How a Missed 2,000-Kueh Order Pushed HarriAnns Toward AI

A missed order for 2,000 kueh made an administrative risk impossible for HarriAnns Nonya Table to ignore. After a hotel chef called about the omission, third-generation owner Alan Tan turned to AI-assisted order management, according to The Straits Times.2

Where an order could get lost

HarriAnns handles hotel and individual orders alongside its cafe business. An employee’s published description of the work lists receiving purchase orders, entering work orders, sending instructions to separate kitchens, preparing labels and arranging delivery paperwork. Each transfer matters: an order that does not reach the production tally cannot be fixed by calculating that tally more accurately.43 HarriAnns had eight outlets as of February 2026.49

The account of an email-to-printed-purchase-order-to-Excel process points to a plausible failure mode: a hotel’s order can be received but overlooked before it becomes a kitchen instruction. The available sources do not establish precisely where the 2,000-kueh order was missed, how overnight production or deliveries were rearranged, or what effect the incident had on staff morale or individual cafes. Those consequences should not be treated as documented facts.

Why the fix has to meet orders where they arrive

The operational challenge is to bring incoming orders into one checked production plan without requiring every hotel to adopt a supplier’s new portal. Hotel procurement systems can include their own approval routes, supplier catalogues and invoice reconciliation; asking buyers to enter the same order again would add another hand-off rather than remove one.33

Extracting products, quantities and delivery dates from incoming documents, then matching differing product names before tallying demand, is a sensible way to address that challenge. Staff would still need to verify uncertain entries and check an apparently absent repeat order with the customer. The provided reporting does not independently confirm that HarriAnns’ system has every one of these features, including alerts or matching across more than 120 product codes.

What changed—and what the timeline cannot settle

The Straits Times reports that Knovel Engineering began developing HarriAnns’ platform around July 2025. It cost less than S$50,000 to develop, with Enterprise Singapore’s Enterprise Development Grant covering half.2 Separate reporting in February 2026 said an AI management system had already reduced the work from receiving orders through invoicing and kitchen instructions from a full day to one or two hours. HarriAnns also reported fewer errors and missed orders and said more accurate quantities helped it allocate resources and take additional orders.47

That February account describes a system introduced in mid-2025, while the later report dates development of the Knovel platform to around July 2025.247 The sources do not clarify whether these accounts refer to phases of one rollout or to different systems. A first rollout in June 2026 therefore cannot be stated with confidence.

A heritage business modernising its back office

Prime Minister Lawrence Wong cited HarriAnns, whose beginnings date to the 1940s, in his Budget 2026 discussion of helping businesses adopt AI-enabled solutions.1 The grants are distinct: the reported funding for development of this platform was the Enterprise Development Grant; the Budget discussion also refers to the Productivity Solutions Grant as support for digital adoption.12

HarriAnns’ case is less about changing how traditional kueh is made than about making sure confirmed demand reaches the people making it. The reported time savings are significant, but the lasting test is whether the ordering process reliably catches omissions while leaving staff able to verify exceptions.247