Harmonized
Common standards give concepts a consistent meaning across sources, so downstream analysis can work with the clinical information rather than reconcile its formatting.
THOUGHTS FROM OUR CEO AI STRATEGY / REAL-WORLD DATA
Your AI advantage starts with where you invest.
AI can help clean the data, connect the infrastructure, and build the intelligence. The strategic question for pharma and biotech: which work should your team make its own?
A perspective from Sundeep BhanCEO, Prognos Health
Hear Sundeep on where organizations can focus their AI investment.
AI can work across the entire stack.
Your investment still needs a strategy.
Your investment still needs a strategy.
Building every layer yourself can mean repeating work that other teams are already doing.
The question behind the tokens
What deserves your team's
time, attention, and AI resources?
In Sundeep's view, the decision goes beyond the cost of running a model. It includes the people designing the workflows, resolving data inconsistencies, and interpreting the results.
For many real-world data programs, the early work is similar across companies. The opportunity is to build on a trusted foundation and focus your own resources on decisions, applications, and insights that are distinctive to your business.
Compute
The AI resources behind each workflow.
Time
The effort between acquiring data and using it.
Attention
The expertise your people bring to the problem.
A framework for AI investment
Three layers.
Different places to create value.
Select a layer to see the work involved—and what it means for your team.
The data layer
Give different sources a common meaning.
Assemble records, resolve differences in naming and structure, and add the context needed to use the data consistently. Even individually clean sources still need to work together.
WHY IT MATTERS
A test can look different across laboratories. A common interpretation helps your analysis follow the clinical concept, rather than the source's formatting.
Why the foundation matters
The same clinical concept
can arrive in different forms.
Laboratories can use different test names, codes, units, and free-text conventions. Without the right context, an AI workflow may treat equivalent concepts as unrelated—or miss a distinction that matters.
Harmonization is more than moving records into one place. It brings clinical meaning into a structure that downstream teams and applications can interpret.
ILLUSTRATIVE SOURCE VARIATION
Source A
Standard test codeCODESource B
Internal abbreviationNAMESource C
Free-text descriptionTEXTClinical interpretation + harmonization
A common concept with the context to use it
A common concept with the context to use it
Conceptual illustration. Clinical equivalence requires interpretation; similar-looking records are not automatically interchangeable.
Beyond a clean file
What should AI-ready
actually mean?
Sundeep proposes four practical criteria. Together, they describe a foundation that gives an AI application more than a set of values to process.
Hear it from Sundeep
Start with what
AI-ready really means.
Before building the next application, consider the data foundation beneath it. Watch Sundeep discuss why defining AI-ready data matters—then explore the four criteria below.
Complete and linked
Address available gaps and connect records at the patient level where possible, supporting a more coherent view across the participating data.
Self-describing
Make field definitions, derivations, and limitations understandable, so applications have the context needed to interpret what the data represents.
Accessible to agents
Provide appropriate connectivity for AI applications to query the data and its context. The article highlights MCP as one approach to that connection.
AI readiness requires both usable data and the context to interpret it.
Where to build your advantage
Choose what to build.
Be deliberate about what to reuse.
The case for shared foundations is strongest when multiple companies are solving similar problems with the same underlying sources. That creates room to focus internal investment on the intelligence above the foundation.
A foundation you can build on
Data and infrastructure
- Source assembly and harmonization
- Patient-level linkage and enrichment
- Definitions, context, and access
Evaluate partners on transparency and auditability, as well as usability.
Where your business can stand apart
Proprietary intelligence
- How your team prioritizes engagement
- How alerts inform field activity
- How insights guide commercial choices
Apply your strategy, expertise, and workflows to the questions that matter to your business.
There are valid reasons to own more of the stack. Sundeep notes that control, audit requirements, and unique proprietary data can justify an internal approach. The choice should follow the value of the work to your organization, rather than an assumption that every layer needs to be rebuilt.
Which work creates something distinctive for your business—and which work repeats an industry-wide effort?
Spend them where they make you different.Read “Where Are You Spending Your Tokens in RWD?” on LinkedIn ↗
For pharma and biotech teams
Put the foundation to work
on your commercial questions.
A connected clinical foundation supports the work above it. The intelligence comes from how your organization applies it to its own priorities.
Engagement strategy
Which HCPs deserve a closer look?
Bring clinical signals and provider context into the way your team evaluates engagement opportunities.
Field execution
How should your team act on an alert?
Develop the rules, workflows, and follow-through that connect a relevant event to your commercial strategy.
Resource decisions
Where should your next investment go?
Use your market understanding and business objectives to decide which analytical applications to build next.
A foundation built around clinical laboratory data.
Explore with Prognos →Prognos harmonizes and enriches laboratory data, connects available patient and provider context, and helps teams create a usable foundation for analytics and commercial applications.
Adapted from Sundeep Bhan's perspective on AI investment in real-world data, published October 14, 2026. Read the original article for his full argument.
Let's continue the conversation
Where is your organization
investing its AI effort?
Tell us what your team is building—and where data preparation, context, or connectivity is slowing the work down. Let's explore the foundation that can support your next step.