We started building our own AI agents because the legal AI products we tried were not suited for pharma and biotech IP work. We have developed our own AI tools that we use ourselves, day to day, and which solve problems that we and our clients actually have.
AI agents, written by our own attorneys
interconnected skills & modules
months of testing by senior attorneys before anything touched client work
of Evolve attorneys, paralegals and implementors using AI
Basis checks, data mapping, office actions, portfolio reviews. Our attorneys do all of this themselves, as they always have. The difference is that our AI agents are able to efficiently gather the information for attorney review. We can look at more, in more depth, and earlier in the process.
The tools collect and prepare. Our patent attorneys still make the decisions: what does the data supports, how to strategically amend or argue against an objection, and how to align the IP and commercial strategy.
Our AI tools are designed to help us provide advice based on deep-analysis early on, when it can still be acted on. This means before full due diligence, opposition or litigation proceedings.
Our AI tools use frontier models, are built by our own attorneys and implementors, and are pharma/biotech specific.
Every AI patent tool on the market runs on models from the frontier LLM labs. We believe that putting someone else’s software between the attorney and the model adds cost and restriction, and very little else.
An off-the-shelf tool ties you to the developer’s idea of how patent work should go, and it stops your attorneys picking up the one skill that now matters, which is instructing the model well. We write our own instructions and workflows, and the case law and drafting rules inside them are ours.
Ask an LLM to argue from what it happens to remember and it will make things up, including citations and data. Our tools only see what the attorney gives them: the specification, the prior art, the prosecution history, the lead compound or sequence, the strategy and our instructions.
A sequence listing has to match the data exactly, and a language model is the wrong tool for that. So sequence listings, compound extraction and Markush interpretation are done in code. We keep the language models for work that use language.
Our duty of confidentiality applies to AI exactly as it does to email or the cloud. We build directly on the frontier labs under enterprise terms. Nothing is kept, nothing is used for training, and there is no third party in the middle whose arrangements we would also have to vet.
The patent attorney steers each step and can see the working, so there is no black box behind the advice. We test every tool against the latest decisions before we rely on it, and when the law or the model moves we update it ourselves, usually within days, rather than waiting on a vendor.
We proudly consider ourselves experts in pharma IP:
“As a new member of the team who has not previously used AI systems in my day-to-day work, I have been thoroughly impressed by the sophistication of Evolve’s AI tools. They are not just useful for sense-checking finished pieces of work but also help streamline the preparation of my work.”
“Using the AI tools has made me much more efficient, along with reassuring me that the judgement calls I am making are right. I love being able to see a bottleneck or an area that can be improved in our workflows and work with the AI to find a solution.”
Prosecution. Translates and summarises the office action and sets up the groundwork for our response.
Validity and opposition. Novelty, inventive step, support, added matter, priority and data strength, laid out for us to review.
Drafting. Claim sets, amendments, specification sections and ST.26 sequence listings, prepared to our instructions.
Amendments. Claim amendments and basis tables for EP, US and other jurisdictions.
Quality. A second read of specifications, sequences, compounds and responses, with each issue graded.
Invention spotting. Finds inventions and disclosure risks in draft publications and competitor filings.
Due diligence. Gathers and organises the evidence on coverage, validity, ownership and contract terms.
Patent estate and loss-of-exclusivity charts that show how long the protection runs.
In pharma, validity nearly always comes down to the data. Albert is a set of modules, one for each question we tend to ask of a patent, whether it belongs to a competitor, a licensing target or one of our own clients. It prepares:
We use Albert for competitive intelligence, freedom to operate, oppositions and diligence. Albert allows us to gets us to a strategic view quickly, and we go deeper only where a particular risk warrants it. More on Albert and Albert Construe.
A competitor’s 300-page patent arrives. Albert pulls out each embodiment with its data and its claims. We then do the analysis we would always have done, only sooner: is this patent a real threat, where is it weak, and is a full FTO opinion worth the money.
The arguments that persuade the EPO are not the ones that work at the USPTO or the JPO. Popper carries a knowledge bank we wrote ourselves on the evidence needed for clinical-use inventions, sequence claims for biologics, comparative data, and the habits of each office. When an office action comes in, Popper prepares:
The attorney then decides whether to push on or slow down, narrow or broaden, and how to keep the family consistent across countries. The name comes from Sir Karl Popper, who insisted that a scientific theory must be falsifiable. It seemed a good fit for an agent whose job is science-backed rebuttals.
An office action arrives in Japanese. Before the attorney has opened it, Popper has translated and summarised it, pulled out the cited sequences, and set the objections against the file history, your strategy and the Japanese case law. The attorney starts with the whole picture in front of them.
