An insurer has rolled out Microsoft 365 Copilot. A claims handler does not want to type up a forty-minute client call, so she uses the free transcription service she also uses at home. An underwriter asks ChatGPT to summarise a valuation report. In IT's records, Copilot is "the company's AI tool".
That is how it goes at almost every firm that has provided one AI tool. The other tools have not gone away; they run alongside it, mostly on free accounts, and none of them has been decided on. That is shadow AI: self-service, without a decision and actually in use. In financial services it matters more, for three reasons, and you can track it without monitoring anyone.
What runs alongside the approved environment?
The list differs from firm to firm, but some tools turn up almost everywhere. They are there because, for the task at hand, the approved tool is not the obvious choice. A separate article looks at why employees use other AI tools when the firm already has Copilot.
| What gets opened | Why people choose it | What the firm needs to know about it |
|---|---|---|
| ChatGPT on a free account | Writing up a meeting note, rewriting a letter, answering a complaint | Whether input is used for training, and whether an IBAN was in the text |
| Perplexity or another AI search engine | Looking up tax rules or product terms, with sources | Free or business account, and what the query said about the client |
| Transcription of client conversations | Not having to type up an hour-long advice meeting | Where the recording is processed, and that the whole conversation was about the client |
| Tools that read documents | Reading payslips and valuation reports from an application file | Whether there is a data processing agreement, and where the document goes |
| AI features in existing software | The advice software or policy system gained an assistant in an update | No new domain, no purchasing decision, but a new processing operation |
| Browser extensions | Summarising an email or checking a text, right in the tab | Which read permissions the extension has in the browser |
The last two rows are the hardest: nobody requested anything new, yet text now goes to a model it did not go to last year. A separate article explains why an extension with read permissions is an AI tool too.
Why does it matter more in this sector?
The first reason is the content. A meeting note contains a client's income and two IBANs. In this sector, client data is the everyday material of the working day. A free account comes with no data processing agreement and input may be used for training; our comparison of free and business AI accounts sets out the differences between account types. The Dutch Data Protection Authority reported several data breach notifications in which employees had put personal data into an AI chatbot.
The second reason is regulation. Regulators such as the AFM and DNB in the Netherlands look for sound and controlled business operations, and careful handling of client data is part of that. An AI service nobody has decided on, and that nobody knows is in use, does not fit the word "controlled". AI in financial services covers what a regulator will then want to see.
The third reason is DORA. The EU's Digital Operational Resilience Act has applied since January 2025 and requires financial entities to manage ICT risk and the ICT services they obtain from third parties, with a register of information on those arrangements. Exactly which obligations apply depends on the type and size of the entity. The direction is clear, though: an AI service employees use every day is a third-party ICT service, and if nobody knows it is in use, it is not in any register.
Why does a survey fall behind?
A mortgage advisory firm with forty advisers has allowed Copilot. In February, the IT lead surveys every team and finds, alongside Copilot: ChatGPT on free accounts, Perplexity, a transcription service for client conversations, a tool that reads application files and the summarising feature the advice software gained in its latest update. Two of them do not appear in the register at all. That is not an incident. It shows that the register is not keeping up with the working day.
By May, the February list is already out of date. The team using the transcription service has switched to one that works better with Teams. And the portal of an insurer the firm works with now has an AI assistant. Nobody did anything wrong. The survey described February.
Every survey is a snapshot; our article on why tool catalogues go out of date explains the mechanisms behind that. Our guide to discovering which AI tools employees really use compares the methods and what each one misses. An advisory firm has one more problem: the proxy logs see the office, not the adviser's laptop at the client's kitchen table.
How do you keep track without monitoring employees?
By recording only what the decision needs. The question is which AI services are opened in the organisation, how often and whether a decision has been made. The service's domain is enough to answer that. Who opened it, which page and what was on it do not change the decision, but they do change the measurement, from a count of services into a record of people. AI monitoring versus employee monitoring explains that distinction and what it means for the GDPR and the works council.
That is why the BeeSensible extension records only the registrable domain, at most once a day per device, with no user ID, no path and no page content. The domains it observes are matched against a catalogue of over 1,000 AI tools, listing the vendor, country, training policy and a risk score for each. In the Inbox, an admin sees which tools have no decision yet and records one; BeeSensible never makes that choice itself.
