At this stage, Agent Pay is what Mastercard calls \u201chuman in the loop\u201d technology with \u201cBook for me\u201d and \u201cConfirm and pay\u201d buttons that need to be clicked. <\/p>\n
But Mastercard is also planning a fully autonomous option, where an AI agent can buy things on your behalf. <\/p>\n
It will offer the ability to pre-set pricing parameters if you say, want an agent to join a queue for concert or sports event tickets but are wary of a Fifa World Cup-style dynamic pricing blowout.<\/p>\n
A key reason to use your credit card for online purchases is the full refund guarantee \u2013 in most cases \u2013 if something goes wrong. <\/p>\n
\u201cIf you buy something and it doesn\u2019t come, or if you buy something and it turns out to be a scam website, you have zero liability as a consumer,\u201d Simons says.<\/p>\n
That promise also includes \u201chuman\u201d in the loop Agent Pay transactions.<\/p>\n
What about a transaction made by a fully autonomous agent? Will those sorts of transactions have the same zero liability?<\/p>\n
\u201cYes they will,\u201d says Simons. <\/p>\n
\u201cThat\u2019s our brand promise. The important thing is going to be trust. If you\u2019re paying on a Mastercard, it doesn\u2019t matter if it\u2019s an agent or a \u2018future intent\u2019 agent or online or in a store, you have zero liability and that will absolutely stay in place.\u201d <\/p>\n
Last year, when the Herald polled various players, policies about AI agent purchases were fuzzy around the edges. Simons has crystalised her company\u2019s position.<\/p>\n
She adds that once autonomous agent buying is here, you\u2019ll also be able to review a list of your \u201cfuture intent\u201d purchases, be they a flight to Wellington as long as it\u2019s below $300 and before 9am, or a pair of black denim pants from a particular retailer if the price falls below $150. They can be tweaked or deleted.<\/p>\n
When will it go mainstream?<\/p>\n
Simons expects \u201chuman in the loop\u201d Agent Pay to go mainstream over the next nine months or so as more partners come onboard at every level.<\/p>\n
Mastercard is in talks with the other major banks (similar showcase transactions in Australia earlier this month also included Westpac, plus ASB parent CBA).<\/p>\n
On the LLM (large language model or AI chatbot) front, while both the Aussie and NZ transactions featured an agent made by Maincode (the company behind Australia\u2019s sovereign AI, Matilda), Simons says \u201cglobally we are working with all major players \u2013 including OpenAI, Microsoft, Google\u201d. <\/p>\n
Multiple merchants are also being lined up.<\/p>\n
Recognising agents<\/p>\n
While she talked up scam protection, Simons also said her firm\u2019s Agent Pay framework, which builds on its existing security token effort and Click to Pay technologies, is initiated with a \u201chandshake\u201d between the retailer and the agent, ensuring everything\u2019s above board.<\/p>\n
As well as preventing fraud, the \u201chandshake\u201d process is also important for reducing false-positives \u2013 or AI transactions being declined because an agent seems dodgy. <\/p>\n
Simons says an agent is very rapid \u2013 a characteristic shared by the bots deployed by scamsters. <\/p>\n
\u201cIt doesn\u2019t look like a normal human behaving in a normal way.\u201d<\/p>\n
Agent Pay helps to establish an AI agent is registered and legitimate, and that it is following a customer\u2019s authorised intent (which also helps protect a retailer from claims of unauthorised use by a card holder).<\/p>\n
She says retailers and service providers of all stripes have to prepare for a world where many \u2013 and ultimately most \u2013 visitors to their sites and apps are AI agents, who behave differently, and customers who, like herself, are now more likely to hit up an LLM like ChatGPT with a shopping query than hit Google.<\/p>\n
Mastercard can work with big retailers on their \u201cdiscoverability\u201d, Simons says.<\/p>\n
She adds, \u201cfor small businesses as well, it\u2019s not going to be about how many dollars you\u2019re spending on Google AdWords anymore. It\u2019s about, how do you become more discoverable for these agents\u201d?<\/p>\n
\u2018Addresses a genuine gap\u2019<\/p>\n
\u201cPeople are using AI tools to search, compare and shortlist purchases right now. Most of the commerce ecosystem cannot see when that is happening. From a merchant\u2019s system, an AI tool browsing a catalogue looks identical to a bot,\u201d says Deloitte NZ AI director Dr Amanda Williamson.<\/p>\n
