I Audited My Own Travel Software. Here Is Where Humans Still Click.

Short answer first: about half. When I measured my own product, NAMA, against the daily work of a travel agency, roughly 50 percent of the minutes an operator spends are already handled by the system. (How I measured: timing each step at a demo agency's volumes, an internal pilot estimate, not an industry benchmark.) Getting enquiries in, sorting them, and keeping everyone informed is close to 90 percent automated. The manual minutes hide in four places: building quotes, filling in bookings, making vouchers, and moving money. This article walks through all thirteen steps of an agency's back office, shows exactly where humans still click, and lists the ten fixes that chart the path from 50 percent toward 85 to 90. It is an honest audit of my own product, and I hope it teaches you something about automating an agency's back office even if you never buy anything from me.

What do I mean by "percent automated"?

I am not measuring whether a feature exists. I am measuring minutes. Out of all the time a human spends running the back office on a typical day, how much of that time does the software now do for them? A step counts as automated only if a person no longer touches it in the normal case. The numbers below come from timing each step at a demo agency's volumes. They are estimates, not lab results, but they are honest ones.

What does the daily chain of work look like?

Every booking in a travel agency travels through roughly the same chain:

1. Enquiry intake (LIVE). A traveller writes in on WhatsApp, email, or a website form, and it becomes a lead in the system. About 7 a day. Automated: a webhook plus a sweep of the mailbox every 15 minutes catches them, and the enquirer gets an automatic acknowledgement. About 95 percent automated.
2. Triage (LIVE). Triage means sorting new enquiries by urgency and value, the way a hospital sorts patients. An AI reads each enquiry on arrival and scores the hot ones. About 90 percent automated.
3. Rate-sheet ingestion (PARTIAL). A rate sheet is the price list a hotel or supplier sends, usually a spreadsheet or PDF. Every row the system extracts is still reviewed by a human, 5 to 15 minutes per sheet, a few sheets a week. About 15 percent automated.
4. Quote build (PARTIAL). Turning an enquiry into a priced offer. The system drafts priced quotes automatically every 15 minutes, but only in approve mode, and enquiries missing dates or group size get nothing at all. Reviewing an auto draft takes 15 to 20 minutes; building one by hand takes 1 to 3 hours. About 30 percent automated.
5. Quote send (HUMAN BY DESIGN). Fully human, on purpose. More on why below.
6. Booking conversion and ops fill (PARTIAL). When a client accepts a quote, someone re-types the supplier and service details into a booking, even though the accepted quote already knew them. 10 to 15 minutes per booking. About 10 percent automated.
7. Supplier confirmations (PARTIAL). Chasing the hotel or driver to confirm happens offline, across phone, email, and WhatsApp. Ten or more minutes per booking, spread over days. About 10 percent.
8. Vouchers (PARTIAL). A voucher is the document a traveller shows a hotel or driver to prove a service is booked and paid for. Each booking needs 3 to 5 of them, created and issued by hand. Only the final combined PDF is automatic. 10 to 25 minutes per booking. About 10 percent.
9. Invoicing (PARTIAL). Drafting, issuing, and emailing an invoice are three separate manual clicks, 5 to 8 minutes per booking. About 20 percent.
10. Collections (PARTIAL). Money coming in. Online payments record themselves; offline receipts are keyed in by hand, 2 to 5 minutes each. About 40 percent.
11. Vendor payables (PARTIAL). A payable is money the agency owes a supplier, such as the hotel's bill. 2 to 4 per booking, keyed in by hand, 3 to 5 minutes each. About 10 percent.
12. Follow-ups (PARTIAL). The daily queue of nudges to clients. The system drafts them, a human approves and sends, 1 to 2 minutes each. About 50 percent.
13. Awareness (LIVE). The morning digest, the ops audit, and traveller care messages. About 90 percent automated.

Which parts are already automated?

The front door and the dashboard. Intake, triage, and awareness all sit around 90 percent. In practice that means every enquiry is picked up and acknowledged within minutes in the normal case, hot leads are scored as they arrive and surface in the morning digest, and the owner sees the day's picture without asking anyone for an update. If you run an agency today and your team still copies enquiries from WhatsApp into a spreadsheet by hand, that is the cheapest problem to fix. Tools for this exist, mine included, and this part of the problem is largely solved across the industry.

Where do humans still spend their minutes?

Four legs of the chain, and they share one pattern: the system already knows the information, and a person moves it by hand anyway.

