Square’s new AI features include a website and restaurant menu generator

Square, the financial services platform, is embracing generative AI in a very visible way.

After announcing earlier this year that it would bring AI features to drive retail sales, Square this morning took the wraps off of new ten — count ’em, ten — generative AI capabilities focused on customer content creation, onboarding and setup.

All are available as of today, albeit some gated behind Square’s beta program

The push, it might be said, is an attempt by Square — and parent company Block — to reinvigorate the Square platform after a difficult, downward-trending year and change.

Revenues from Block’s Cash App, the peer-to-peer payments service, have declined steeply. Meanwhile, the buy now, pay later service Afterpay, which Block acquired in 2021 for $29 billion, has posted serious losses. Block’s Bitcoin revenue has fallen corresponding with the fall in the price of the cryptocurrency last year. And Square faces growing competition on multiple fronts. including from Fiserv’s Clover, Toast and Stripe.

Investors are displeased. Square stock has retreated around 30% so far in 2023, as Block founder Jack Dorsey prepares to take the reigns from former Square head Alyssa Henry.

So Square’s giving stockholders generative AI.

One of the new AI-powered features, Menu Generator, allows restaurants to create a “full menu” on Square in “just minutes,” the platform promises in a press release.

“This is a valuable tool in particular for new restaurant sellers who don’t have or aren’t ready to upload a menu during onboarding,” a Square clarified to me via email. “With Menu Generator, restaurants can now create a full menu on Square … with the flexibility to come back and change or update that menu later after they’ve finished other set-up tasks. This gives them — or any business looking to expand into food and drink offerings — valuable momentum when launching operations on Square.”

Given generative AI’s tendency to go off the rails, I’d be wary of using it to publish a menu — particularly considering that gross inaccuracies could land restaurants on the hook for lawsuits over false advertising. But Square emphasizes that the process — which relies on a range of third-party generative AI models including OpenAI’s GPT-3, GPT-3.5 and GPT-4 — isn’t fully automated and that customers are afforded the chance to make edits before a menu hits the web.

“All of our AI-generated content is offered as suggestions to sellers, giving them the ability to save as-is or make further edits,” the spokesperson said. “The human review process comes from sellers themselves, who approve or modify all copy created to ensure it helps them meet their business goals.”

Human review is also baked into Square’s new generative email copy feature, according to the spokesperson, which taps generative AI to create personalized email messages to customers. And it’s a component of Square’s newly-launched website copy generator, which writes headlines — and entire blogs — given a brief text prompt.

I asked Square whether it’d taken steps to ensure that copy from its website tool wouldn’t result in downranking by search engines who don’t look kindly on certain, obviously duplicative forms of AI-generated content. In response, the spokesperson pointed to Square’s partnerships and integrations with Google to “help small businesses overcome … discoverability hurdles” — but didn’t answer the question directly.

As a part of the raft of new generative AI capabilities, Square’s point of sales system can now auto-generate item descriptions for seller catalogs. Square pitches this as a massive time saver — but color me skeptical. eBay recently rolled out a similar item description generator and the feedback has been less than stellar, with users complaining about repetitiveness and fluff in the AI-generated descriptions.

When presented with these concerns, the Square spokesperson noted that item descriptions from the generator are created “based on inputs from sellers, including keywords to keep descriptions focused, and length and tone guidelines to avoid overly verbose or fluffy results.” But it remains to be seen if sellers adhere to these guidelines.

Elsewhere on the AI front (not necessarily generative AI, despite how Square refers to them in the press release), the Kitchen Display System, Square’s kiosk for restaurant order management, can now auto-assign menu items to kitchen categories and station screens. Square’s scheduling system, Appointments, can automatically import service names, descriptions, durations and prices during onboarding, meanwhile. And Square’s namesake point-of-sales platform now suggests items for sellers to adopt “based on insights about their business.” (It’s not clear exactly which insights — the press release doesn’t specify.)

Square Team Communication, which adds announcements and messaging to Square Team, Square’s app for employee scheduling and time tracking, is now able to generate and send out announcements to let employees know about new products and upcoming promotions. (The topic, length and tone are adjustable, Square says.) And Square Messages, Square’s business-customer messaging platform, can now suggest “sophisticated” AI responses, with the ability to personalize messages with replies that prepopulate customer names. (This upgrades Square Messages’ existing reply generator, which Square claims is generating 450,000 messages per month.)

In the press release, Saumil Mehta, Square’s head of point of sale, is quoted as saying that the new generative AI capabilities put Square “at the forefront of technology.”

“Square is uniquely positioned to be the technology partner that enables the most seamless, intuitive applications of AI,” Mehta said.

Me, I’m not so sure. But if the sheer breadth of today’s rollout is any indication, Square’s ambitious if nothing else.