Skip to content
Systems of record · 01

Pharma OS

Pharma has no shortage of systems.

A commercial organisation typically runs six or more, and every one of them holds part of the answer.

Veeva CRM
Calls, HCP interactions, territory activity
Rx data
Prescription volume and share
HCP master
Prescriber identity and affiliation
SAP
Supply, orders, distribution
Brand plans
Strategy, targets, allocation
Medical systems
Evidence, MSL activity, enquiries

None of them is missing data. What is missing is a layer that can reason across all six at once.

Plate data

Industry
Pharmaceutical
Status
Product
Agents
06
Conformance
Level 3
Approval
Declared per agent

Built, deployed and operated by us. The first Industry OS on this architecture.

The four questions · 02

The questions a cycle meeting is actually about.

  1. Why did this brand slow in this region?
  2. Which territories will miss next quarter?
  3. Which HCPs should this rep see next cycle?
  4. Which system carries out the change, and who signs it off?

Today these are answered by an analyst with an extract and a week. The knowledge is present. The reasoning is manual.

The ontology · 03

Pharma is not a generic enterprise.

Molecule, brand, indication, HCP, territory, call, prescription, formulary, guideline, promotional claim. These are entities with rules attached, not fields in a database.

Ontology fragment · pharmaceutical

ENTITY        brand
RELATIONSHIP  brand -> prescribed_by -> hcp
EVENT         prescription_written
METRIC        territory_growth
RULE          off_label_promotion  (enforcement: block)
DECISION      hcp_prioritisation
ACTION        crm_task_create

Because the rules sit in the ontology, a recommendation that breaches one never reaches a human to approve.

The agents · 04

Six specialists, each with a declared boundary.

Commercial

Which HCPs should this territory prioritise next cycle?

Medical

What is the evidence position on this indication, and where is it weak?

Field

What changed in my territory since my last call plan?

Brand

Why is this brand growing in one region and flat in another?

Supply

Where will demand and supply diverge in the next quarter?

Executive

What are the three things about this portfolio I should know today?

Commercial
Reads CRM, Rx data, HCP master, brand plans. Recommends prioritisation and next best action. Creates CRM tasks.
Medical
Reads medical systems, published evidence, enquiry logs. Recommends evidence gaps. Executes nothing without medical sign off.
Field
Reads CRM, Rx data, territory structure. Recommends call plan changes. Generates call plans.
Brand
Reads Rx data, brand plans, territory performance. Recommends allocation changes. Recommends only.
Supply
Reads SAP, order history, demand signals. Recommends supply actions. Creates supply alerts.
Executive
Reads across all of the above. Recommends nothing. Explains everything.
The pharma pack · 05

What ships with it.

Ontology
Pharma entities, relationships, events, metrics and rules, modelled and maintained.
Knowledge graph
The industry structure, ready to be grounded in a company's own brands and territories.
Reference workflows
Cycle planning, targeting, call planning, brand review, supply review.
Agent templates
Six agents with default manifests and approval boundaries.
Metrics
Defined once, so two agents cannot disagree about what growth means.
Connectors
Veeva, Salesforce, SAP, Rx data providers, HCP master.
Evaluations
Test sets for each agent, including compliance cases designed to fail safely.
Guardrails
Regulatory rules enforced before a recommendation is produced.
What it connects to · 06

Read paths and write paths, declared separately.

Per system, per agent. No system of record is migrated, and no master data moves.

Veeva CRM
Read · write: task creation only
Salesforce
Read · write: task creation only
SAP
Read · write: alerts only
Rx data
Read · no write
HCP master
Read · no write
Brand plans
Read · no write
Deployment · 07

Where it runs, and what it never widens.

Where it runs
Inside your environment, or in a dedicated tenant. Company knowledge does not leave your boundary.
Model choice
Set per workload and per data boundary. Changing model does not change the OS.
Identity
Your directory, your groups, your existing permissions. The OS never widens access.
Audit
Every reasoning step and every action, retained and exportable.
How a deployment starts · 08

Narrow enough to judge in a cycle.

  1. One brand, one region. Scope narrow enough to judge in a cycle.
  2. Grounding. Your brands, territories, definitions and policies loaded into the Company Pack.
  3. Boundaries agreed first. Approval lists are signed off before any agent runs.
  4. Evaluation before rollout. Agents are tested against your own historical decisions.
Governance and approval · 09

In pharma, the boundary is the product.

Every agent declares what it may never do alone. The runtime enforces it, and the enforcement does not depend on model behaviour.

requires_human_approval:  external_communications  material_strategy_changes  incentive_change  promotional_claim
Blocked before reasoning
Off label recommendations are never produced, not filtered afterwards.
Held before action
Anything on the approval list is routed to a named human.
Recorded after action
Agent identity, approver, and reasoning trace, retained together.

Conformance Level 3. Nothing less is usable here.

READ NEXT · 10

Read next.

Industry Operating Systems
The same architecture applied to four more industries.
Agent Manifest
The format every agent on this page declares itself in.
Packs
What the Pharma Pack contains, and where the company half stays private.
For enterprises
Runtime, connectors, governance and how a deployment runs.

One architecture. Twenty pages. Same document.

TALK TO US · 11

Bring one brand and one region. We will show you the rest.

A conversation, not a demo script. We ask which decisions your commercial team repeats every cycle, and which of them are still made from a spreadsheet.

Talk to us

A conversation, not a demo script. We ask what your systems of record are, what they cannot answer, and which decisions actually matter. If an Industry OS is not the right answer for you, we will say so.