Quote build. An active agency builds about 2 quotes a day. The auto-quote feature prices only the enquiries that arrive complete, and it runs in approve mode only, so a human still reviews every single draft. Enquiries missing dates or group size, which are common, get no draft at all and fall back to a fully manual build of 1 to 3 hours. The tool exists, but the trust dial is set to zero.

Booking fill. When a client says yes, the accepted quote already knew the supplier, the service, and the rate. The booking asks for all of it again, and someone in ops types it in a second time. 10 to 15 minutes per booking of pure re-typing.

Vouchers. Each booking needs 3 to 5 vouchers, one per service. The confirmed booking already holds the traveller's name, the dates, the hotel, the inclusions. Yet each voucher is created and issued by hand, and only the step that merges them into one PDF for the traveller is automatic.

Money. Three separate jobs. Invoices: draft, issue, and email are three clicks done by a person. Payables: each supplier bill is keyed in by hand, 2 to 4 times per booking. Collections: online payments record themselves, but every offline receipt is manual entry. Add supplier confirmations, where someone chases the hotel or driver by phone and WhatsApp over several days, and you have the full picture. Legs 6 through 11 of the chain, booking fill through payables, make up roughly 80 percent of the remaining manual minutes in an active agency.

Why does the quote still go out by hand?

Because approval before send is the product promise, and it stays. A quote is a price commitment to a real customer. One wrong automated quote can cost real money and real trust, and no agency owner I know wants software making pricing promises on their behalf. So the send button stays human by design. The goal of every fix below is to remove the typing and chasing around the human decision, never the decision itself.

What are the ten fixes, ranked by time saved?

1. Let good auto-quotes send themselves, within limits (ROADMAP). For FIT quotes, meaning small custom trips for individual travellers rather than big groups, under a rupee threshold the agency sets itself, the system moves from approve mode to auto mode. Small, standard quotes stop waiting in a queue.
2. A reply bot for incomplete enquiries (ROADMAP). When an enquiry is missing dates or group size, the system asks for the two or three missing fields automatically, on the same channel the traveller used. Today those leads get nothing, and they are a large share of the manual builds.
3. One-click quote to booking (ROADMAP). Carry the supplier, service, and rate from the quote's line items straight into the booking. No re-typing.
4. Auto-create and issue vouchers (ROADMAP). Confirmed booking items already contain everything a voucher needs. Generate and issue them automatically, and keep the combined PDF the traveller receives.
5. Auto-draft the invoice on booking confirmation (ROADMAP), and email it when issued. Three clicks become zero in the normal case.
6. Auto-create vendor payables from booking items (ROADMAP). Same data, one more place it should never be re-keyed.
7. A trust ladder for rate sheets (ROADMAP). After the same vendor's extracted sheets have been approved enough times in a row, the system applies new ones automatically and spot-checks one in ten. Human review shrinks as accuracy proves itself.
8. Follow-ups send themselves after 24 hours untouched (ROADMAP), with a delay the agency controls. A drafted nudge that nobody reviews for a full day is just a slower send anyway.
9. A supplier-confirmation chase bot (ROADMAP). The system emails or messages the supplier, reads the reply, and updates the confirmation status. Ten minutes of chasing spread over days becomes a watched thread.
10. Match receipts to invoices automatically (ROADMAP). Incoming payments get matched to open invoices, and only the unmatched ones queue for a human.

What does the agency look like after these fixes?

The math first. Legs 6 through 11 are roughly 80 percent of the remaining manual minutes. Fixes 3 through 6 and 9 through 10 clear most of those. Fixes 1 and 2 clear the quote side. The path leads toward 85 to 90 percent automation by operator minutes, a target to reach, not today's number, and it does not touch the approval promise.

In plain terms, staff stop acting as copy-paste machines between screens that already hold the data. The work that remains is judgement: approving a quote, handling the odd unmatched receipt, talking to the client who needs a person. That is the work agencies actually value, and it is the work that was being crowded out by typing.

Why publish an audit of my own product?

I sell this software, so admitting where it still makes people click might look odd. Two reasons. First, agency owners deserve honesty about what automation can and cannot do today, from someone who has measured it instead of marketing it. Second, the gap list above is the roadmap. Every fix named here is specific, small, and grounded in data the system already has. None of it needs magic. One more detail, because it is fitting: this article was itself drafted by AI, approved by a human, and published through NAMA's own blog engine, the same approve-before-send workflow described above.