Traveller expectations have moved on, and the way most agencies handle a customer request has barely changed in twenty years. An email arrives, a consultant reads it, re-types the details into a booking engine, builds a quote and sends it back. Multiply that by hundreds of requests a week, across more channels than ever, and you have the real constraint on agency growth today.
Travellers are not waiting. Extensive use of ChatGPT and similar AI apps for trip planning more than doubled in a year, from 13% in 2024 to 30% in 2025, according to Skift research highlighted by McKinsey. The same analysis notes that travel still trails other industries in AI maturity, held back by siloed data and outdated systems.
That gap is both the threat and the opportunity. Below are the six operational challenges we see most often across agencies and TMCs in more than 60 markets, and what AI can realistically do about each one.
Requests arrive by email, WhatsApp, phone follow-ups, shared mailboxes and system notifications from GDSs and suppliers. Most of them land in someone's inbox. When that consultant is busy or out, the request waits. Nobody sees the full queue, so managers discover a missed request only when the customer chases.
Where AI helps: the first job is not intelligence, it is consolidation. Once every channel feeds a single shared queue, AI can read each incoming request, recognise the customer, classify the request type and set its priority before a person opens it. Ownership becomes visible, and response times stop depending on who happens to be at their desk.
A consultant's value is judgment: knowing which hotel suits a honeymoon, which connection is too tight, which corporate traveller always wants an aisle seat. Yet much of their day goes on reading emails and re-keying destinations, dates and traveller details from one screen into another.
Where AI helps: this is exactly the work language models do well. A message such as "We'd like a hotel in Lisbon from 12 to 19 July for two adults and our son" can be turned automatically into structured trip data: Lisbon, 12 to 19 July, two adults, one child, family leisure request. AI can then draft the reply in the agency's tone and the customer's language. The consultant reviews instead of retyping.
Customers often ask several agencies at once. The first good, relevant offer usually wins. Building a proposal by hand means searching inventory, comparing options, copying details, prices and conditions into a document or an email, and checking it all against the client's budget. For a multi-option family holiday that can take a large part of a working session.
Where AI helps: AI agents, rather than chatbots, can take a request, search the agency's own inventory, rank options against the brief and budget, and return a ready-to-send proposal with photos, ratings, room tables and refund rules. The consultant adds what only they know, such as where the family stayed last year, approves and sends. Minutes instead of hours, without losing the personal touch.
In most agencies, more requests mean more consultants, because every request costs roughly the same manual effort. But experienced people are scarce. In a 2024 survey of North American agencies by American Marketing Group, 59% said hiring qualified advisors was "not easy at all", and 56% of agencies that lost advisors during the pandemic had not replaced them. Agencies across Europe, the Middle East and Africa describe the same pressure.
Where AI helps: AI breaks the link between volume and headcount. When sorting, extraction, drafting and first-draft quotations are handled by AI, the same team can serve more requests, and new hires become productive faster because the routine work is already done for them. The goal is not fewer people; it is putting the people you have where their expertise earns margin.
Most agencies know their booking volumes and margins. Very few know how long a leisure quote, a corporate change or a refund actually takes to handle, or what that time costs. Generic helpdesks report ticket counts, not effort. Without that visibility, it is impossible to know where automation would pay off.
Where AI helps: when AI and people work in the same system, every action is measurable. Leaders can see cost per request and how the work splits between AI and people, by request type. That turns AI from an act of faith into a business case: you can see where it is paying off and where to extend it next.
Many agencies have already tried AI: a chatbot on the website, a writing assistant in a separate tab, a pilot that never left the IT department. Adoption stalls for two reasons. The tools sit outside the daily workflow, so consultants forget to use them. And leaders worry, rightly, about AI sending the wrong thing to a valuable client.
Where AI helps, if it is done right: AI has to live where the work happens, on every request, not in a separate window. And control has to be a dial, not a switch. Agencies should be able to decide, per request type, whether AI only drafts, acts with human confirmation, or resolves routine requests end to end, with any human action pausing the AI.
Across these six challenges, the same four principles separate AI that changes operations from AI that becomes another unused tool:
| Principle | What it means in practice |
|---|---|
| Inside the workflow | AI works on every request where consultants already work, not in a separate app |
| Built for travel | It understands trips, travellers, PNRs and bookings, not just generic tickets |
| Does the work | Agents complete tasks such as building offers, rather than only suggesting text |
| Under your control and measurable | Autonomy set per request type, with cost and time tracked for AI and people |
These principles are what we built ARIA around. ARIA is an AI-native workspace where travel consultants and AI agents handle customer requests side by side. Email, WhatsApp Business and system notifications arrive in one queue; AI classifies each request, extracts the trip details and drafts the reply; and specialised AI agents take on the repetitive work.
The first of these, the IRIX Offers Agent, builds ready-to-send flight and hotel proposals from IRIX content in minutes. Agencies decide how far AI goes for each agent and request type, and nothing goes out unless their rules allow it. ARIA runs as the operating layer on top of the dcs plus platform, including TINA for mid- and back-office, and we estimate that agencies using it can achieve more than 50% operational efficiency growth.
Travel remains a human business. A good consultant's judgment is not something software replaces. What AI changes is everything around that judgment, so your people spend their time on travellers instead of on inboxes.
Want to see what this looks like with your own requests? Book a live ARIA demo and bring your hardest inbox scenarios.