Direct answer: under the ABTA Code of Conduct (current member PDF, September 2026), Clause **5B** requires Members to deal with all correspondence with Clients as promptly as possible and, in any event, within fixed clocks — **(i)** an acknowledgement not later than **14 days** from receipt, and **(ii)** a **detailed reply**, or a reply containing a **detailed explanation for any delay**, not later than **28 days** from receipt. ABTA’s own Holiday complaint FAQs restate the same Member duty in plain language. Clause **5D** requires every reasonable effort to reach a **speedy solution** in a dispute (including with a Client’s formally appointed representative). Clause **5E** requires Agents for an Organiser or other Principal to handle minor complaints where possible, and otherwise to **forward** complaints to the Principal **without undue delay** and use best endeavours as intermediary. Clauses **5F** / **5G** keep the ABTA arbitration / ADR path available when amicable settlement fails. An AI complaint, CRM, or email agent that auto-acks without a durable receipt clock, invents a “we’ll get back in 48 hours” SLA the Code does not state, closes a thread before a detailed reply is due, or treats a principal-level complaint as “resolved in chat” without forwarding is not “customer-experience automation”. It is a **member-conduct correspondence control failure**.
That is the operator question Kaize is answering here. We are a UK AI product studio for travel and hospitality. This is **not legal advice**, not a substitute for the ABTA Code, the Guidance on the application of the Code of Conduct (February 2026), your membership terms, booking conditions, or counsel. It is a **control-stack** article: what must already be true in the complaints state machine before AI is allowed to acknowledge, draft, escalate, or close client correspondence for an ABTA Member. It sits beside — and does not replace — our CCR 2013 pre-contract information gates, our PTR Regulations 12–14 termination and refund controls, our CMA AI-agents consumer-law controls, our PECR soft opt-in enquiry follow-up gates, and our ICO ADM / profiling gates for significant CRM decisions. This article is narrower: **ABTA Code correspondence and dispute clocks must be encoded before AI may handle Member complaint traffic**.
Why generative complaint agents keep colliding with Clause 5B
Most ABTA Members already know the 14- and 28-day clocks exist. The 2026 failure mode is a **generative complaints assistant** that can ingest an email, draft a sympathetic acknowledgement, promise a callback, open a ticket, and mark the case “responded” **without** proving when the correspondence was received, whether the ack went out inside 14 days, whether the 28-day artefact is a **detailed reply** or a **detailed delay explanation**, whether the matter is agent-level or must be forwarded to a Principal under 5E, and whether ADR / arbitration signposting is required when deadlock is reached. ABTA’s consumer-facing Code page states customers of Members can expect a response to any complaint within 28 days — that is a **membership conduct expectation**, not a prompt tip to “sound helpful”.
Five ABTA facts teams must encode before any auto-ack or “smart complaint reply” tool is armed on Member client correspondence:
- **14-day acknowledgement** — Clause 5B(i): acknowledgement not later than 14 days from date of receipt.
- **28-day detailed reply or detailed delay explanation** — Clause 5B(ii): detailed reply, or reply containing a detailed explanation for any delay, not later than 28 days from receipt (restated on ABTA’s Holiday complaint FAQs).
- **Speedy dispute resolution** — Clause 5D: every reasonable effort to reach a speedy solution; treat formally appointed representatives the same way.
- **Agent forwarding without undue delay** — Clause 5E: minor/general complaints handled where possible; principal-level complaints forwarded without undue delay with best endeavours as intermediary.
- **Arbitration / ADR path preserved** — Clauses 5F / 5G and ABTA’s dispute process: amicable settlement first; arbitration scheme remains available under scheme rules when settlement fails (payment of awards has its own clocks — do not invent them in chat).
Map the case state before the model writes
Complaint SLA compliance is not a tone-of-voice prompt. Your assistant needs structured fields for: receipt timestamp (durable, timezone-aware); correspondence channel and thread id; ABTA membership capacity on this booking (Member as Principal vs Agent for named Organiser/Principal); case stage (new / acknowledged / detailed-reply-due / delay-explained / forwarded-to-principal / deadlock / ADR-registered / arbitration); clock status for 14-day and 28-day duties; whether a human case owner has approved the detailed reply; and whether the draft is an acknowledgement, a detailed merits reply, or a detailed delay explanation. Those fields must be **sourced from CRM/case systems and membership configuration**, not invented in chat. A model that replies “we aim to respond within 2 working days” as if that replaced Clause 5B is inventing a Member SLA ABTA does not publish as the Code minimum.
