All documents

How we use AI

Effective 5 November 2026

RouteZero Ltd · 71-75 Shelton Street, London, England, WC2H 9JQ · Registered in England and Wales · Company no. 14190222
Contents

This page explains how RouteZero Ltd ("RouteZero", "we", "us") uses AI with the travel data you share with us, and the protections that apply when we do.

We publish it so that you, and your security and data protection teams, can see where AI fits into our work before you decide whether to share data with us. It sets out what AI does for you, what it may see, what it doesn't do, and the safeguards around it.

1. Who this applies to

This page applies to any organisation ("you") that shares travel data with us under an agreement for our services.

It covers the travel data you share with us under that agreement, such as the exports of your booking providers, expense systems, and cards. Our privacy policy covers other personal data we hold, such as your contact details and visits to our website.

2. What we mean by AI

The terms below have the following meanings on this page:

  • AI means large language models, a type of software that reads and writes text, such as Anthropic's Claude.
  • Our AI provider means Anthropic, the company that provides the AI models we use.
  • Personal details means information in your data that identifies a person, such as a traveller's name or email address.
  • A pseudonym is a stand-in value that replaces a personal detail. The same detail gets the same pseudonym, so records about one person can still be grouped without showing who they are.

3. How AI helps you

AI lets our team do more careful work in less time. You benefit in four ways.

More accurate results

Travel data arrives in many formats, and each one has to be read correctly for your figures to be right.

AI lets us check far more of your bookings, and each booking in more detail, than our team could by hand. As a result, we find and fix more processing errors before your results reach you.

Results sooner

Before we can report on your data, we need to work out what each of your files holds and how to process it correctly. AI helps us set up and check your data faster, so you receive your results in a shorter engagement.

Lower cost

AI reduces the time our team spends on your data, and we pass that saving on to you in our fees.

No need to change your exports

Travel data formats change over time. A booking provider may update its export, or you may move to a new booking tool or card provider. Keeping to an old format for us can be awkward, and it can mean losing detail that the newer format holds.

Our software is designed to read data in many formats. AI helps us adapt to a new format quickly, so you don't need to change how you export your data for us.

4. What we use AI for

We use AI to help us set up how we process your data, and to check a sample of the results, booking by booking. We do this while we import your data.

When we set up your data, AI helps us work out what each of your files holds. It then helps us configure our software for your files, and extend it where a file holds something our software doesn't yet read.

When we check the results, AI compares a sample of processed bookings with the records they came from. It flags anything that doesn't match, which helps us find and fix processing errors.

5. What AI may see

To check that a booking was processed correctly, AI needs the details of what was booked, but not who travelled.

AI may therefore see the details of a booking that describe the travel itself, such as dates, routes, the suppliers used, and amounts.

Travellers' names and other personal details are removed, replaced with pseudonyms, or reduced to a part that doesn't identify the person, before AI sees them. Card numbers are removed altogether. Free-text notes are removed, or scrubbed of email addresses, phone numbers, and card numbers.

The example below shows what this looks like for a single booking.

DetailIn your dataWhat AI sees
TravellerErin Hollowayvalue_3f9a1c07b2e4d856Pseudonym
Emailerin.holloway@outlook.comvalue_8c21d4e0a97b3f15Pseudonym
Address14 Mill Lane, Leeds LS1 2ABMill Lane, LeedsReduced
Card number4111 1111 1111 1111Removed
NotesTeam offsite, Erin's 40thRemoved
RouteLondon Heathrow to AmsterdamLondon Heathrow to AmsterdamKept
Travel date12 March 202612 March 2026Kept
Amount£268.42£268.42Kept
An example with made-up values

We set what AI may see for each type of detail your files hold, and we can tailor this with you during our engagement.

6. What AI doesn't do

AI helps us set up and check how our software processes your data. It doesn't produce your results.

The figures we report are produced by our software, not by an AI model.

The only AI provider we use with your data is Anthropic.

Only RouteZero staff use AI tools with your data.

7. Three layers of protection

Your data is protected by three layers, each working on its own. The first, our contract with our AI provider, covers everything our AI provider receives. The second and third reduce what it receives in the first place. By the time any data reaches AI, there is as little of it as possible, and its personal details have been removed, replaced, or reduced.

1. A contract that protects your data
Covers everything our AI provider receives
2. AI sees only what it needs
The least data each task needs
3. Personal details removed first
Removed, replaced, or reduced before AI sees them
Your data

Layer 1: A contract that protects your data

Every use of AI with your data happens under a contract that sets out what our AI provider may do with that data, how long it may keep it, and how it must protect it.

We use Anthropic's Claude models on a Team plan under Anthropic's Commercial Terms. Under those terms, Anthropic acts as our processor, which means it processes your data on our behalf and only on our documented instructions.

Anthropic doesn't train its models on your data. It deletes the data we send it within 30 days, unless it needs to keep data longer to enforce its usage policy or to comply with the law.

Anthropic's security is independently certified. It holds ISO 27001:2022 certification for information security and ISO/IEC 42001:2023 certification for AI management systems, along with SOC 2 Type I and Type II reports. Data is encrypted in transit between our devices and Anthropic.

Layer 2: AI sees only what it needs

The less of your data AI sees, the less there is to protect.

AI sees only as much of your data as a task needs, and very often a summary rather than your records. Our tools also limit how much data any one request can return, so a single request can't bring back a large part of your data.

Layer 3: Personal details removed first

We also remove or replace personal details in the data AI does see, as the example in section 5 shows.

Our AI tools reach your data only through our own software, which removes or replaces personal details first.

We create each pseudonym with a cryptographically secure method that uses a secret key we keep. Without that key, nobody can work out which value a pseudonym stands for, even by trying likely values one by one. Because the same detail gets the same pseudonym, AI can still tell when two bookings belong to the same traveller, without ever learning who that traveller is.

8. People stay in charge

AI supports our team's work. Our team keeps control of it at three points.

A team member approves what AI may see before any AI reads your data.

A team member reviews what AI finds before we change how your data is processed.

When we set up your data, we check our figures against the totals in your own data, and a team member reviews the results.

9. Where your data is kept

The primary copy of your data is stored in the UK, in Microsoft SharePoint. We process it on company devices in the UK. Identifiable customer data is never stored in our code.

When we use AI, Anthropic receives only the data a task needs, after personal details are removed or replaced. Anthropic stores that data in the United States, and deletes it as section 7 describes.

The EU Standard Contractual Clauses and the UK International Data Transfer Addendum, in Anthropic's Data Processing Addendum, cover this transfer. Data whose personal details have been replaced with pseudonyms is still personal data under UK data protection law, so these safeguards apply to it. They are standard safeguards the UK GDPR recognises for sending personal data outside the UK. Our sub-processor list sets out where each company we use stores data, and the safeguards for each transfer.

10. Questions

You can contact us about how we use AI with your data at hello@routezero.com.

We can share more detail during your security review.