Orcana.AI
Building Pharma’s Commercial Org Twin, Run by AI Agents

Pharmaceutical companies still make their biggest commercial decisions on a calendar. Brand plans are set once a year, targeting is refreshed every six months, and campaign results arrive at the quarterly business review, often after the next campaign has launched. Meanwhile the market keeps moving. A competitor wins a label expansion, a payer adds a step edit, patients stall between approval and first fill. By the time a decision is revisited, the moment has usually passed.

Arjun Bidesi, Founder and CEO of Orcana.AI, believes that rhythm belongs to the world before AI. “Every industry will run on agents, and the companies that start first will compound fastest,” he says. “We are building that for pharma.”

Orcana.AI’s answer is a new way of working. Instead of being made once and reviewed months later, each decision runs as a loop. Agents watch the data around the clock, bring the signals that matter to the person who owns the decision, act once that owner approves and measure what happened. Then the loop runs again. Owners stay in control of every move while the agents do the watching, the analysis and the follow-through. Because every result feeds the next decision, outcomes do not simply improve once. They compound.

“Today, a question becomes a project, and a decision gets revisited next quarter,” Bidesi says. “With Orcana, it becomes a loop that runs every week. The team that runs the most decision loops wins.”

From Consulting Projects to Decision Loops

Bidesi saw the cost of the old rhythm firsthand. Over more than a decade in data science and pharma consulting, he watched the same high-value questions return across brands two to four times a year, each one a new consulting project at $150,000 to $300,000, each one delivered eight to twelve weeks later, after the market had moved, with none of the thinking kept.

“The problem was around running same repeatable grunt work of cleaning the data, staging the data across various projects to get to the exciting part of a consulting project, which is the insights and the recommendations,” Bidesi explains. “Orcana.AI aims to reduce the dependence while preserving the rigor needed to support consequential commercial decisions.”

Orcana.AI began by taking that repetitive work off analysts’ plates. With Orcana 2.0, it has set out to change the rhythm itself, by giving every pharmaceutical commercial organization a twin that runs on agents.

A Commercial Org Twin, Run by Pods of Agents

The commercial org twin is a working mirror of a company’s commercial organization, in which every team, signal and decision is worked by AI agents around the clock. Each team, from brand and marketing to sales operations, market access, patient services and the field, gets its own pod of custom agents. As in an orca pod, each agent has its own role, and the pod works as one.

The pod watches the team’s sales, CRM, claims, hub, payer and media data 24/7. When something moves, such as a dip in new patient starts or a competitor’s label expansion, it diagnoses why and recommends a response. Once the owner approves, the agents carry the work into the tools teams already use, re-ranking targets in Veeva or Salesforce, re-planning a district’s week or shifting media spend, and then measure the result against a matched control group. A brand team that once tested four ideas a year can test dozens, keep what works and stop what does not.

This is what sets Orcana.AI apart from chatbots, dashboards and next-best-action engines. Those tools answer a question or display a number. Orcana.AI’s pods close the loop from signal to measured outcome, with a person approving every move.

AI That Reasons, Code That Calculates

Agents are only as trustworthy as the numbers beneath them. In Orcana.AI’s architecture, generative AI reads the question, coordinates the work and explains the findings, while every calculation runs through deterministic workflows: validated, versioned Python that people have tested and reviewed like software. Metric definitions are locked, data health checks run before an agent touches a table, and the rules of pharmaceutical analytics, from privacy to who can see which data, are built into the platform rather than left to a prompt.

“We are the first architecture to make structured analytics fully deterministic,” Bidesi says. “We do not let AI do the math.” A competitor, he argues, would need to rebuild the engine, the hand-offs between AI and code, and the rule book before shipping its first answer.

A Pharma Model, Fine-Tuned Into Your Own

Beneath the agents sits the Orcana pharma model, trained on harmonized public and licensed partner data, from drug labels, clinical trials and literature to provider, prescribing, payer and de-identified patient journey data. It works out of the box and comes in three sizes: Orcana-5B, light enough to sit inside every agent and triage signals all day; Orcana-20B, the analyst that turns a business question into a defensible study design; and Orcana-30B, now in training, for launch, portfolio, pricing and loss-of-exclusivity decisions. All three are smaller than frontier general-purpose models.

Each customer then receives its own private model, the Orcana model fine-tuned on the company’s sales, CRM, hub, specialty pharmacy and media data and on its past decisions. It learns the company’s brands, markets and definitions, improves with every decision loop and belongs to the customer, IP included. Everything runs inside the customer’s own cloud on Snowflake, Databricks, BigQuery, Redshift or Azure, and no data leaves that environment.

From Weeks to Minutes in Production

Orcana.AI entered pharma through a competitive proof of concept in which seven vendors were given the same data warehouse and the same commercial questions. Bidesi says Orcana.AI stood out with fully repeatable results and zero hallucinations. The client then gave Orcana.AI a campaign-measurement study it had planned to hand a consulting firm for $150,000. Orcana.AI built an agent using retroactive experiment design, matching exposed physicians with comparable unexposed ones and tracking both over time, and the client has since moved all of its campaign measurement to Orcana.AI.

A real-world data provider tells a similar story. Each of its pitches to large pharmaceutical companies took three to four weeks of analysis, from defining patient cohorts to sizing the opportunity. Orcana.AI turned that work into workflows that run in two to three hours, and the provider can now deliver as many as three data-backed pitches in a day.

Customers have noticed. A director of analytics at a $30 billion pharmaceutical company called Orcana.AI “the most sophisticated offering we’ve seen in the market.” A senior vice president and general manager at a real-world data company went further: “If this continues to deliver at scale like the first project, this will be a true technological marvel for the pharma industry.”

Customers typically fund Orcana.AI from existing consulting budgets, at roughly a fifth of the cost. “It’s displacement, not new spend,” Bidesi says. Teams usually start with a single decision loop running on their own data within 7 to 14 days, then add pods for more teams.

The Team Behind the Vision

Bidesi brings about 15 years of data science experience, including a decade in pharmaceutical consulting, where he led data science teams and helped build one of the first AI-driven pharma marketing platforms for a major client. His co-founder and CTO, Reehan Shah, is an expert in natural language processing and generative AI who built AI products for large enterprises long before generative AI went mainstream. The two met while studying operations research at the University of California, Berkeley.

Their premise is simple: commercial decisions should not wait for the next quarter, and pharmaceutical companies should not keep paying for the same analysis on data they already own. Every commercial team will soon run on its own agents, and the companies that start first will pull ahead. For building the commercial org twin that makes this possible, Orcana.AI has been named Top AI Pharma Agentic Data-Science Platform 2026.