Simple at the base.
Strong in the structure.
Powerful at the top.
BXL79 AI builds autonomous agents on infrastructure engineered to hold — so the complexity you feel is the complexity you chose, not the complexity you inherited.
About BXL79 AI
We build AI the way you'd build a bridge — load-bearing parts first.
BXL79 AI is a Brussels-based studio designing autonomous agents for operational work: the orchestration, the customer conversations, the monitoring, the parts of a business that run whether or not someone is watching. We don't ship demos. We ship agents that are still running correctly a year later, because we build in the order things actually fail — base, then structure, then the capability on top.
01 — Base
Simple at the base
Clean data, well-defined interfaces, no hidden state. Every agent we ship stands on infrastructure a new engineer can understand in an afternoon.
02 — Structure
Strong in the structure
Orchestration that survives a bad API response, a missing field, a retried request. The middle layer is where most AI systems quietly fail — ours is the part we over-engineer.
03 — Apex
Powerful at the top
Once the base holds and the structure survives contact, the agent at the top can take real initiative — because everything under it is trustworthy.
The agents
Four agents, one chain of command
Each agent has a narrow job and a clear escalation path — to another agent, or to a person. None of them guess when they should ask.
ORIN
Base layerOrchestration agent
Reads incoming work, breaks it into steps, and routes each step to the right tool or agent. The coordination layer everything else in a BXL79 deployment sits on.
CASSIA
Structure layerAnalytics agent
Watches operational data as it arrives, flags what changed, and drafts the explanation before a human asks for it.
FENN
Structure layerConversation agent
Handles customer and partner conversations end to end, escalating to a person only when the exchange genuinely needs one.
KESTREL
Apex layerOperations agent
Sits above the other agents, watches for drift and failure across the whole system, and decides what gets attention first.
Services
Engagements built around what fails, not what demos well
Agent design & deployment
We scope the job an agent is actually meant to do, build it against your real systems, and put it into production with monitoring from day one.
Working agent, deployed and owned by your team
Systems integration & data pipelines
Most agent failures are integration failures. We connect your data sources, APIs, and internal tools so agents act on current, correct information.
Pipelines and connectors documented and versioned
Applied research & model tuning
When an off-the-shelf model underperforms on your domain, we fine-tune, evaluate, and iterate until accuracy holds under real conditions.
Evaluation report and a model that clears your bar
Managed AI operations
Ongoing oversight of deployed agents: usage review, drift detection, incident response, and quarterly recalibration.
A standing operations relationship, not a one-off handoff
Contact
Tell us what's supposed to run itself
Describe the workflow, the team, or the bottleneck. We'll reply with which layer — base, structure, or apex — needs attention first.
- Studio
- Brussels, Belgium
- hello@bxl79.ai
- Response time
- Within two working days