Work piles up between systems
People move information from inboxes to documents, ERP, CRM, and spreadsheets. The process takes hours and creates room for missed details.
We turn a valuable, repeatable process into a governed AI workflow—built in your cloud accounts on European infrastructure you control, with people in the loop where judgment matters.
A short fit call. If the work is a match, the next step is a free diagnostic.
Illustrative specialist-manufacturing example. Other workflows use the same shared rules, with their own access, checks, and approval points.
Why now
AI can process information at machine speed. The advantage comes from systems that turn it into dependable work, shaped by your context and controls.
People move information from inboxes to documents, ERP, CRM, and spreadsheets. The process takes hours and creates room for missed details.
Quotes, exceptions, and customer decisions rely on experience that is difficult to find, share, and apply consistently.
AI in separate products brings separate context and instructions. Passing work between them can add calls, delay, and cost without creating one dependable process. A well-designed workflow uses only the context and tools the job needs, with clear permissions and checks.
If a critical workflow depends on one model or cloud provider, a change in its terms, access, or location could disrupt the work.
The work, made concrete
A first workflow takes a specific, repeatable process and turns its inputs into a checked result. These are examples, not client case studies.
Read specifications and drawings, check feasibility and pricing rules, prepare a draft, and list unanswered questions.
Check a requested change against inventory, supplier commitments, production or delivery capacity, and customer terms. Surface conflicts and prepare options for the team.
When a supplier, site, or service is disrupted, assemble affected orders, customers, dependencies, and relevant obligations. Prepare a prioritized response for review.
Before building, we set a baseline and acceptance tests with you. Depending on the workflow, we might measure time per case, correction rate, turnaround, or backlog. At handover, we review the result against the agreed measures.
How it works
For each workflow, we define its business context, what data and tools it can access, where information is processed, and when a person must review an action.
Shared rules govern the workflow. The workflow runtime carries out one defined process within them.
A clear first step
Start with a useful process. Establish the shared context and controls it needs. Decide what to expand after you see it working.
We agree the workflow, systems, measures, and acceptance tests before the build. At handover, your team has the code, configuration, and operating guide.
50% at signing; 50% when the agreed acceptance tests pass. If a test fails, we fix it and retest.
Your IT contact or a freelancer runs it using your access, the source code, and the operating guide.
Monitoring and support for the workflows we built, in your cloud accounts. +€500/month + VAT per further workflow; three-month minimum, then monthly notice.
Fixed scope and price, quoted for each workflow on the shared rules already in place.
The free diagnostic, in practical termsA 90-minute session, followed by a two-page write-up within a week: three ranked workflow candidates, the recommended first workflow scoped for the 30-day build, and how it will use systems and be tested. Yours to keep, whether or not you proceed.
Choose the right first process
A strong first workflow repeats often, pulls information from different places, and has a result the business can measure. You don’t need an in-house AI team. You do need someone who knows the process and can help define what good looks like.
Invoices, enquiries, reporting, document checks, or another process your team handles again and again.
The work crosses email, documents, and business systems. People spend time finding, copying, checking, and re-entering it.
Time, delays, backlog, rework, or missed opportunities make this process worth improving—and give us a way to measure progress.
An owner, CEO, or managing director at an owner-led European company—typically 30–250 people—can name the process and bring its owner into the work. Your team keeps responsibility for business decisions; we design, build, and hand over the workflow.
A general-purpose chatbot, a strategy deck without implementation, or unchecked AI decisions about consequential business matters.

Ivan Sanz
Built by an operator
My work spans the infrastructure behind modern business: hardware and supply chains, data centers, global networks, and software that coordinates work. At Fastly, I led network strategy and capacity across more than 150 Tb/s, focusing on resilience, failover, and vendor diversity. Through Listo Digital, I advised carriers, infrastructure operators, and investors on network and data-center strategy, satellite infrastructure, market entry, and technical due diligence across North America, South America, Europe, Africa, Asia, and Australia. I’ve also built software and operating processes for businesses I run.
I studied political science at UC Berkeley and earned an Executive MBA in finance and strategy. That mix shapes how I see AI: its value depends on infrastructure, ownership, regulation, capital, and geopolitics. Berlin Operators helps European companies put AI to work while keeping control over how it runs and how they adapt.
Questions, answered
A workflow is only useful if the business can trust it, operate it, and change it.
We talk through what runs slowly, where information gets stuck, and whether the process fits a first workflow. If it does, we offer a free diagnostic. If it doesn’t, we’ll tell you.
No. The diagnostic uses conversation only. Files, credentials, exports, and system access come after a contract and data-processing agreement.
Before the build, we agree what data and systems the workflow can access, which model handles each task, where information is processed, and where human review is required. The system runs in cloud accounts your company controls, with a European provider by default. EU-only processing can be specified and checked for the full deployment path.
Model routes are configurable. A replacement is tested against the agreed workflow checks before going live. Moving hosting providers or infrastructure is a planned migration, not an instant switch.
If a test fails, we fix the workflow and run the agreed tests again. The second 50% is due when those acceptance tests pass.
Someone needs to handle access, updates, monitoring, recovery, and model changes. Your team can do that with the handover materials, or you can choose the monthly run service.
Start with the bottleneck
Talk it through in 20 minutes. If there’s a fit, the diagnostic maps the right first workflow and how we’ll measure it.