Four stages. Fixed order. Documented output at each step.
The same method is used for automation, digitalisation and security work. Each stage ends with a deliverable you can read and sign off. Durations below are typical for a single workflow or site.
STAGE 01 · 2–5 DAYS
Process Audit
Processes, tools and data flows in scope are mapped with the people who do the work. Repetitive, rule-based tasks are listed and ranked by hours saved and risk. Output: automation map with the first candidates.
STAGE 02 · 2–5 DAYS
Blueprint
Each workflow is designed step by step: triggers, AI steps, rules, approvals and the systems it touches. Data handling and access rights are fixed before any build. Output: blueprint with fixed scope and price.
STAGE 03 · 1–6 WEEKS
Build & Integrate
Agents and workflows are built on frontier or open-source models, in the cloud or on local infrastructure, with Python, n8n or other tools chosen by the project’s needs. They are connected to your systems and tested on real data with your team. Output: live workflows with documentation.
STAGE 04 · ONGOING
Run & Improve
Every workflow is monitored, with alerts when something fails. Changes and new automations are handled on a monthly support plan. Output: run log and a monthly improvement report.
Where AI fits, and where it does not
AI is used where it is reliable: reading documents and emails, classifying requests, extracting data and drafting replies. Anything that must be exact, such as prices, amounts or stock, runs on fixed rules. Steps with financial or customer impact go through human approval, and no client data is used to train shared models.
WHY SECURITY IS BUILT IN
Automations touch customer data, money and production. A workflow that works in a demo can still leak data, double-book an order or fail silently at night. Every SourceGround build defines access rights, data handling and logging before the first run, the same way plant systems are built. That is also why the same workflows can later pass a security review or an audit without rework.
Built by an engineer who has run production systems
Who does the work
Whoever wins the bid, often learning the tool on your project.
Project staff, slow to start and priced for large companies.
A former Tier-1 Senior IT/OT Lead, personally on every engagement, with a lead IATF auditor brought in when the scope requires it.
Output
A working demo. Documentation and error handling are extra.
A long project plan before anything runs.
Live workflows with monitoring, documentation and a named owner.
Reliability
Breaks quietly when a tool or file format changes.
Office-IT assumptions applied to every system.
Monitored runs, alerts on failure, changes planned around operations.
Security & data
API keys in shared accounts, data sent wherever the tool sends it.
Security handled by a separate team, later.
Access rights, data handling and logs defined before the first run.