€800M
2020–2023 experience · pre-agentic
pricing decisions supported
Rebuilt ERP pricing logic in Python during an SAP migration, then created a transactional database and decision dashboards.
Python · SQL · Power BIBusiness transformation in the AI era
Turn your business know-how into software that moves your company forward.
Less retyping, better-informed decisions and teams that can move forward. I start with how you actually work to design, build and help your teams adopt software that improves operations.
Julien Vaughan · 20 years in operations · 7 years building applications
Agentic engineering puts AI agents to work to accelerate development. I remain accountable for business understanding, architecture, verification and delivery, through to adoption by your teams.
20 years in operations · 7 years building
Operational proof
The outcomes — and how they were achieved.
These results come from earlier professional experience, before my current agentic-engineering approach. Agents now increase my delivery capacity; they do not rewrite the past.
50+
2020–2023 experience · pre-agentic
countries served consistently
Standardised the pricing process and automated email and PDF communications for local teams.
Python · email/PDF automation · data pipelines
95%
2020–2023 experience · pre-agentic
manual workload removed in three weeks
During the first month of my work-study placement at Carrier, repetitive preparation, transformation and distribution steps were replaced with an automated, controllable flow.
Python · SQL · Power BI
27×
Earlier experience · no coding agents
ROI from a targeted digital campaign
€1.3k in spend generated €35.1k in sales over nine months and 80+ new customers through targeting built from business and public data.
SQL · Python · modelling · commercial targeting
25k → 100k
Earlier experience · no coding agents
qualified, rankable prospects
Built a prospecting model combining internal CRM, INSEE and cadastral data to rank the addressable market.
Python · SQL · ETL · public dataThe method
From know-how to software people use
Six stages, concrete deliverables and your teams involved throughout.
I immerse myself in your business, the problems you need to solve and the optimisation you are looking for. Twenty years across sales, hospitality, logistics, airport operations, credit-card back-office operations, pricing and business intelligence help me connect the human, operational, commercial and technical dimensions quickly.
The highest-return opportunities are often overlooked because they live closest to execution: in tasks repeated every day by frontline teams. A few minutes saved, one error prevented or one better-supported decision then compound at scale.
Understand real work
Connect processes, wiki content, interviews and observed usage to business objectives. Plan migration from the start.
Build living specifications
Let domain experts clarify rules, share a glossary and track open questions. A dedicated dialogue workspace is planned.
Prototype with users
Put future workflows in users' hands, test exceptions and adjust priorities before building.
Build and prepare migration
Develop with agents using representative anonymised real data. Rehearse migration before switching over.
Verify against real operations
Where feasible, replay operations and compare results to distinguish regressions from intended improvements.
Deploy, adopt, improve
Open a pilot, improve interfaces and workflows, then migrate progressively while measuring outcomes.
Feedback can revisit any earlier stage. Documentation follows decisions, prototypes and real usage throughout the project.
I lead business discovery, architecture, development with AI agents, testing, migration and deployment. Your domain experts validate rules and results; your users shape how the software works in practice.
Explore the methodConfidential programme · €80M-turnover business
Turning business knowledge into testable architecture.
Existing documentation and first principles become decisions, controlled specifications and then prototypes future users can genuinely challenge.
● In progress: architecture and user-prototype validation.
- Operational knowledge → traceable decisions
- Controlled specs → testable prototypes
- Architecture validated before production
Systems already in the wild
Credibility is also what keeps running when nobody is watching.
Two public products designed, deployed and operated end-to-end — user experience through infrastructure.
OaService SIRET
Bulk enrichment for B2B databases: import, resilient matching, confidence scoring and reusable export.
Why it existsIt began with a practical need: match a business database with official records without checking every line by hand. Time needed to go to ambiguous cases, not repeated lookups.
See the system ↗OaService Rappels
SMS, WhatsApp and email reminders designed to reduce missed appointments without bypassing consent rules.
Why it existsIt was first built to help an osteopath friend re-engage patients, reduce missed appointments and request reviews without manual follow-up.
« With automated reminders scheduled every four weeks, our retention rate has improved significantly and our no-shows have fallen considerably. Within a few months, we received hundreds of additional Google reviews, with no manual effort. OaService Rappels is simple to set up and runs on its own. »
Boina MZE, osteopath D.O. · La Maison de l'Ostéopathie, Meximieux · 349 reviews · 5.0 ↗
Field experience
“I also write”
Articles written with AI assistance from my real-world experience — data, automation and development.
18 August 2026
Cheaper code is making custom software economically viable again
Agentic engineering changes the economics of IT projects: friction that was once too small to justify software can now generate a very concrete return.
Read the article ↗ 20 March 2026A missing SIRET is not a detail. It is a decision waiting to happen.
B2B enrichment only matters when it advances a commercial, accounting or operational decision.
Read the article ↗ 24 March 2026A missed appointment is a system failure, not just a forgotten date.
Effective reminders are more than an SMS: they make the next action clear and respect the customer relationship.
Read the article ↗ 28 March 2026AI accelerates code. Specifications accelerate the system.
Agentic leverage comes less from a spectacular prompt than from explicit decisions, constraints and verification.
Read the article ↗The Process Teardown
Send the process costing you time, money or energy.
Answer a few questions to get an initial structured view of the friction worth validating. It is a working hypothesis, not an automated diagnosis or quote.
Do not enter secrets, passwords, confidential documents or sensitive personal data.
Or contact me directly
contact@oaservice.fr
France · Canada · United Kingdom · remote