Guide to Purchasing AI for Biopharma: An Overview

We've been having a number of conversations with big pharma execs recently about how to think about purchasing AI. It's actually fairly difficult because if you don't work in tech, it feels like every vendor is approaching you making claims that sound pretty similar. Here's a quick guide, agnostic of the product being sold:

1) Legacy Providers: There are a number of companies in biopharma that have been around since before the start of the millennium, in some form or another. Here's what you can likely count on from these companies: strong processes, security certifications, documentation, and excellent sales teams. They're also probably quite good at whatever they initially set out to do.

However, they have legacy codebases, probably have weak engineering teams (because the best engineers tend to gravitate to the hottest AI companies to build cutting edge tech - or start their own), and their claims of AI are usually really just AI layered on top of legacy architecture rather than built into the system from the ground up. Or, put differently, more often than not they're appending natural language search to a legacy codebase, which misses the benefits of an AI that can really connect the dots for you.

2) Foundation Models: This includes ChatGPT (OpenAI), Claude (Anthropic), Perplexity, Gemini, etc. These models are exceptional for generalist use cases. If you just want a tool that will uplevel everyone's writing and perform strong analyses in general, these are a solid choice. They have strong security procedures at this point and are building out their sales/customer support rapidly.

However, their analytical capabilities start to break down in highly technical domains. Biopharma is unquestionably a highly technical domain. Over time, these models will get better for various verticals (you can see them starting to move in that direction), but for specialized needs and analysis in biopharma, they're not great yet.

3) Older Tech Companies: In every industry there are companies that were started 10 or 15 years ago that were once hotshot SaaS startups and are now struggling to adapt to the new AI paradigm. These companies often have fairly good engineering talent but in order to take advantage of AI, they basically have to rebuild their company around it from the ground up to make their databases and processes AI-ready. Few execs are willing to take that leap and start over. So these companies are usually stuck somewhere between the legacy providers and the AI native companies in terms of their AI capabilities.

4) AI Native Applications: There are an increasing array of AI tools that are native to biopharma. These tend to be the true cutting edge: best-in-class AI built specifically for this vertical. If you want to understand the caliber of the tool, start by going on LinkedIn and looking at the caliber of the engineering team that built it. (Not the number of engineers - some of the best teams are actually reducing their number of engineers as AI improves - but the quality of their experience/talent.)

These companies, however, are going to feel more risky and haven't been around long enough to feel "safe". Fewer processes, fewer security certifications, less polished sales and marketing.

So, in a world where no one gets fired for buying the legacy provider, you have to decide if you want the safe but disappointing bet that won't get you fired, or the risky bet that could give you a real edge. It's not always an easy decision.

If you're thinking it through, we're always happy to discuss. And, of course, start your search with One Zyme :)