The crisis of reuse: Why 65% of experiments are being repeated

The rise of the AI Lab Notebook blog series (Part 1)

This post is part of the AI lab notebook (AILN) series. In this blog, we look at experiment reuse and why ELN data findability issues can drive repeat experiments.

You can also read about shadow AI in labs and why scientists use public generative AI tools to get work done. Additionally, explore the lab AI maturity model, a roadmap from passive to active labs (URL).

Get the full research report here.

Experimental data reuse breaks down when prior results are hard to find, compare, and trust in context, and that friction can show up as avoidable repeat experiments. An AI lab notebook (AILN) is a third-generation ELN designed to embed science-aware assistance inside governed workflows so reasoning and decisions can be captured with the experimental record.

Modern science moves on interpretation

Scientific progress does not rely solely on whether results can be recorded. Instead, it depends on whether prior data can be found, understood in context, and reused quickly enough to shape the next experiment.

Many labs have modernized documentation and compliance. The harder problem now is what happens after results are documented. A second-generation electronic lab notebook (ELN) can preserve the record well. However, scientists are left stuck when they need to compare outcomes, spot anomalies, or decide what to do next.

In practical terms, “interpretation” is the day-to-day work of pulling relevant previous work, comparing outcomes across similar conditions, spotting outliers, connecting results back to methods, and making the reasoning behind a decision easy to inspect later.

The duplication tax on discovery

Our 2025 research, based on responses from 150 lab professionals across the U.S. and Europe, surfaced a costly pattern. Sixty-five percent report repeating experiments or assays because they cannot easily find or reuse previous results with confidence.

It is worth being precise about what this does and does not mean. Repetition is sometimes exactly what good science requires. Replication can reinforce or disprove a hypothesis, validate a method change, or meet regulatory expectations. The concern here is repeat work driven by friction. In these cases, prior outcomes may exist but are not findable, comparable, or trustworthy enough, quickly enough, to reuse.

Why ELNs record data but do not support experiment reuse and interpretation

The findings point to a consistent distinction between what many notebooks do well and what scientists increasingly need from them.

• 62% say their ELN enables efficient day-to-day work.
• 81% agree their ELN records data but does not help interpret it.
• 74% describe their ELN as “passive” in how it supports scientific thinking.

Together, these results suggest a familiar reality. Documentation works, but the steps that turn documented results into decisions often happen elsewhere.

One friction point is configurability. Seventy-one percent say ELNs are hard to configure, and 56% say complexity slows them down. When changes require specialist help, iteration becomes dependent on handoffs and queues.

Another friction point is tool disconnection. Fifty-one percent say they spend too much time importing and exporting data between the ELN and other tools. Manual movement increases the odds of lost context, fragile versioning, and reasoning that never fully returns to the record.

This is also visible in autonomy at the bench. Only 7% can configure assays or templates without support. Only 5% can analyze experimental data without specialist help. When routine interpretation depends on availability elsewhere, even straightforward iteration starts to stall.

The mandate from the bench

Scientists want the notebook to participate in the loop of hypothesize, design, plan, act, and analyze. They also want the notebook to do it inside governed workflows.

• 96% agree future ELNs should help interpret data.
• 99% want ELNs to act as intelligent research partners.
• 95% want conversational, text-based interactions to simplify the interface.

This is where third-generation ELNs and science-aware AI notebooks, including products like Sapio ELaiN, are being aimed. The goal is to bring interpretation and decision support closer to where work is documented and governed. That ambition is separate from the root cause of the repeat experiment statistic, which is primarily about findability and fragmentation across tools.

Before we get to maturity models, we need to address what labs are doing right now to keep moving. In the next post, we look at shadow AI in labs and why public generative AI tools have become the default workaround when people need answers quickly.

Download the full research report here

In summary

• If previous work is hard to locate or validate quickly, reuse loses to rerunning on speed alone.
• Disjointed tools fragment context, so confidence in what already exists becomes harder to establish.
• The next generation of notebooks is being shaped by two needs: better reuse of prior work and better support for interpretation once results are documented.
• An AI lab notebook (AILN) is a third-generation ELN that embeds science-aware assistance inside governed workflows. This way, reasoning and decisions can be captured with the experimental record.