·
3 minute read
BD&L teams lose time in the assembly phase, not the decision phase. At Fierce Biotech 2026, Ferma CEO Sriram Subramanian walked through a live BD&L evaluation end to end - opportunity screening, valuation modeling, and landscape analysis to show what changes when Agentic Workflows handle the assembly.
The full session recording is now available.
The Problem: The Last Mile Is Always Manual
Take a canonical opportunity screening workflow for a mid-to-large pharma BD team. The work involves pulling data from multiple sources, consolidating into Excel, filtering row by row, and writing a screening memo. A two-person analyst team takes approximately one week per screening pass. And the moment the memo is written, new data has already come in.
The tools provide raw ingredients. The synthesis is always manual. That's the bottleneck.
Conversational AI vs Agentic AI: What's the Difference?
The session drew a direct distinction that most BD teams haven't fully internalized. Most teams are already using conversational AI - ChatGPT, Claude, Gemini, where you ask a question and get an answer. Agentic AI takes actions on your behalf.
In BD&L, that means:
- Collecting data across clinical trials, conference presentations, filings, and pipeline sources
- Reasoning through that data within the structure of a specific BD workflow
- Producing a finished deliverable - not search results, but something you can bring to a licensing committee
Ferma's Agentic Workflows are built on this principle, combining a life sciences data foundation with workflow logic shaped by BD&L practitioners from the buy side, sell side, and investment side. As Sriram noted in the session: AI is only as good as the data accessible to it. Ferma's data includes sources that go beyond what standard platforms carry, including native-language China pipeline data, and is refreshed on a near-real-time basis.
The Live Demo: Three Agentic Workflows
1. Opportunity Screener Agent
The demo opened with a live screener build: ADCs targeting a specific target, filtering on reported ORR greater than 20%, favorable safety profile, and global rights availability for partnership. The agent generated a multi-step execution plan and ran it - pulling efficacy data, safety data, deal information, and catalyst events in parallel.
Starting from 41 ADC programs, the workflow filtered down to 5 targets through a four-step funnel:
- Scientific data must be available
- Reported ORR greater than 20%
- Favorable safety profile
- Global rights available for partnership
Four of the five shortlisted assets were China-based, surfaced because Ferma's embedded China team indexes native-language pipeline data. Each asset linked directly to the underlying conference data, including the latest readout from ESMO 2025.
The same scope a two-person analyst team would spend a week on ran in minutes. The output stays live as new data comes in.
2. rNPV Valuation Model
The second workflow demonstrated a real-time, risk-adjusted NPV model. The demo scenario: an asset with a Phase 3 readout expected in 2026. Inputs were sourced directly from Ferma's database rather than entered manually. The output was a parameterizable model with bear, base, and bull scenarios, with a projected 2029 launch and peak share by 2035 in the demo scenario and all parameters adjustable in real time.
The practical value: BD teams can run an interactive model into an investment committee discussion without a separate financial modeling build. For smaller companies without dedicated modeling teams, this type of analysis becomes accessible on demand.
3. Competitive Landscape and White Space Analysis
The third workflow mapped the GLP-1 landscape by mechanism class and indication cluster. The output showed where crowding exists - obesity and type 2 diabetes, and where activity is sparse, including the neurodegenerative side. The view was also configurable by competing company, enabling a side-by-side look at how specific organizations are positioning across the indication space.
What Changes When Assembly Is Automated
Agentic Workflows don't replace BD judgment. They remove the assembly work that delays it.
The three workflows demonstrated at Fierce Biotech 2026 - screening, valuation, and landscape mapping cover the first phase of any deal process. Each one addresses a distinct bottleneck:
- Opportunity Screener - eliminates the week-long manual screening pass, keeps the output live
- rNPV Valuation Model - removes the dependency on a separate modeling team for a first-pass valuation
- Competitive Landscape - surfaces white space and competitive positioning without a manual compilation cycle
When this phase runs in minutes and the output stays current, BD teams spend their time on conviction rather than data collection.
Watch the Full Session
The 17-minute session covers all three live workflow demos and the audience Q&A.
Watch: AI Agents for BD&L Workflows - Fierce Biotech 2026
See It on Your Indication
The Fierce Biotech demo ran on ADCs and GLP-1. If you want to see the Opportunity Screener or Competitive Landscape workflows running on your own therapeutic area, request a Ferma demo.
Ferma is the AI-powered intelligence platform for life sciences from ZoomRx, trusted by 100+ leading biopharma companies across the world.
