In this episode of The Biotech Startups Podcast, Jon Chee sits down with Alex Telford, founder of Convoke, who spent nearly seven years in biopharma strategy consulting before what he calls "AI psychosis" set him on a different path. Alex traces the arc from biochemistry at Bristol and synthetic biology at UCL to discovering consulting existed only after graduating, joining a boutique Swiss firm two weeks before it was acquired, and learning one core lesson: clients want a recommendation to push back on, not a data dump.
🧬 The Biotech Startups Podcast is powered by Excedr—helping life science startups accelerate R&D and commercialization with founder-friendly equipment leasing. Skip the upfront costs, stay lean, and focus on breakthrough science.
As a TBSP listener, you can get exclusive perks through Excedr's partner network—special savings, promotions, and more. Explore these offers today: https://www.excedr.com/partners.
"The models are like this bunch of fractured entities — depending on how you prompt them, you can access one fractured space or another."
In this episode of The Biotech Startups Podcast, Jon Chee sits down with Alex Telford, founder of Convoke, who spent nearly seven years in biopharma strategy consulting before what he calls "AI psychosis" set him on a different path. Alex traces the arc from biochemistry at Bristol and synthetic biology at UCL to discovering consulting existed only after graduating, joining a boutique Swiss firm two weeks before it was acquired, and learning one core lesson: clients want a recommendation to push back on, not a data dump.
Key Topics Covered:
- The Slot Machine Problem: Why LLMs fail to get it right the first time — data scarcity, epistemic unknowability, and model architecture.
- AI's Good Fits in Life Science: Where LLMs work in biopharma — document analysis, regulatory submissions, clinical data structuring — and where they break down in pure discovery.
- What a Harness Is: A plain-language breakdown of AI operating environments, with Cursor and Harvey as examples.
- The Consultant's Real Job: Why effective strategy means reducing decision friction, not dumping analysis — and why LLMs can't thread that needle yet.
- Coordination Tax in Pharma: How siloed functions and unknowable drug biology make decisions stall — and what AI might actually fix.
If you enjoy The Biotech Startups Podcast, please consider subscribing, leaving a review, or sharing it with your friends. Thanks for listening.
Links:
Subscribe to the Podcast:
Apple: https://podcasts.apple.com/us/podcast/the-biotech-startups-podcast/id1679591994
Spotify: https://open.spotify.com/show/2PSL162R2s0zOwU5913Zv4
YouTube: https://www.youtube.com/@thebiotechstartupspodcast
Website: https://www.thebiotechstartupspodcast.com/
Find our guest, Alex Telford at these links:
LinkedIn: https://www.linkedin.com/in/alexander-telford/
Website: https://www.convoke.bio/
Find our host, Jon Chee, at these links:
LinkedIn: https://www.linkedin.com/in/jonchee/
Learn more about Excedr:
LinkedIn: https://www.linkedin.com/company/excedr/
Website: https://www.excedr.com
Intro & Outro Songs Created by OkKyojin, Owned by Excedr:
Website: https://flow.page/kyojin
Resources & Articles:
AlphaFold: Highly Accurate Protein Structure Prediction: https://www.nature.com/articles/s41586-021-03819-2
AI-Generated Poetry Is Indistinguishable from Human Poetry: https://www.nature.com/articles/s41598-024-76900-1
LLMs in Drug Discovery and Development — A Primer: https://pmc.ncbi.nlm.nih.gov/articles/PMC11984503/
Top Equipment Needs in AI-Driven Drug Discovery Labs: https://www.excedr.com/blog/top-equipment-needs-in-ai-driven-drug-discovery-labs
Electrification and the Productivity Lag — Why Retooling Takes Decades: https://www.weforum.org/stories/2019/04/the-latest-technology-isnt-enough-you-need-the-business-model-to-go-with-it/
PD-1/PD-L1 Checkpoint Inhibitors in Tumor Immunotherapy: https://pmc.ncbi.nlm.nih.gov/articles/PMC8440961/
Companies, Universities, & People Mentioned:
Convoke: https://www.convoke.bio/
Charles River Associates (CRA): https://www.crai.com
Cursor: https://cursor.com
Harvey AI: https://www.harvey.ai
AppLovin: https://www.applovin.com
University of Bristol: https://www.bristol.ac.uk
University College London (UCL): https://www.ucl.ac.uk
Maged Ahmed (Co-Founder, Convoke): https://www.linkedin.com/in/mageda
Vikas Velagapudi (Co-Founder, Convoke): https://www.linkedin.com/in/vikas-velagapudi
Adam Foroughi (CEO, AppLovin): https://en.wikipedia.org/wiki/Adam_Foroughi
Timestamps:
00:00 Intro
02:26 Career Pivot and the Accidental Discovery of Consulting
05:01 Growing Up Without Good Advisors
07:31 Boredom, Shortcuts, and the Slot Machine Problem
09:33 Why LLMs Get It Wrong: Data, Unknowability, and Architecture
13:55 Good Fits vs. Bad Bets for AI in Life Science
20:42 StarCraft Now, Cookie Clicker Later
24:36 What Is a Harness?
29:34 Joining C1 Associates Mid-Acquisition
34:23 First Projects: PD-L1, Lung Cancer, and Learning to Have an Opinion
38:51 Coordination Tax and What AI Might Actually Fix
44:55 Outro