There's a subtle but momentous shift on display in our February release in the rollout of "data tags". Think of them as better wiring: a more precise way to get the right source data to the right place in the dossier. Everyone's familiar at this point with GIGO (garbage in, garbage out). Well, it takes on a new form in the LLM world when it comes to document generation. In short: the better context (source data) you can provide, the better output you'll get. Sure, you can throw a mess of source docs at an LLM and hope it'll sort it out. But when you are operating at the edge of the models' capabilities and context windows, in a technical domain, every advantage matters to get output of the quality you need. Multiply the challenge by 100 when accounting for the complexity and magnitude of an IND, and you'll appreciate why we had to come forward with data tags. I've seen many folks so consumed tweaking prompts they ignore this crucial part of the equation. It was a hard won lesson for us, too, and one that AutoIND users may now benefit from.
📢 #AutoIND February Release: Data Exploration ➡️ Publishing Our February release is just the thing to shake you out of the January doldrums. We're significantly extending workflow support with Publishing v1 and Data Exploration, covering both earlier and later stages of your submission workflow. Content quality is also getting a big boost with Data Tags, offering greater precision and flexibility when mapping source data to templates. Highlights: 🔗 Add links throughout the dossier that always stay up-to-date 📄 Better first drafts and more powerful templates with Data Tags (especially in CMC/Module 3) 📊 Instantly tabulate your entire Data Room however you wish with the new Data Room Overview ❓ Ask any question of your source documents and get your answer as a summary, table, list, or letter …and more! ➡️ Review the full release here: https://lnkd.in/g8XjcfCs #WeaveWednesday #AIinPharma #AIinBiotech #AIinRegulatory #AIinnovation #regulatoryautomation
GTM @ Glean
1w100% agree Brandon. LLMs are only as good as the context they have access to. Most approaches seem to be pretty silo'd to vertical tools like Jira/Asana, CRMs, Figma, Call recorders, etc. but miss the context from the other tools. Think a horizontal AI platform, especially for a cross-functional area like Product makes a lot more sense, but comes with typical hurdles of permissions, relevancy and automation. Thoughts?