gets automated, in what order, and whether it should be automated at all comes from watching your operation, and from running and rebuilding the workflows myself. The tools are AI's. The judgement and the accountability are mine.
- A human decides what gets automated
- A human answers for it running
- The tool is AI — the accountability is mine
06 / Who you're working with
I am Olaolu Matty, and you work with me directly
I started in US Army global logistics, running freight across supply chains on several continents, then spent years on business transformation for Phoenix Contact and Merck KGaA across EMEA. Both taught me the same lesson: the expensive part of an operation is rarely the work itself. It is the handoffs, the re-keying, and the systems that do not talk to each other.
My credibility is operational, not technical. I can read a process and I can build the thing that fixes it — which is a rarer combination than it should be.
I work solo, from Lisbon. The assessment, the build and the ongoing support all come from the same person, and that person is still there when something needs debugging.
Background
- US Army — global logistics, moving freight across supply chains on several continents
- Phoenix Contact and Merck KGaA — business transformation consulting across EMEA
- Master's degrees in Global Logistics & Supply Chain Management and International Business
- Certified in n8n and Claude AI
- Lisbon-based, working EMEA first and US second
07 / Who this is for
Mid-to-large operations teams with more process than capacity
Logistics and freight forwarding companies, and pharma and manufacturing organisations, running complex operational processes across multiple teams. I work EMEA first, US second.
What I look for is how much work is running at once: 10 or more projects on the go at the same time, or a heavy load of admin and coordination where skilled people spend their week moving information between systems instead of making decisions.
Who I speak to
- Operations director — owns the process and the people it takes to run it
- VP of delivery — carries the targets for how much work gets through
- COO — answers for margin and growth
08 / Who this is not for
Being clear about the fit saves everyone a call
Small, single-project teams
If one team runs one project at a time, the audit finds less than it costs, and the build comes too early.
Anyone wanting a generic AI strategy deck
I do not produce general overviews of the AI market or presentations on how ready your business is for AI. What you get is a measured process and a working build.
Software in search of a problem
I start from an operations process that is already costing you money in manual hours. If there is no such process, there is nothing to automate.
09 / Next step
Book an intro call
Tell me how your operation runs today — the projects in flight, how much coordination it takes, and the process that eats the most hours. I will tell you whether an audit is worth doing.
Prefer email? olaolu.matty@matty.consulting