Everything you need to build, run and trust Orcana agents
Agents, deterministic workflows and dashboards, governed and running in your cloud.
One question, every team, one loop from start to finish
Every agent runs that same loop, and you approve every move
Watch
Sales, CRM, claims, hub, payer and media feeds, 24/7/365.
Diagnose
Runs a deterministic workflow to find out why.
Ask
The owner gets the finding and the move in Teams, Slack or email.
Hand off
Calls the agents that do the work, straight into CRM and media tools.
Learn
Tracks every action against its result.
Then it goes round again. Every loop starts smarter than the last, so outcomes compound.
Behind every loop, four building blocks your teams use every day
A plain-English analyst
Ask in plain English. Get the answer, the working and a deck.
Saved workflows
The analyses you repeat, from sizing to campaign lift, saved and rerun in minutes.
Briefs, dashboards and alerts
Each owner gets their numbers on their schedule, plus a heads-up when something moves.
Loops that run themselves
A pod of agents chains the other three together, working as one.
Underneath every agent, locked metric definitions + rules + deterministic workflows
AI at every step
How most AI analytics tools work
Orcana
AI where it's strong, code where it has to be exact
definitions.pyHuman-tested code · 100%data_checks.pyHuman-tested code · 100%line_of_therapy.pyHuman-tested code · 100%@workflow("line_of_therapy", version="3.2.0", validated="2026-10-12")
def line_of_therapy(claims: Table, defs: Definitions) -> Table:
"""Assign every patient's line of therapy from claims"""
regimens = build_regimens(claims, gap_days=defs.regimen_gap)
lines = regimens.assign_lines(
advance_on=defs.new_agent_added,
ignore=defs.supportive_care,
)
check(lines).no_gaps().one_line_per_patient_per_day()
return linesEvery workflow is a Python file
Versioned in your repository, reviewed like software and run the same way every time.
One definition per metric
What counts as a call, a new start or a reached target lives in one tested file, used everywhere.
Bad data gets caught first
Every table passes its health checks before an agent touches it. When a sample is small, it says so.
18 workflows live inside pharma today, and more every quarter
Field and targeting
5 live · 2 nextPatients and access
5 live · 2 nextMarketing and brand
4 live · 3 nextPlanning and launch
4 live · 3 nextLiveNext on the roadmap
Built with your teams, by pharma AI engineers
Customize
Tailor workflows, definitions and rules to how your company works.
Deploy
Stand up the agents and your private model inside your VPC.
Maintain
Retrain, monitor and revalidate, so models and workflows stay accurate.
Scale
Add the next decision loop, and the next team, on the same foundation.