Work out a lab method once and keep it for the lab

AI for biotech lab documentation should make a lab learn each method once: the analysis rule one scientist settles on a Tuesday is the rule a colleague's analysis uses on Friday, without anybody explaining it again. Giving every scientist an AI assistant does not get a lab there, because each assistant opens every conversation on an empty chat window, so the lab pays for the same method to be worked out again and again while what it learns stays with whoever happened to learn it.

A method worked out in a chat window is gone when the tab closes

Every lab has these. One scientist settles how to normalise a qPCR plate and which two reference genes to trust, another works out that the outer ring of wells on a 72-hour viability plate loses medium to evaporation and has to be dropped, and a third writes the prompt that turns a plate reader export into a clean table of fits. Each of those took an afternoon, and each now sits in one person's chat history, where it disappears when they close the tab, go on holiday or leave the lab.

The cost shows up twice. The next scientist opens an empty chat and works the same analysis out again, paying for it a second time, and sometimes lands on a different cutoff, so the lab ends up with two results from the same assay that it cannot compare.

Elnora starts from what the lab has already settled

Elnora, the AI agent for lab work, keeps what a lab works out where the next piece of work starts. Conventions that belong to the whole team, such as the reference genes, the edge-well rule or the Ct value above which a well counts as no amplification, go into a shared notes file that every member of the organisation and Elnora can read and edit, and Elnora reads it back on every turn. What it learns about one scientist's own way of working stays private to that scientist.

So a rule is taught once. A colleague who uploads a 72-hour viability plate three weeks later and asks only for IC50 values gets a fit that leaves out the edge wells, because Elnora read the group's notes before it started, and the colleague did not have to explain a thing. When a convention changes, anybody in the lab edits the note and the next analysis follows the new rule.

The steps of a method are kept the same way. A scientist works through an analysis with Elnora once, fold changes by the delta-delta Ct method against the group's two validated reference genes with technical replicates flagged when they spread by more than 0.3 cycles, and once the numbers are right they ask Elnora to keep what worked. Elnora saves it as a named skill, a written method that lists which tools the job needs and in what order, and a method the scientist runs three times is saved into their toolbox without being asked. A named protocol template does the same for the shape of a protocol, so the group's house format holds from one experiment to the next.

Every result lands where the next scientist looks

Each file Elnora writes lands in the lab's shared Knowledge Base under the folder it belongs to, on the day it is produced. When a colleague asks something, Elnora runs the whole question against the Knowledge Base and brings back the matching passages with the file each one came from, so the colleague can open the note a rule was written in before trusting the number.

Two scientists often work on the same method at once. When a teammate saves, restores or uploads a file, the file tree and the open viewer update in everybody else's workspace straight away, and when a teammate saves a protocol while Elnora is editing it, Elnora reads the newer version, checks what its own change would do to it and merges both before saving, so neither person's edit is lost.

Keeping what a lab learns also keeps the bill down, because a method that is already saved is one nobody pays to work out again. Elnora sends each task to the size of model it needs, the hard reasoning to a large model and the routine turns to a small fast one, which is where most of a lab's token spend goes.

No model arrives knowing a lab's reference genes or its edge-well rule, and a lab that writes each one down once, where Elnora reads it, does not have to settle it again.