Drafting a pharma application means working out what the data will support before you decide what to claim. Darwin is built around that order of events, for small molecules, biologics and cell and gene therapy. It prepares:
Gilbert then reads the whole application through. The attorney settles what the data supports and what the business needs the patent to do. See how we work with start-ups and established pharma, or read how to read a biotech patent.
A new set of results needs protecting. Darwin drafts the claims with the attorney one decision at a time, then the specification, then the sequence listing. Gilbert reads the whole thing for inconsistencies and the attorney sorts them out before filing.
IP sits at the heart of most biotech deals, and the diligence nearly always arrives with too little time attached. Cassandra keeps a single record of every fact, where it came from and how far it can be trusted, which leaves our attorneys free to spend the time on judgement rather than bookkeeping.
Cassandra also reads the IP terms in commercial agreements and flags what the attorney should look at.
A deal is on the way. Cassandra goes through the information with us, logging each fact with its source and building the family table, so our attorneys can test coverage, validity and ownership claim by claim and have the risks ranked before signing.
Patent documents are long, repetitive and unforgiving. One inconsistency between the claims and the description, or a sequence that does not quite match the lead, can sink a patent years later. Gilbert is our second or third pair of eyes before anything goes out:
Each issue is graded critical, high, medium or low and the attorney decides what to change. The result is fewer formalities objections, fewer unnecessary claim fees, and the odd sequence error caught before it reaches the register. Related reading: the unforgiving gold standard.
A response is ready to file. Gilbert reads it against the law, the science, your instructions and the attorney’s notes, and grades each issue from critical down to low. The attorney decides what to change.
In a pharma or biotech company, inventions tend to be made faster than anyone notices them, and the next abstract or manuscript is often the real filing deadline. Franklin is built for that gap between the lab and the patent department. It prepares:
The attorney decides what to file, when, and what the paper should say. Related reading: IP strategy for cell therapy FAQs, and how we build invention capture into Fractional IP counsel.
A draft paper lands two weeks before submission. Franklin sets it against your pending cases and earlier disclosures, and the attorney reviews it with the risks already marked. The paper goes out on time. So does the filing.
A spreadsheet or a complex array of arrows and dotted lines will not tell your board or an investor a clear story about your patents strategy. We build highly customizable and adaptable ammonites to help us tell the story about the IP that we want to tell. We’ve found that once you start using the ammonite, you can’t go back. We use it for:
The board wants to understand the status of the patent portfolio before launch and agreeing to further investment in new indications for the product. They want to know: What patents do you have? When do they expire? What do they cover? When is Loss of exclusivity? How much patent term is left for each of the cases? What are the life cycle management versus core product filings? Are the patent pending or granted? How does patent expiry compare to regulatory exclusivity protection? All of this an more is communicated clearly on the ammonite.
We are a regulated practice, bound by duties of confidentiality, competence and care, and those duties do not stop at the AI. In practice that means:
Good attorneys, but the AI is either absent or an off-the-shelf product bolted on afterwards. How much analysis you get is limited by how many attorney hours you can afford.
Limited capabilities in pharma and biotech, slow to adapt to changes in practice, each new tools requires you to upskill your team on how to use the software.
Pharma patent attorneys who develop their own tools and use them daily. The same analysis we have always done, started earlier and finished faster, with a qualified attorney signing the advice and a regulated firm standing behind it.
No. The tools are how we do our work, not something we sell. Every client benefits from them on their own cases, including Fractional IP clients.
No. Every AI patent product on the market sits on a model from an LLM lab, and a wrapper simply puts a third party between you and that model, with its own confidentiality terms, its own roadmap and its own price. We build directly on the frontier models under enterprise terms, nothing you send is kept or used for training, and our attorneys write the instructions and workflows themselves.
A qualified patent attorney, every time. The tools do the groundwork. The analysis and the advice are the attorney’s, and you get the same professional duty of care as on any other piece of our work.
Our tools can draft claims and specification sections quickly, and we use them for exactly that. What they cannot do is decide what the data supports, what an examiner will accept or what the business needs from the patent. That is the attorney’s job, and every draft goes through one before it goes anywhere.
No. Evolve Intellectual Property is a firm of UK and European patent attorneys in Cambridge working only in pharmaceuticals and biotechnology. The US company called Evolve IP has nothing to do with us.
We also write an AI in the patent industry newsletter, with practical notes on workflows, prompting and the tools we use. Subscribe, or get in touch and we will add you to the list.