Plan & build · Discover
Idea Hub
Decides what's worth automating — captures candidates, estimates TCO/ROI, and tracks each automation from idea to retirement.
Explore Idea Hub →The Platform
Intelligent automation platform — from rule-based RPA to AI agents
Primo spans a spectrum of executors — deterministic robots, scenario agents, and autonomous agents — so every task runs on exactly as much intelligence as it needs. On-premises, across the systems you already operate.
In production since 2018
100+ enterprise integrations
Native SAP, Oracle & legacy
The Platform Model
RPA and AI: how the platform divides the work
Three responsibilities — plan and build the automation, run it under control, and execute it at the right level of intelligence — over the systems of record you already operate. Orchestrator runs the robots, Primo Epic runs the agents, and AI Server reads documents for both. This is how we think about the platform, not a wiring diagram of every component.
The system is built around one idea: intelligence is a dial, not a default. The same platform runs pure-rule robots and reasoning agents — you choose how much intelligence each task is worth paying for.
The Executor Spectrum
01
Adds: Execution
Robot
Runs a fixed route by rule. No understanding — fast, cheap, stable. The trade-off: it stops when it meets something unexpected.
02
Adds: + Understanding within a step
Scenario agent
The route is still fixed, but an LLM works inside each step: reads free text, answers from corporate knowledge, judges whether a document fits. Built no-code in Primo Epic.
03
Adds: + Choice of the route
Autonomous agent
The model decides the next step, not just the step's content — for work whose path can't be written in advance. More compute, more setup, a human checkpoint on its decisions. Built and run in Primo Epic.
Intelligence has a price — in compute, setup, and oversight. So the platform lets you pick the cheapest executor that solves the task. Efficient automation usually needs all three — not autonomous everywhere.
The model above shows how the work is divided. Here is the full platform — every part, including the ones a model diagram leaves out.
What Each Part Is For
The full platform: RPA tools, document AI and monitoring
Plan & build · Build
Studio
Where automations are built — no-code, low-code, and full-code in one project that deploys to the same robot and orchestrator.
Explore Studio →Run & control · Robots
Orchestrator
Runs the robot fleet under control: scheduling, queueing, monitoring, and an end-to-end audit trail.
Explore Orchestrator →Run & control · Agents
Primo Epic
Builds and runs AI agents, scenario and autonomous, inside your perimeter. Company policies, human approval before changes and a record of every decision. Calls robots through Orchestrator.
Explore Primo Epic →Run & control · Monitoring
ART
Synthetic and real-user monitoring: robot-driven transactions run through your applications on a schedule and correlate with real user activity to catch issues before users do.
Explore ART →Runtime · Execute
Robot
Where work runs. Deterministic execution on Windows and Linux, attended or unattended, including Citrix/RDP/VDI and legacy interfaces — the tier-1 executor.
Explore Robot →Document AI
AI Server
On-premises document AI — OCR/IDP, document classification and LLM text processing. Turns invoices, contracts and scans into structured data for robots and agents.
Explore AI Server →Looking for fit to a specific process or industry? See Use Cases →
The Architecture Argument
Why it's split into modules
Reliability beats novelty
The deterministic robot core stays auditable. AI sits on top of it, where it pays off — it never replaces the part you can trust to behave the same way every time.
Replace one piece, not the platform
Modules are adopted and upgraded independently. You don't re-platform everything to change anything — which is what makes a multi-year commitment defensible.
Reach systems without an API
Where there's no API, a robot performs the action the way a user does — Citrix, virtual desktops, legacy screens. The reach of RPA, under the control of the orchestration layer.
By Design — Platform Properties
What the architecture is built to give you
1,000s
Robots in parallel
Designed to scale unattended execution across the enterprise — not a handful of bots on a desktop.
Win + Linux
Runtime coverage
Robots run on both, attended or unattended — including back-office systems usually out of RPA scope.
On-prem
On-prem AI stack, no data scientists required
The full AI stack runs inside your perimeter; analysts configure it without data scientists.
Modular
Replaceable independently
Adopt or swap a single module without re-platforming the rest of the estate.
Design-capability defaults; actual capacity depends on deployment topology.
Related reading
RPA vs AI Agents: Cost Comparison and When to Use Each →
Choosing between RPA and AI agents is a cost decision, not a technology one. A practical framework to match each process step to the cheaper executor.
Strategy8 min read