A panel of patent attorneys, in-house counsel, and a patent agent discussed how AI now handles chemical structures, biological sequences, and messy inventor data in life sciences patent work. Solve demoed features for genus validation, CDR extraction, and sequence listings. The panel also flagged where the technology still falls short and where human review remains essential.
Life sciences has long been considered one of the hardest areas of patent practice for AI to serve well. Chemical structures, biological sequences, and Markush claims don't behave like ordinary text, and the tools that work for mechanical or software patents tend to fall short the moment a chemist or molecular biologist gets involved.
We partnered with IPWatchdog to bring together a panel of in-house counsel, private practice attorneys, and patent agents actively using AI in their day-to-day life sciences practice to talk candidly about where these tools stand today, where they fall short, and where the profession is headed next.
As Gene Quinn put it in his opening remarks: “the professionals in this space who have particularized needs are now starting to see those needs met, and the pressure to adopt is coming from clients, CEOs, and CFOs alike.”
We were joined by:
The panel agreed early on that life sciences patent work simply doesn't tolerate the sloppiness that AI drafting tools were originally built to handle. James Whittle noted that in this field, attorneys file fewer applications than in other industries, and each one carries more weight. That leaves little room for error.
Three technical obstacles came up repeatedly:
Erin Hill pointed out that most tools historically treated chemical structures as static images rather than something attorneys and inventors could actually manipulate. Comparing Markush genera, checking whether a list of species falls within a proposed genus, and identifying where white space remains for a narrower claim all require more than a picture. Erin also described the strategic problem this creates.
Biotech and pharma face pressure to file before the development candidate is settled, so a backup compound may later move forward that sits in a thinly described corner of the genus. A narrow claim to that species may still be available, but capturing meaningful space around it is much harder. This is where interrogating the compound list against the claimed genus earns its keep.
Rick Timmer noted that sequences are textual and therefore easier for AI to process than structures. He added that the real value lies in handling the complexity of large, multi-part applications, freeing attorneys to focus on strategy rather than manual bookkeeping.
Sequences and structures rarely turn up in a clean, filing-ready table. They arrive spread across PowerPoint slides, Word documents, and spreadsheets, often carrying modifications, and normalizing them has traditionally fallen to the attorney or a vendor. Gene Quinn flagged sequence and data handling in his opening as one of the pain points he hears about most.
For more information on data security at Solve, click here.
For chemical structures, Solve Intelligence can auto-generate a genus from a set of uploaded species, flag any species that fall outside it, and export a genus coverage validation spreadsheet. This gave the panel a concrete way to move from a static drawing to an actual claim strategy. Rick Timmer added that doing this inside one interface also makes it easier to check consistency in R-groups and terminology across a full application.
For sequences, James Whittle called out CDR extraction as one of the clearest time-savers Solve Intelligence offers for antibody work:
“Particularly antibodies, it really has an incredible tool to identify the CDRs, which we used to always have to ask the inventors to do.”
He added that this removes a real hassle from client relations. The same tooling extends to consensus sequence generation and to producing WIPO-compliant ST.26 sequence listings, both shown live in the demo.
James also described a feature that pulls file history across jurisdictions and compares claim amendments and arguments made in each one, flagging any place where a position taken in one country might contradict what's pending in another. Dennis Parad separately named this same portfolio-wide context as one of the biggest improvements he's seen in Solve Intelligence.
Beyond drafting, the panel called out several other uses:
Rick noted that Solve Intelligence's capabilities can be scaled to experience level, with certain features toggled on or off so less experienced attorneys can use the tool without getting lost. He sees this as a useful training aid as well as a production tool. Dennis put it simply: the platform saves time.
James raised a concern worth sitting with. AI-assisted drafts now look superficially perfect, which removes an old signal of how carefully something was drafted, like typos or inconsistent structure. He'd like to see better tools for flagging what's AI-generated versus human-reviewed. It's a good reminder that this panel isn't treating AI as a black box. They're actively deciding when to trust it and when to override it.
What stood out across the conversation is that the practitioners on this panel aren't outsourcing their judgment to AI. They're actively shaping how it's used, deciding when to trust it, when to override it, and when to keep it out of the room entirely. That discipline, more than any single feature, is what's letting life sciences practices start to realize the same gains that mechanical and electrical practices have enjoyed for years.
As Gene Quinn put it in closing:
“I know a lot of you have been waiting for these types of tools to be deployable in your space, and that day is here now.”
Book a demo to see Solve Intelligence's life sciences workflows on your own applications.
