A buyer's decision guide — what each approach costs you, what it's best for, and an honest "when not to use us."
Fixed automation repeats one hard-coded protocol fast but can't change. Flexible automation can be reconfigured for several workflows but still needs programming and integration work. Plain-language (physical-AI) automation runs protocols you describe in natural language on your existing instruments. Choose fixed for one ultra-high-volume assay, flexible for a few stable workflows with engineering support, and plain-language for a changing, high-mix bench.
| Fixed | Flexible | Plain-language | |
|---|---|---|---|
| How you set it up | Hard-coded for one protocol | Scripted / reconfigured per workflow | Describe the protocol in natural language |
| Cost to change | Very high (re-engineer) | Moderate (reprogram + re-integrate) | Low (re-describe) |
| Who runs it | Automation engineer | Trained operator / engineer | Any scientist |
| Throughput | Highest for its one task | High | High, across a changing mix |
| Works with existing instruments | Usually purpose-built | Sometimes | Designed to |
| Best for | One unchanging, huge-volume assay | A few stable workflows + engineering support | A high-mix bench that changes often |
If you run a single assay at industrial volume and it will not change for years — think a dedicated production or screening line — fixed automation's raw speed and reliability are hard to beat, and the rigidity doesn't matter because nothing changes.
If you have a handful of stable, well-characterised workflows and an automation engineer (or vendor support) to program and maintain them, flexible automation gives you reconfigurability without giving up much throughput.
If your bench is high-mix — protocols change, samples vary, and the people closest to the science aren't programmers — plain-language physical AI fits the way the lab actually works. You describe the run, it adapts to your bench and instruments, and changing the protocol is a sentence, not a project.
Plain-language automation is the wrong choice if you have one fixed ultra-high-throughput assay where every second of cycle time matters and nothing will ever change — a dedicated fixed line will out-run it. It's also not a fit while a brand-new protocol is still being invented and changes every single run; stabilise it manually first, then automate.
