The case is analyst time. Static detectors record far more files than people can check, and many are noise. AI software such as BioSonic removes noise and suggests the species, so ecologists spend their hours on the records that need judgement. Make the case with a timed trial on one of your own surveys: hours before and after, errors found when checking, and whether IT accepts where the data is kept.
We make BioSonic. This page is written to be forwarded to a director or an IT team, so it includes the limits as well as the benefits.
The summary for a director
| Question | Answer |
|---|---|
| What changes? | Recordings are uploaded in a browser and come back with noise separated and a species and confidence for each bat. An ecologist still reviews and signs off. |
| What can it save? | In the Wharton case study one night of 4,200 files went from about 2 hours 30 minutes to about 50 minutes. One team and one night, so test your own. |
| How good is the identification? | 98.7% F1 against 83.0% for Kaleidoscope Pro in the Wilder Sensing benchmark on 5,325 predictions from UK recordings. Myotis and other difficult groups still need checking. |
| How is it bought? | You buy a package of analysis gigabytes. The bigger the package, the lower the price per GB. There is no charge per seat. |
| What is the risk? | Wrong or missed identifications. Control it with the validation and manual checks in the BCT guidelines. |
The summary for IT
| Question | Answer |
|---|---|
| Supplier | Biosonic AB, Sweden, organisation number 559472-8056. |
| Installation | None. It runs in a web browser. |
| Where is the data? | Stored in the EU, mainly in Sweden. |
| Encryption | TLS in transit. Uploaded audio and off-site backups are encrypted. |
| Sign-in | Microsoft Azure SSO is supported. Two-factor sign-in is required for BioSonic staff. |
| Privacy | EU and UK GDPR. A Data Processing Agreement and security questionnaire answers are available on request. A DPA can also override the default AI training behaviour. |
| Retention | Recordings are kept for 9 years in the standard agreement. You can ask for deletion at any time. |
| Certifications | No SOC 2, ISO 27001 or independent penetration test is claimed today. They are on the roadmap. |
The full detail is on the security page and in the privacy policy.
A trial that gives a clear yes or no
- Pick a survey you have already analysed and note the hours it took.
- Agree what would be a yes before you start, for example hours saved and no loss of rare species.
- Analyse the same recordings again and time the checking.
- Compare species lists, a random sample of common species and a random sample of noise.
- Send IT the security page at the start, not at the end.
A message you can forward
I would like us to trial AI bat call analysis on one survey we have already completed. We will measure analyst hours and identification errors against our current method, and an ecologist will still verify the results. The supplier information for IT is at biosonic.io/bat-analysis-software-business-case. Can we agree on the survey and who signs off?
Common question
Does AI replace the ecologist?
No. BCT asks for an experienced analyst on every project and manual verification of the records that carry most risk. The software reduces the files that person has to open.
Sources
- Bat Surveys for Professional Ecologists, 4th edition (Collins, ed., Bat Conservation Trust, 2023, amended March 2024), section 10.2.
- BioSonic benchmarks, including the Wilder Sensing benchmark on 5,325 predictions from UK recordings.
- Wharton case study, published by BioSonic.
- BioSonic security.
Related: Security · Trust Center · How pricing works
