Since the release of ChatGPT, there’s been no shortage of speculation about how AI in general and ChatGPT in particular will upend, if not replace, many professions. That includes scientists working in pharma and biotech. Potential applications have been proposed, ranging from accelerating literature reviews to automating documentation. But despite all the hype, today’s potential for the existing level of AI models in biotech comes from augmenting existing tools, rather than completely replacing them (or the scientists who use them).

For context, here are a few examples that have been published about potential applications of AI in pharma/biotech:

In these examples and others, the pattern seems to be a clear tradeoff: ChatGPT and other AI models have more ready access to the technical details of biology than a human scientist could ever hope to remember. But while they have some ability to interpret and reason about these details, it’s nowhere close to the level of a professional scientist.

Of course, it’s possible that the next wave of AI will completely change the game. But at the level where today’s AI models have plateaued, they’re more like smart research assistants who remember everything they learned in class. They haven’t yet learned the nuanced reasoning, hands-on expertise, and creative problem-solving that come from years of actually doing the science. They know all the facts, but you wouldn’t want to trust them on their own.

And that’s OK, because for most bench scientists, reasoning and problem-solving are the parts of the job that they enjoy the most. They’re not looking to give up those responsibilities, and with today’s AI, there’s no reason to. Any speculation about how that will change in the future is just that—speculation.

The parts of a scientist’s job where today’s AI can do better than a human scientist are mostly the smaller tasks that sit between deep thinking and the more mundane aspects of science:

  • Translating experiment designs into formal protocols—Scientists think in natural language. AI has shown promise translating this into precise technical instructions.
  • Retrieving data based on intuitive descriptions—Manually searching through databases is slow and exhausting. AI has proven effective at retrieving data based on the scientific context.
  • Tracking inventory and consumables—AI can monitor lab supplies based on protocol details, giving scientists one less thing to worry about.
  • Helping users navigate complex technical systems—AI-powered assistants can guide users through complex technical tasks on new tools.

Leveraging AI for these tasks allows scientists to stay cognitively at the level of science rather than allowing these mundane tasks to pull them out of “the zone.”

But it’s not just about the tasks—it’s also about the tools. These capabilities are only beneficial if scientists can take advantage of them within the same interface they already use to reason about the science. These capabilities must integrate seamlessly into the workflows that scientists are already comfortable with.

Unfortunately, many of the tools that scientists use today are built on platforms that are hard to update and even harder to integrate with cutting-edge tools. These platforms are unable to incorporate new technology like AI and leave users few, if any, options to do it themselves.

That’s why Sapio built its ELN/LIMS platform with data analysis and AI in mind from day one. (Even before it was cool.) Sapio’s science-aware AI, ELaiN, allows users to leverage AI agents to drive complex R&D workflows. And thanks to the unified platform, ELaiN integrates seamlessly with Sapio’s ELN and LIMS, allowing scientists to do more work, within the context of the science, without leaving the platform.

While the world will continue to speculate about the future of AI, Sapio is making AI-enabled science that complements scientists a reality today.