Your data, Your expertise, You’re in control.
Tripcerto sits between your content, your systems and the model, so your business can put its own data to work with AI. It decides what the model may read, recommend and do, records what the model read and did, keeps each business’s data apart and leaves the commercial decisions with your experts.
How information moves through the system
Tripcerto reads your content, your product data and your customer records to run the workflow you configured. The main paths, and how long each record is kept, are set out below; the full list of suppliers is available on request. None of it is used to train a model, by us or by our suppliers.
Your content and product data
Drawn on, never trained on
The documents and product records you provide are indexed so a conversation can draw on them. Each product or place a visitor is shown as a recommendation comes from matching against the catalogue set for your account.
The conversation
Written by Anthropic’s model, which keeps nothing
Anthropic runs the model that writes each reply, under terms that rule out training, and keeps nothing once the reply is written. It receives the conversation and any document uploaded to it. We attach no contact details, but whatever a customer types or uploads goes with it, and in Workspace so do the travellers’ names. A second model, from OpenAI, turns short phrases into the vectors used for matching. It never sees a whole message, and it does not train on what it receives but does keep it.
Your customer records
Held in London, for a set time
When a visitor asks to be contacted, their name, contact details and the conversation become an enquiry. It is stored in a database in London, emailed to your team through a mail service in the United States, and deleted after 24 months, or sooner on request. Deleting it removes it from the database at once and from the daily backups within seven days, and we ask your team to delete the copy in its inbox. A conversation that never becomes an enquiry is deleted after 90 days without activity, unless the visitor uploaded a file or saved a plan. Trips and conversations in Workspace are kept until you ask for them to be deleted.
Your systems
Authoritative, and left where they are
Your CRM, reservations, pricing and availability stay the record, and Tripcerto does not replace them. It writes to none of them unless you approve a connection that lets it.
Matching chooses, the model writes, your expert decides
The products and places Engage recommends are chosen by matching against the catalogue set for your account, and each one shown is recorded. Workspace builds and prices the trip, and your expert checks it and decides what goes to the customer.
Matching chooses
Each product or place a visitor is shown as a recommendation comes from matching against the catalogue set for your account, not from the model’s own knowledge.
The model writes
The model turns what matching found into a reply, in your voice, and is told never to claim to be human; the chat window tells every visitor that it is an AI assistant. It is instructed not to suggest products or places from its own knowledge. What it may do is a fixed list of actions, set for your account, and each action it takes is recorded.
Your expert decides
In Workspace, changes to a trip are logged with the person or the process that made them. Nothing reaches the customer until your expert has checked the prices, the exceptions and the recommendation.
Each business kept apart, every request checked
Each business we work with is a tenant of its own. Every request is checked against its tenant before any data is read, and changes are logged with who made them.
Separation
A check on every request
Every request is checked in the application against the tenant it belongs to, at every endpoint, before any data is read. Your consultants see your trips and no other business’s.
Access
Denied by default, granted against a need
Access to each of our systems is granted per person against a stated need, recorded the same day and removed within one working day of the need ending, with a second factor on every system that offers one. Access is reviewed every quarter, most recently in September 2026.
Machine access
A credential names your business, never a person
Your own system calls the API with a credential that names your business. Each of your customers is identified by your own reference, and the credential reaches your data and nothing else.
What is recorded
Events, changes, sign-ins and deployments
Application events are logged at the boundary with what happened, how long it took and the outcome, and sent on to a monitoring service that alerts both directors. Replies are recorded with what the model read and the actions it took, and those records are kept for 30 days. Sign-ins are recorded, changes to a trip are logged with who made them, and every deployment is tied to a commit, an author and a time.
How the system changes
A pull request, checked by the pipeline
Every code and schema change goes to production as a pull request that runs type checking, linting, tests and an architecture check, and the pipeline applies each schema change from a committed file. Secrets and service settings are set by hand, by the two directors.
An information security management system aligned to ISO/IEC 27001
The management system is written, approved and operating, and certification is what it is run towards. Tripcerto holds no certificate today and claims none.
Written and approved
September 2026
Scope, policy, risk method, a Statement of Applicability across all 93 controls of ISO/IEC 27001:2022, and the procedures the standard asks for, approved by both directors.
Operating on a schedule
Quarterly, whether or not anything has happened
Risk, access and supplier registers are reviewed every quarter. Incidents are logged, objectives are measured, and the first management review was held and minuted by both directors in September 2026.
Suppliers, chosen against four tests
A register of everything that receives personal data
Every supplier that receives personal data is recorded with what it receives, where it processes it where that is known, and what assurance it publishes. A supplier is chosen for receiving no more than the job needs, publishing where it processes, offering a second factor and publishing an independent report; a shortfall is recorded as a risk, not waived.
Availability
Checked from outside, backed up daily
External checks run around the clock from outside our own infrastructure, and their history is published on the status page. The database is backed up daily, and a restore was performed and verified in September 2026.
On the register
Information Commissioner’s Office, ZC233726
Tripcerto Ltd is registered with the Information Commissioner’s Office, the UK regulator for data protection.
The route
Cyber Essentials, then assessment
Cyber Essentials first, then an independent internal audit, then the two-stage ISO/IEC 27001 assessment. Until a certificate is issued, this page will say that none is held.
Processing terms, suppliers and your rights
What is published, what is provided on request, and how a request about a person’s data is handled.
Privacy notice
Published
What is collected, why, and the rights that go with it.
Terms
Published
The terms on which the service is provided.
Processing terms and suppliers
On request
Data processing terms, and the list of suppliers that process personal data with what each receives and where, are available on request.
Your rights
Access, a copy, erasure
A request for access, a copy or erasure is answered within the statutory month, and an erasure removes the record and the uploaded files with it.
Live status
The checks on the status page run from your own browser when it loads. The history beneath them is recorded by our monitoring service.
A pilot begins with your security questions answered
Your security questionnaire is welcome before anything is built.