The six gates before AI may handle an ABTA Member complaint
Gate 1 — Lock receipt time and channel before any acknowledgement
Before any generative turn emits an acknowledgement, bind the case to a durable **date of receipt** and channel. Gate 1 fails if the assistant “acks” verbally on a phone note without writing an acknowledgement into the correspondence log, or if receipt time is guessed from model context instead of the inbound message metadata / case system.
Gate 2 — 14-day acknowledgement clock is a hard gate, not a vibe
Clause 5B(i) is a maximum: acknowledgement not later than 14 days from receipt. Gate 2 fails if the agent queues acknowledgements behind unrelated marketing drafts, or if an automated “thanks, we got this” is never actually sent / logged within the clock.
Gate 3 — 28-day artefact must be detailed reply or detailed delay explanation
Clause 5B(ii) does not accept a one-line “still looking into it”. The 28-day artefact is a **detailed reply** or a reply with a **detailed explanation for any delay**. Gate 3 fails if the model sends holding fluff without substance, or marks the case closed after only the 14-day ack.
Gate 4 — Agent vs Principal routing (Clause 5E) before “case closed”
If the Member is an Agent, minor/general complaints should be handled where possible; principal-level complaints must be forwarded **without undue delay**. Gate 4 fails if the assistant settles (or rejects) a hotel/flight principal dispute in chat without forwarding, or invents that the Agent “is not responsible” as a substitute for 5E duties.
Gate 5 — Speedy dispute resolution and appointed representatives (Clause 5D)
Clause 5D requires every reasonable effort toward a speedy solution and equal treatment of formally appointed representatives. Gate 5 fails if the agent stalls for model “more information” loops past the Code clocks, or refuses to engage a solicitor/claims handler the Client has appointed.
Gate 6 — Deadlock and ADR / arbitration signposting — humans confirm, AI does not settle awards
When amicable settlement fails, ABTA’s dispute / arbitration path exists under scheme rules (see ABTA’s resolving-disputes materials and Code 5F/5G). Gate 6 fails if the assistant invents compensation awards, pretends chat settlement is arbitration, or auto-pays / auto-refuses an award without human finance and legal controls. Late arbitration-award payment has driven real Code enforcement — do not let a model own that path.
Operator checklist before you arm the complaints agent
- Durable receipt timestamp + channel + thread id on every inbound complaint.
- 14-day acknowledgement job with escalation if unsent; ack content logged, not only “intent to ack”.
- 28-day detailed-reply or detailed-delay-explanation artefact with human approval for merits decisions.
- Capacity flag: Member as Principal vs Agent for named Organiser/Principal on this booking.
- 5E forwarding workflow with undue-delay SLA and intermediary notes — not “tell the client to email the hotel themselves” as the only step.
- Appointed-representative flag so 5D equal treatment is enforced in routing.
- Deadlock / ADR registration path with human case owner; AI may prepare packs, not invent awards.
- No invented SLAs (“2 working days”) that replace or dilute Clause 5B minima in customer copy.
- Logging: clocks, artefact type, forward events, approvals, and gate failures stored for ABTA / audit review.
- Do not conflate ABTA Code correspondence with PTR refund rights, ATOL certificate duties, or PECR marketing consent — those gates remain separate.
What this article is not
- Not a substitute for the ABTA Code PDF, Guidance, membership conditions, or scheme rules for arbitration fees/limits.
- Not advice on whether a specific complaint is a Code breach versus a contractual compensation claim.
- Not a claim that Kaize has implemented ABTA complaint SLAs for a named Member — no invented case studies.
- Not an update to homepage, services, or pricing pages (those remain approval-required).
How Kaize frames the build
Kaize designs travel and hospitality AI so **control planes sit in front of models**. ABTA Code correspondence readiness is one of those gates: the complaints path does not emit acknowledgements, detailed replies, principal forwards, or ADR signposts until receipt clocks, 14/28-day artefacts, Agent/Principal routing, 5D representative handling, and human approval for merits/deadlock steps are live on structured case data. That is how UK ABTA Members keep generative complaint handling useful without turning Clause 5B into an ungoverned chatbot timer.
Sources for operators to read next: ABTA Code of Conduct (hub); ABTA Code of Conduct PDF (September 2026); Guidance on the application of the Code of Conduct (February 2026); Holiday complaint FAQs; Resolving disputes; related Kaize gates on CCR 2013 pre-contract information, PTR termination/refund controls, and CMA AI-agents consumer law.