AI automates the plate design between a question and the bench

Between a scientific question and the first pipette tip sits a long list of decisions that have nothing to do with the biology itself: which reagent goes in which well, what dilution scheme to run, how many replicates are enough, where the controls sit, and whether the whole design fits on one plate or needs two. Individually none of these decisions is hard, and each one takes only a moment, but a scientist works through dozens of them before an experiment starts, and doing that from scratch for every run is where much of a working day actually goes. An AI lab protocol generator turns out to be well suited to exactly that part of the job: not deciding what the experiment should test, that stays a judgement call, but turning a description of the experiment into something specified down to the well.

Turning that decision into a plate is the part that moves to software. A scientist still explains an experiment the way they would to a colleague; what changes is that the explanation itself becomes the plate.

From a description to a plate map

Elnora's plate maps hold what each well contains, and the instrument's own output file binds to that map the moment a run finishes, so a set of conditions described in a conversation comes back as a table organised by well, not a blank grid waiting to be filled in by hand. The controls belong to that same table from the start: a dose-response fit weighs every treatment against the control sitting on that plate, and any well the fit set aside is named along with the reason, so another scientist can check the design without rebuilding it. Describing the conditions, generating the map and fitting the curve all happen in one conversation, the same one that goes on to write the method.

The materials list carries its own numbers

The same shift shows up once the plate has a materials list attached to it. A catalogue number in an Elnora protocol is checked against the manufacturer's own product page before the line is written, and a distributor's entry is traced to the item it actually resells, so a scientist can place an order straight from the protocol and get what the page describes. That materials list comes out of the same conversation that produced the plate, priced and ready the moment the layout is settled.

Onto the deck itself

A lab whose plates get filled by a liquid handler carries the same specification one step further. Elnora turns the finished plate design into the instructions a connected instrument actually runs, and before any of it moves, a scientist reviews a simulated pass through those instructions and signs off on it. What used to mean loading a plate from a printed layout becomes checking, after the run, that the deck carried out precisely the steps the design called for.

What none of this touches is the part of the job that was never mechanical. Whether the experiment is worth running, whether one more condition changes what the result would mean, whether a fit makes biological sense against everything else known about the system: those calls stay with the scientist. What has changed is how much of the ground between a question and a plate has to be covered by hand, and that ground used to make up most of the job.