OpenAI Case: Franchisee Cuts Cash Reconciliation Time 94% With ChatGPT

Three hours of monthly cash reconciliation now take 5 to 10 minutes. Annual menu repricing that used to fill five days is done in one to two hours. Those are the numbers OpenAI published on October 20, 2026, in a customer story about one restaurant owner, and they describe tasks that, in a larger company, belong to bookkeepers, payroll clerks and operations managers.

The subject is Ted Taylor, an independent franchisee who runs three co-branded Schlotzsky's and Cinnabon restaurants in New Mexico after 25 years in the semiconductor industry. The story was written by OpenAI, which sells the ChatGPT Work product it describes; every figure in it is Taylor's own account as reported by the vendor, without independent audit. [Claim] below marks those self-reported results.

What the owner moved to ChatGPT Work

[Claim] Menu repricing: once a year Taylor feeds in recipe costs, item sales history, supplier prices, local competitor prices, portion sizes and separate in-store and delivery price lists, states the overall increase he wants, and gets back a spreadsheet of recommended prices. Five days became one to two hours, a 95% cut. On the first run he did the analysis by hand as well and found the tool had flagged increases that would have left some items uncompetitive locally.

[Claim] Cash reconciliation: each month he uploads each store's cash transaction data and bank records, and the tool matches transactions to deposits. Three hours became 5 to 10 minutes, the 94% in OpenAI's headline. [Estimate] Over a year that is roughly 36 hours of reconciliation reduced to one or two, our arithmetic rather than OpenAI's.

[Claim] Payroll: summarizing payroll inputs with the tool removes four hours of manual work from every payroll cycle.

[Claim] Labor and scheduling: with read-only access to sales and labor data, the tool compares scheduled, budgeted and actual hours against forecast and actual sales over the previous 12 weeks and flags overtime, shifts that ran long and coverage gaps. For one location it now drafts the full weekly schedule from a roster of roles, training, availability and pay. The store manager checks it, adjusts it and enters it into the scheduling system. Taylor estimates better scheduling could eventually trim hourly labor by one to two percentage points of sales, worth $50,000 to $100,000 a year in free cash flow across the three stores. [Estimate] Those two figures together imply combined annual sales near $5 million, about $1.7 million per restaurant, again our arithmetic.

Which occupations this touches

No employee lost work in this account. The tasks moved were ones the owner did himself, and the one task that reached an employee, scheduling, ended with the manager still reviewing, correcting and entering the result. The exposure sits in the task list rather than in the headcount of three small restaurants.

In a chain or a franchisor's back office, the same task list is the daily work of bookkeeping, accounting and auditing clerks (reconciliation), payroll and timekeeping clerks (payroll input summaries) and food service managers (pricing, labor review, scheduling). A 94% to 95% time cut on a task differs from a 94% cut in a job: Taylor's reconciliation went from 36 hours a year to under two, but the person who uploads the files, reads the exceptions and signs off is still him.

Taylor's method is the piece a worker can copy: the same data and rules he would use himself go to the tool, its output is compared with his own work, and only then does he expand it to the next task, "always reviewing the results before implementing," in OpenAI's words.

What to prepare

For bookkeepers and payroll clerks, the remaining work in this story is the input (clean transaction files, bank exports, payroll inputs in a consistent format) and the exception review. Practice both: know what a mismatch between a cash log and a deposit looks like, and be the person who explains it. For restaurant and food service managers, the scarce input was the rule set: when to add staff in a rush, which items can absorb a price increase. Writing those rules down is what made the schedule draft usable, and the tool did not supply them on its own. For anyone in a small operation, the parallel-run habit of doing a task by hand once, comparing, then handing over is how Taylor decided what to trust.

Limits

This is a vendor-written story about a single owner-operator with three stores; the time savings are self-reported, no baseline measurement method is given, and the $50,000 to $100,000 labor figure is a projection rather than a realized result. Adoption rates across restaurants or any occupation cannot be inferred from one case. Our mapping to occupation pages is ours, not OpenAI's.

Sources

• OpenAI, "Franchisee reconciles cash in 94% less time with ChatGPT Work," published October 20, 2026: https://openai.com/index/ted-taylor-restaurant

About this analysis

This article was written with AI assistance from the full text of OpenAI's customer story. Annual-hour and sales figures are our calculations from the published numbers and are marked [Estimate]. OpenAI's text is quoted briefly; the mapping to our occupation pages is ours.