November 7, 2026. Travel Code has opened a free travel policy health check. Send your travel and expense policy as a PDF and you get back a score out of 100, the rules that contradict each other, the rules too vague to enforce, the topics the document never covers — and a machine-readable version of the same policy that an AI assistant can apply at booking time. It is free, there is no card and no trial, and a Travel Code specialist reviews every report before it is sent.
Why policy length stopped being a style problem
In The State of Corporate Travel Policies: U.S. and Canada 2025, produced by GBTA with the travel management company ALTOUR, 168 corporate travel managers were surveyed between 19 November and 10 December 2025. Three findings from it describe the document most companies are about to hand to an AI assistant:
- 51% of travel policies run longer than 10 pages, and 24% run longer than 20.
- Only 14% of travel managers say their policy has got shorter over the past three years.
- 32% name unfamiliarity with the rules — employees simply have not read them — as a reason bookings go out of policy.
The same study found that 46% of policies tell employees to book a "reasonably priced" hotel, while only 30% set an actual per diem or rate cap. For a human traveller that is an inconvenience. For software it is a dead end: "reasonably priced" cannot be evaluated, so the rule is either ignored or guessed at.
That is the gap the 2027 outlook published by The Business Travel Magazine keeps returning to. Mark Parker, Regional Director of Sales EMEA at ALTOUR, expects travel policy to be delivered as guidance inside the booking flow rather than as a document, and warns about what happens if the document goes in unchanged: "The risk is automating ambiguous rules and making poor decisions faster." In the same piece Steve Banks, CEO of the Travel Division at Identity, puts the first job plainly — getting policy "into a form an AI can actually follow", with the travel manager governing the output instead of policing a booking tool.
Both describe the same prerequisite. Before a policy can be automated, somebody has to find the places where it contradicts itself and the places where it says nothing measurable. That is a reading job, and on a 19-page document with three appendices nobody volunteers for it.
What the check returns
Six things, each tied to the exact quote and page number it came from, so the report can be argued with rather than taken on faith.
- A score from 0 to 100. Calculated by fixed rules from what the check finds, so the same document always gets the same score and two policies can be compared.
- Length and reading time, against the 10-page mark from the GBTA study.
- Contradictions — pairs of rules that cannot both be followed, with both quotes and both page numbers. Most often an old amendment against the main text, or an appendix against a list.
- Rules too vague to enforce — "reasonable", "where possible", "with the permission of management" — each with a concrete wording you could actually put into a system.
- Gaps, against a 15-point checklist: approvals, cabin class, price caps, hotel limits by city, per diem, expense reporting, booking channel, duty of care, unused tickets and more.
- A machine-readable version of the policy: limits, cabin classes and approval thresholds as configuration.
A worked example from a real report: a 19-page policy scored 51/100. One amendment on page 19 required per diem to be paid even when the host covers travel, lodging and meals, while page 11 said the sending party does not pay per diem in foreign currency. Elsewhere, an exception was allowed "with the permission of the head of the organization" — no limit, no named approver. The suggested replacement was specific enough to configure: up to 150% of the cap, approved by the CFO.
If you want to see what a complete document looks like before sending yours, our corporate travel policy guide and template and the ten real Fortune 500 policy examples cover the same 15 points the check scores against.
The machine-readable part
The last item is the one that changes anything operationally. The check converts the policy into limits, cabin classes and approval thresholds that Robert, the Travel Code AI assistant, applies when a trip is actually booked: a traveller asks in chat, every option is checked against the company's own rules, and anything out of policy goes to the named approver before money moves. The underlying agent, Policy Builder, is in beta today, and every rule it produces is reviewed by a human before it goes live.
The same rules apply when the booking does not come from our interface at all. Since Travel Code opened its MCP connector for write access, an employee can book from Claude, ChatGPT or Gemini — and the policy is evaluated on our side, not on the assistant's. A vague rule is therefore not a documentation problem any more; it is a rule the booking path cannot check.
Two of our own guides go deeper on the mechanics: turning policy rules into automation and compliance workflows, and approval workflows for the C-suite, where exception thresholds tend to be written most loosely.
What the check is not
It is a readability and enforceability review, not an audit and not a legal opinion. Specifically:
- It does not certify the policy against any standard, and it issues no certificate or statement of conformance.
- It does not review how the policy is being followed — that needs booking data, and the check needs none. The PDF is all we ask for.
- It does not rewrite the policy for you. It shows the lines that break and suggests enforceable wording; the decisions stay with the travel, finance and HR owners. Mark Parker's point about policy design being shared across travel, HR, risk and technology applies here too.
- It is not automated end to end. Our AI reads the document, then a Travel Code specialist checks the report before it is sent — usually within one business day.
On the file itself: it is used only to produce your report, it is not used to train AI models, and it is deleted within 30 days. The full terms are in our privacy policy.
How to send your policy
Go to travel-code.com/lp/policy-check and attach the document. A PDF up to 25 MB and 100 pages, without a password; scanned pages are fine. Name, company and work email are required, phone is optional. One check per company email. The report arrives by email from robert@travel-code.com, with the machine-readable version attached.
You do not need to be a Travel Code customer, and nothing about the check depends on becoming one. If you do use Travel Code, the machine-readable file is what makes the policy start working on the next booking instead of sitting in a PDF nobody opens. If the compliance side is what you are measuring, our guide to corporate travel policy compliance sets out what to track once the rules are enforceable.
Frequently Asked Questions
Is the travel policy health check really free?
Yes. There is no trial, no card and no obligation. One check per company email.
How long does it take?
Usually within one business day. Our AI reads the policy and a Travel Code specialist checks the report before it reaches you, which is why it is a day rather than minutes.
Who sees our travel policy?
Our AI reads it, and one Travel Code specialist reviews the report before it is sent. The file is used only for that report, is not used to train AI models, and is deleted within 30 days.
What file formats work?
A PDF up to 25 MB and 100 pages, without a password. Scanned pages are fine — the check reads them.
Does this work for a combined travel and expense policy?
Yes. The 15-point checklist covers expense reporting, per diem and receipt rules alongside booking, cabin class and approvals, so a combined T&E document is scored on all of it.
Is this a compliance or duty-of-care audit?
No. It reviews the document for contradictions, unenforceable wording and missing topics. It does not assess your program against any standard and it issues no certification. Duty of care appears only as one of the 15 topics a policy is expected to address.
What is the machine-readable version, in practice?
Your rules expressed as configuration — price caps, cabin classes, hotel limits by city, approval thresholds and who approves what. Loaded into Travel Code, those are the rules Robert checks every option against at booking time, including bookings made through an AI assistant over the MCP connector.
Do you need access to our booking data?
No. The check reads the policy document and nothing else.
Sources
- GBTA and ALTOUR, The State of Corporate Travel Policies: U.S. and Canada 2025 — survey of 168 corporate travel managers, fielded 19 November to 10 December 2025, published 5 March 2026.
- The Business Travel Magazine, Top trends for 2027 — contributions from Mark Parker (ALTOUR) and Steve Banks (Identity).
- ALTOUR, Why 10-page travel policies fail.
