Tai Custom Agent Templates

These are the agents we've built and tested with customers during the private preview. Every one of them is available to deploy into your environment either as-is, or adapted to your workflows.

To request any of these, raise a support ticket with the agent name and we'll set it up for you. Or use the **Draft with AI** feature in the Workbench to build your own version: simply paste the description below into the prompt box, hit Draft with AI, and it will generate the system prompt and skill configuration for you as a starting point.

Locates affected customers fast when an incident, disaster, or emergency occurs at a destination. Enter a destination and the agent immediately identifies all customers currently in-country, arriving within 48 hours, and recently departed — including passengers who may be stranded at stopovers en route.

For each affected booking it surfaces full passenger details (lead traveller and co-travellers), hotel, travel dates, customer contact, and agency contact — everything needed to act without manually searching individual bookings.

When to use: As soon as an incident is declared at a destination. Answers "who do we have there right now?" and produces a contact sheet ready to work from immediately.

Pulls every booking currently in Action Required status, surfaces exactly which components are still unresolved, and prioritises by urgency — bookings that entered Action Required in the last 7/14/21 days versus older unresolved items. Links directly to Cockpit so your ops team can act immediately.

When to use: Run as a scheduled report so nothing slips through to departure undetected.

Sweeps all confirmed bookings departing in a configured time frame and checks each against a pre-departure checklist. Flags outstanding unpaid balances, components in Failed or On-Request status, incomplete passenger details (missing names, DOBs, placeholders like "TBD"), open internal remarks, and transfer times that don't align with booked flights.

Outputs a destination summary, a full booking list sorted by departure date, a per-flag action note for every issue, and a pax-vs-destination heatmap.

When to use: Run weekly, or daily as the departure window closes. Payment issues and component failures are significantly harder to resolve inside two weeks of departure.

Paste a booking reference and the agent fetches the full component picture, diagnoses what failed and why, calculates financial exposure from cancellation penalties, flags time pressure, and tells you exactly what to do — and what not to touch — before taking any action.

When to use: Any booking in Action Required with a component failure. Particularly useful for complex multi-supplier bookings where the root cause isn't obvious from the error message alone.

Analyses component booking outcomes across all suppliers over a specified period, surfacing failure rates, on-request rates, and cancellation rates by supplier and component type. Reports two cancellation metrics separately: customer-facing cancellation rate and cancel API failure rate — often the more operationally significant of the two.

Low-volume suppliers are excluded from main rankings but shown in a sub-table. A monthly variant produces a fixed rolling three-month HTML report of booking failure rates and top error messages — designed to be shared as-is with no editing.

When to use: Ahead of supplier renegotiations or product reviews. Run the monthly variant on a schedule and share with ops leadership.

Analyses booking change events over a specified period — breaking down volume and net revenue impact across price-increasing, price-decreasing, and neutral changes, and splitting by who initiated the change (travel agency, consumer, tour operator) with a weekly trend view.

When asked to drill into losses, it analyses the negative-impact subset across every available dimension and pulls full booking history for the largest cases to distinguish real revenue drains from administrative corrections. Surfaces concentration patterns rather than treating all negative events as equally significant, and closes with concrete follow-up actions per finding.

When to use: When ops or leadership wants to understand whether booking changes are net-accretive or a revenue drain, and to identify which bookings or patterns are driving the losses.

Evaluates whether a promo campaign is actually working — not just redemption counts, but whether it's driving incremental bookings worth the discount given away, and how it stacks up against other codes competing for budget.

Answers three questions: is this code earning its keep (redemptions, discount cost, conversion, cancellation rate); is promo traffic behaving differently from full-price traffic; and which of several live codes is performing best when compared side by side.

When to use: Campaign reviews, budget conversations about promo spend, or any time someone asks "is this code actually working."

Pulls together bookings that don't show up cleanly under a normal supplier name — works for any manually booked component type. Groups everything by region and category, shows totals and subtotals, flags anything missing or unusual, and outputs one clean monthly HTML report with PDF export. No chat back-and-forth — the report is the whole output.

When to use: Recurring internal report for tracking bookings that sit behind generic "manual" or "ad hoc" supplier tags — useful for inventory planning and loading decisions.

Designed for product managers and commercial teams. Analyses booking performance and demand trends across destinations and trip templates, then enriches findings with external intelligence to explain the numbers. Ranks destinations by bookings, revenue, and growth rate, always breaking results to template level.

When to use: Quarterly portfolio reviews, before pricing or capacity decisions, or when a destination shows an unexpected demand shift that internal data alone can't explain.

Turns a completed TripBuilder itinerary into a polished, print-ready HTML quote for the traveller — cover section with trip title, dates, traveller names and total price, followed by a day-by-day narrative, inclusions/exclusions, a pricing summary table, and booking terms.

Pricing and dates are taken verbatim from the itinerary (never recalculated or invented) and writes in the itinerary's own language unless told otherwise, and delivers a single downloadable file.

When to use: Generating a customer-facing quote from a completed itinerary, ready to send or convert to PDF.

Compiles a supplier-ready room list for a specific group tour departure. Given a tour template ID and departure date, it fetches all confirmed bookings, extracts traveller data (full names, passport names, DOB, nationality, room type, occupancy, special requests), groups travellers into rooms, and outputs a formatted document ready to hand directly to a hotel or ground handler.

When to use: The morning before finalising supplier communication for a group departure, or before any pre-departure call with a hotel or ground handler.

Any of the descriptions above can be dropped into **Draft with AI** in the Workbench to generate a starting-point agent for your own environment. Draft with AI writes the system prompt, selects the right skills, and flags anything redundant — so you're refining rather than starting from scratch.

For advice on queries that involve large datasets — locating customers across a full country, running reports across thousands of bookings — see Working with large datasets.