Short answer: We take published figures for seven jobs an agency does by hand and run them on its weekly transactions. At 1,385 transactions a week, the mid case comes to 103 hours of work today and 33 with Majeste. It stays a projection until the Audit times your own week.
Which jobs does the estimate count?
Seven jobs that sit with an agency's accounting and ticketing team every week. Each one runs on the weekly transaction count, or on a count you give us, like the corporate accounts that get a monthly report.
| Job | Today | Back |
|---|---|---|
| ARC report and IAR match | 34h 30m | 28h 50m |
| Refunds | 17h 20m | 15h 00m |
| Accounting inbox | 25h 50m | 10h 20m |
| Unused credits | 9h 15m | 7h 45m |
| Client reports | 6h 55m | 3h 25m |
| Commission | 3h 40m | 2h 25m |
| Debit memos | 5h 55m | 2h 20m |
| All seven jobs | 103h 30m | 70h 15m |
Hours back a week, by job
The 1,385 figure is a published case for a mid-size TMC, and it counts transactions of every type. Type your own count into the estimator and every row reruns.
Where do the inputs come from?
Each input is a published figure: minutes a task takes, how often it comes up, and the share of the time that comes back once it runs on its own. Most come from software providers. Every source carries a tag, so you can see which figures are independent, which come from trade press and which are a company's own claim.
A worked example from the ARC report
One published case saves 1.25 minutes per transaction. At 1,385 transactions, that's 1,385 × 1.25 ÷ 60 = 28.85 hours back a week. A bank's reconciliation study puts the share that comes back at 83.6 percent, so the time today works out to 34.52 hours.
How wide is the range?
From 11 to 152 hours back a week at 1,385 transactions. The low end runs every job on its lowest published input, and the high end on its highest. The estimate shows the middle inputs, 70 hours back.
What checks the number?
- APQC's realized median for finance automation is 20 percent, which would give the same agency 21 hours back a week. The mid case assumes 67.9 percent comes back, 3.4 times that median.
- 46 of the 70 hours back rest on software providers' own savings claims.
- Studies of AI across whole jobs find 2 to 3 percent of hours saved in office and finance work. They average everyday use across every task, while the estimate counts named jobs set up from start to finish.
What does the estimate count, and what stays with your team?
- Exchange reissues stay with your travel agents, so they sit outside the count.
- Some jobs share work, such as refunds inside the ARC report. Each job counts in full, so the total can run high.
- Client reports and commission start at 45 accounts and 8 statements a month, the counts in the published figures. Your own counts replace them.
- A year here is 52 working weeks.
What replaces the projection?
Your own week. In the Audit we sit with your team and time each job as they do it. That measured baseline replaces this projection, and every result after it is measured against the baseline with your team.
Common questions
Is the estimate a guarantee?
It's a projection from published figures. In the Audit we time your team's own week, and that measured baseline replaces the projection.
Why do so many inputs come from software providers?
Few independent studies time these jobs inside travel agencies. Every source carries a tag, so you can weigh the independent figures and the company claims yourself.
What does the range of 11 to 152 hours mean?
It runs every job on its lowest published input, then on its highest. The estimate shows the middle inputs, 70 hours back at 1,385 transactions a week.
How do I run the estimate on my own numbers?
Open the estimator and type your weekly transactions. Every job reruns on your count, and you can copy a link to your estimate.
Sources (39)
Show every source
- Accelirate (2024) Company's own claim
- Bank reconciliation paper in CEUR Workshop Proceedings (2025) Independent
- Peakflo (2026) Company's own claim
- Grasp Technologies (2024) Company's own claim
- Grasp Technologies (2026) Company's own claim
- Host Agency Reviews (2018) Trade press
- Travelweek (2018) Trade press
- Airlines Reporting Corporation (2023) Independent
- Built In job posting (read 2026) Independent
- US Bureau of Labor Statistics (2026) Independent
- WNS (2023) Company's own claim
- Noy and Zhang in Science (2023) Independent
- Anthropic (2025) Company's own claim
- TripActions data via Travolution (2021) Company's own claim
- Cornerstone Information Systems (2019) Company's own claim
- Magnatech case study (2024) Company's own claim
- APQC via CFO.com (2026) Independent
- Amadeus (2018) Company's own claim
- UiPath customer story (read 2026) Company's own claim
- Amadeus (2016) Company's own claim
- Sabre Travel Network Indonesia (read 2026) Company's own claim
- IGT Solutions on UiPath Marketplace (2018) Company's own claim
- Cybage (2025) Company's own claim
- Travefy (2026) Company's own claim
- Rezdy (2014) Company's own claim
- Tramada via Travel Weekly Australia (2019) Company's own claim
- APQC via CFO.com (2023) Independent
- AFP and APQC via Vena Solutions (2023) Trade press
- Association for Financial Professionals (2020) Independent
- Ardent Partners via Bottomline (2024) Trade press
- IOFM via APEX Analytix (2024) Trade press
- European Business Magazine (2025) Company's own claim
- SoftCo (2026) Company's own claim
- Airlines Reporting Corporation (2026) Independent
- Stanford and MIT study via CFO Dive (2025) Trade press
- Stanford Graduate School of Business (2025) Independent
- Federal Reserve Bank of St. Louis on office support jobs (2026) Independent
- Federal Reserve Bank of St. Louis on business and finance jobs (2026) Independent
- Accounting system manual, IAR reconciliation chapter (undated) Company's own claim
Checked on October 26, 2026. Rules change, so check the source before you act on a date or an amount.