\u201cFrom a bank\u2019s perspective, it looks like unusual activity. The concept of making AI-initiated transactions identifiable across the payment chain addresses a genuine gap.\u201d<\/p>\n
Williamson sees another pressure for retailers: pricing. A traditional search engine makes it easier for a person to find the best price. An AI agent will get endless quotes.<\/p>\n
\u201cBusiness models built on human effort constraints will face pressure,\u201d she says.<\/p>\n
\u201cBusinesses that are only legible to humans will become harder to find. When software does the browsing and purchasing, it interacts with structured data, pricing logic and policies. It does not read a web page the way a person does. If your catalogue is not readable by an AI tool, your products do not get surfaced.\u201d<\/p>\n
AI expert Dr Amanda Williamson has tried out many early AI agents \u2013 and found them either clumsy and slow or prone to out-of-control spending on your behalf. <\/p>\n
Room to finesse<\/p>\n
The rise of agentic AI has seen various companies release agents that can buy things on your behalf over the past few months.<\/p>\n
\u201cI have been experimenting with agentic AI shopping tools since they became available. They can be extremely erratic,\u201d Williamson says.<\/p>\n
The agents that require human confirmation are too slow and clumsy to be worthwhile. <\/p>\n
And fully autonomous agents have been outright perilous.<\/p>\n
\u201cI asked an AI to find me a T-shirt. It found 50 identical ones, added them all to my cart, and tried to process the payment. Had I not caught it in time, I would be explaining to my family why I spent over $800 on \u2018I [heart] AI\u2019 shirts,\u201d Williamson says.<\/p>\n
Simon\u2019s Agent to Pay transactions were short and snappy.<\/p>\n
Mastercard has built on existing standards and technologies, including tokenisation (replacing a 16-digit number with cryptographically secure credentials ensuring each transaction is traceable and authenticated) Apple Wallet and Google Wallet, PassKey and Click to Pay.<\/p>\n
They were arguably too short and snappy. <\/p>\n
Not many will be comfortable picking upmarket Queenstown accommodation based on a choice of one of three thumbnails.<\/p>\n
And any penny-pinching Event Cinema goers, like your author, would like to log on to the theatre operator\u2019s Cinebuzz loyalty scheme to get a couple of dollars off the price of the tickets and a cheaper coke and popcorn. <\/p>\n
\u201cBuilding loyalty into those experiences will be really important. It will happen,\u201d Simons says. <\/p>\n
Filling in such features will ultimately happen on the app maker and retailer\u2019s side, but Simons says her firm\u2019s Agent to Pay framework makes things more transparent.<\/p>\n
Chris Keall is an Auckland-based member of the Herald\u2019s business team. He joined the Herald in 2018 and is the technology editor and a senior business writer.<\/p>\n","protected":false},"excerpt":{"rendered":"At this stage, Agent Pay is what Mastercard calls \u201chuman in the loop\u201d technology with \u201cBook for me\u201d…\n","protected":false},"author":2,"featured_media":286966,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[2],"tags":[6116,17901,365,137857,10001,4526,63059,1888,2082,78,807,16710,117,111,43,139,69,2744,25530,1243,21694,764,6035,2044,80],"class_list":{"0":"post-286965","1":"post","2":"type-post","3":"status-publish","4":"format-standard","5":"has-post-thumbnail","7":"category-new-zealand","8":"tag-agent","9":"tag-agents","10":"tag-ai","11":"tag-behalf","12":"tag-closer","13":"tag-easier","14":"tag-faster","15":"tag-first","16":"tag-for","17":"tag-in","18":"tag-kiwis","19":"tag-mastercard","20":"tag-new","21":"tag-new-zealand","22":"tag-news","23":"tag-newzealand","24":"tag-nz","25":"tag-pay","26":"tag-safer","27":"tag-shopping","28":"tag-step","29":"tag-tests","30":"tag-things","31":"tag-time","32":"tag-zealand"},"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/286965","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/comments?post=286965"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/posts\/286965\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media\/286966"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/media?parent=286965"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/categories?post=286965"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/nz\/wp-json\/wp\/v2\/tags?post=286965"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}