Lab automation: what to automate, and in what order

Choosing what to automate

1. Measure where work queues. Walk the process and find where samples wait, not where people complain. The constraint is frequently a manual data step, a plate sealing step or a review, rather than the pipetting everybody notices. Automating anything other than the constraint moves the queue rather than shortening it.
2. Islands before integration. Individual automated stations with a person moving plates between them are cheaper, more robust and easier to reconfigure than an integrated workcell. Integrate only where the manual transfer is genuinely the constraint, because every integration multiplies the failure modes.
3. Sample identity must survive the whole path. Barcodes, plate maps and their link to the system of record have to hold across every step. An automated process that cannot say which well came from which sample has automated the wrong thing, and this is the most common serious defect in these projects.
4. Staff for it explicitly. Automation replaces manual work with configuration, maintenance and troubleshooting work, and that work needs somebody whose job it is. Laboratories that automate without assigning the role end up with expensive instruments used manually.
5. Plan the failure modes. Decide what happens when a run fails halfway: whether samples are recoverable, whether the batch is lost, and how a partial run is recorded. Automation concentrates risk into fewer, larger events, and unplanned failure handling is where confidence is lost.

What lab automation companies actually sell

The market divides into instrument makers selling stations, integrators selling workcells assembled from several makers' instruments, and software vendors selling the scheduling layer. Understanding which of the three you are talking to explains most of what they recommend.

An integrator is the right partner for a genuine workcell and adds cost and a dependency. For islands, buying stations directly and keeping the integration human is usually better value.

The data side is half the project

Automation produces results faster than a manual process produces them, and if those results still reach the system of record by hand you have moved the bottleneck rather than removed it. Plan the data path with the physical one.

Where results support regulated work, the automated path and its records carry the same electronic records obligations as anything else, including what happens to a partial or failed run.

Automate a stable protocol, not a developing one

Automation pays where a protocol is stable, repeated and well characterised, because the cost is in programming and validating it. A protocol still being optimised will be reprogrammed every time it changes, and the automation then slows the science down.

The honest test is whether you would be willing to fix the protocol for a year. If not, automate the stable parts, such as plate filling and normalisation, and leave the developing steps manual.

With automated liquid handling systems, the integration is the project

A platform that moves liquid is the easy part. Making it useful means labware definitions that match your consumables, a scheduler, integration with a reader and a thermal cycler, error handling that does not lose a plate, and a path from the instrument's output into your data system.

Ask who does that integration, what it costs, and what happens when a consumable changes. A plate from a different supplier can require every labware definition to be re-taught, which is the hidden cost of an automated laboratory.

Deciding where the hands really go

Measure the protocol before automating it: how much time is pipetting, how much is incubation, how much is moving plates and reading them. Automation that removes five minutes of pipetting from a two-hour protocol has not repaid itself.

Where throughput is the aim, the constraint is frequently the reader or the incubator rather than the liquid handling, and adding a second reader is cheaper than a robot. Where reproducibility is the aim, the liquid handling is exactly the right thing to automate.

An automated liquid dispenser laboratory setup

An automated liquid dispenser laboratory workflow is built around the reagent rather than the robot: a bulk dispenser lays the same reagent into every well quickly, while a pipetting robot moves different volumes between positions. Dead volume, priming waste and the cleaning cycle decide the running cost. Dispense accuracy is verified gravimetrically on the actual plate type before a screen starts.

automated liquid dispensing and the errors it removes

automated liquid dispensing removes operator variation and adds its own: a blocked tip, a mis-set plate height or an unprimed line produces a pattern across the plate rather than random scatter, which is easier to spot and easier to miss. A plate map with dispense order and a uniformity check are what make a screen's edge effects interpretable rather than mysterious.

An automated liquid dispenser, bulk against selective

An automated liquid dispenser in the bulk sense uses peristaltic or syringe channels to deliver one reagent to many wells, so throughput is high and flexibility is low. Cassette based instruments swap the fluid path to change reagent without cross contamination, which is what makes them usable for more than one assay. Minimum dispense volume is the specification that bounds an assay's miniaturisation.

A lab dispenser for solvents and media

A lab dispenser mounted on a bottle delivers repeated fixed volumes of solvent, medium or buffer, and the specifications are chemical compatibility, the volume range and accuracy, and whether it can be autoclaved. Recirculation valves avoid wasting the first stroke. For anything corrosive the seal material rather than the volume range is what decides the choice.

Common questions

What should a laboratory automate first?

The step where work actually queues, which is frequently a data or review step rather than the pipetting. Automating anything else moves the constraint rather than removing it.

Should I buy an integrated workcell?

Usually not first. Individual automated stations with a person moving plates are cheaper, more robust and easier to reconfigure. Integrate only where the manual transfer is genuinely the constraint.

Why do automation projects overrun?

Because the wrong step was automated, because sample identity was not designed to survive the whole path, or because nobody was assigned to configure and maintain the system. All three are decided before purchase.

Does automation reduce staffing?

It changes it. Manual work becomes configuration, maintenance and troubleshooting work, and that needs someone whose job it is. Laboratories that skip this end up running automated instruments by hand.

When is automation worth it?

When a protocol is stable, repeated and worth fixing for a year. A protocol still being optimised will be reprogrammed constantly, and the automation then slows the work down.

What is the hidden cost of liquid handling platforms?

Integration: labware definitions matched to your consumables, scheduling, instrument integration, error handling and a path into your data system. A change of plate supplier can mean re-teaching every definition.

Browse by service class

• Compare Gene synthesis on advertised prices