01 · pattern
AI agents and automation for operations teams
// systemartis · est. 2023
- Founded
- 2023
- Based
- Bucharest, RO
- Engagements
- EU · NA · ME
- Founded
- 2023
- Based
- Bucharest, RO
- Engagements
- EU · NA · ME
// selected clients · partners
// real metrics · real engagements
Numbers from production systems, not pitch deck math.
- 70%Founder dependency reduced
// systemartis
- 20h+Reclaimed per week
// systemartis
- 99.9%System availability
// systemartis
- +40%Sales increase
// innovium
// the problem
Eliminate the bottlenecks that hold you back.
The same four problems show up in almost every business we work with. None of them require a new app. They require a system.
- 02 · pattern
Leads slip away without consistent follow-up
- 03 · pattern
Outdated workflows hurt customer experience
- 04 · pattern
Scaling requires more people and higher costs
// our services
Custom systems. Not templates.
Four ways we ship, from scoping a problem to running production systems. Every engagement is shaped to your business, not a pre-packaged tier.
Card 1 of 4
// shape of a system
We don't have a universal product. Here are four shapes we've built.
Every engagement is a custom backend, a custom dashboard, and a custom set of agents, branching, looping, deciding. Below: four real problems with the architecture you'd understand and the dashboard your team would use every day.
Sales reps drowning in unscored leads. Manual qualification eating four hours a day, half the team is selling to the wrong companies.
// outcome, Every inbound is enriched, scored, and routed in under two seconds, with stale leads automatically pulled back for re-enrichment.
// the system
The tech we build with.
Models, infrastructure, automation runtimes, and the languages that hold it all together. We pick the right tool for the job, then own the integration end-to-end.
// not listed? send the API docs and we'll tell you how it plugs in.
// case studies
Real systems. Real numbers.
Two engagements. One we built for ourselves, one for a client. Both still in production.
Card 1 of 3
// case studies
Real systems. Real numbers.
Two engagements. One we built for ourselves, one for a client. Both still in production.
01 · case2026 
client · independent garage, france
How an AI voice agent answers every call a French garage used to miss
Built for an independent French garage. Identifying details anonymized at the client's request.
An AI voice agent that answers an independent French garage's phone in natural French, checks live calendar availability, and books the appointment during the call, with RGPD (GDPR) compliance built into the data layer.
- €7KMissed-call revenue at stake
- 24/7Inbound calls answered
- 0Callbacks to book
Retell AITwilioGoogle Calendar APIPostgreSQL + DrizzleHono + HTMX02 · case2026 
client · systemartis
How we built an AI operational brain to reduce founder friction
Built for ourselves (internal), then offered to clients.
An internal operational system we built to remove the founder bottleneck, onboarding, approvals, exception handling, knowledge retrieval, all coordinated by a small fleet of AI agents wired into our real backend, not a Zapier flow.
- 70%Reduced founder dependency
- 20h+Saved per week
- 99.9%Availability
Custom backendCustom dashboardLLM orchestrationVector storen8n03 · case2025 
client · innovium
Marketplace pricing intelligence & operations automation for a multi-country seller
A pricing-intelligence and operations automation system for a multi-country marketplace seller. Continuous price monitoring across regions, automated repricing logic, and operations dashboards that replaced nine manual spreadsheets.
- 2×Productivity
- +40%Increase in sales
- 9Automated reports
Custom backendMarketplace APIsPricing engineOperations dashboardScheduled jobs
// pricing
Two ways to work with us. No fixed tiers.
We don't sell off-the-shelf subscriptions because we don't build off-the-shelf software. Every engagement is quoted after a free 30-minute scoping call. You pick the mode; the price is shaped by the scope.
One-time project
We build it, you own it. Once it ships, you host it on your infrastructure and your team takes over the day-to-day.
// What's included
- Discovery, solution design, and architecture
- Custom backend, dashboard, and agents
- Integrations with your existing tools
- Deployment to your hosting environment
- 30 days of post-launch bug-fix support
- Documentation and team handover
// Not in this mode
- Ongoing hosting (you handle)
- Continuous maintenance after the 30-day window
Best for teams with internal engineering capacity.
// need both? a project to ship · then monthly to run · normal, common, often the best mix.
// frequently asked
Questions, before you ask.
If yours isn't here, the answer is probably "yes, let's talk about it."
// writing
Notes from production.
Things we learn while shipping. No engagement-bait hot takes, only what we'd want to read ourselves.
52% of enterprises say they have AI agents deployed. Only 5% worry about tool-calling accuracy. The gap between 'deployed' and 'reliable' is where the real work begins.
Reading is passive. Real value comes from execution. Modern agents treat the LLM as a routing engine, validating inputs, calling tools through MCP, and gating destructive actions behind a human, not as a chat surface bolted onto a product.
Stripe, Salesforce, and your own DB each think a 'Customer' is something different. The real engineering work isn't writing the pieces, it's making ten systems that disagree about reality stay in sync without ever dropping a task or duplicating an order.
Card 1 of 3
// writing
Notes from production.
Things we learn while shipping. No engagement-bait hot takes, only what we'd want to read ourselves.
Insights Most AI Agents Aren't Production Ready52% of enterprises say they have AI agents deployed. Only 5% worry about tool-calling accuracy. The gap between 'deployed' and 'reliable' is where the real work begins.
AI Strategy The Shift to Agentic: Why Basic Chatbots Are DeadReading is passive. Real value comes from execution. Modern agents treat the LLM as a routing engine, validating inputs, calling tools through MCP, and gating destructive actions behind a human, not as a chat surface bolted onto a product.
Insights Why 'Glue Code' Is the Most Important Skill in Modern EngineeringStripe, Salesforce, and your own DB each think a 'Customer' is something different. The real engineering work isn't writing the pieces, it's making ten systems that disagree about reality stay in sync without ever dropping a task or duplicating an order.