It’s been three years since ChatGPT introduced generative AI to the masses. Suddenly, everyone is using AI to find their city’s best burgers (reasonable), ask for a little romantic advice (questionable) or diagnose medical ailments (get a second opinion).

Businesses have also eagerly embraced the new technology. According to a global McKinsey survey, 88% of organizations are reporting using AI for at least one business function, up from just 55% in 2023. This is, in part, because it’s become more widely available. The cost of AI systems has decreased by about 30% annually over the past three years while energy efficiency has increased by 40%.

Private investment in AI in the U.S. reached $109.1 billion in 2024. In the first half of 2025, investment in information processing equipment in the U.S. grew by more than 20% year over year, driven by strong corporate demand and favorable tax treatments, according to the Mastercard Economic Institute’s Economic Outlook 2026 report. The report also anticipates deeper AI integration and targeted fiscal stimulus as key drivers of global growth.

Still, many people remain cautious. Although nearly all businesses use some form of AI, the vast majority of those systems are still in the experimentation or piloting phase. This is especially true in the payments realm, where a single breach of trust can have a longstanding effect on both a customer’s bank account and an organization’s reputation. As has been the case throughout the history of payments, consistent and rigorous security is essential.

Here are a few ways AI has made an impact in the financial world in 2025.

AI has always done its best work by combing through reams of data and uncovering patterns that may not be immediately obvious to humans. The next step in its development is using what it has learned to “reason”: breaking down tasks into structured, actionable steps to act automatically on a user’s behalf. This is known as agentic AI.

Imagine you are planning to travel to London next month and need to book a flight. An earlier version of AI may just present you with a list of flights to choose from. However, an AI agent would look through multiple websites, factor in your specific preferences, such as time, cost, seat location or airline on–time rating and honor your personal airline loyalty or credit card points. Then it goes on to select the best flight for you and book it on your behalf. It will even remember all your unique demands for the next flight you book.

Companies are using agentic commerce agents for everything from booking this type of business travel to stocking their shelves.

There are obvious questions about how these systems work, especially if they go wrong and, say, leave a business with 300 extra cases of toilet paper and no way to return them. Agentic commerce needs secure guardrails and permissions in place to do its job without interference from bots. In April, Mastercard launched Agent Pay, a new program to drive agentic commerce responsibly, and more recently introduced an acceptance framework that enables merchants to support agentic payments with confidence and trust. It’s expected that this technology will only continue to grow in 2026, mostly likely into multi-agent systems, a team of AI agents working together to accomplish a single task.

In the glory days of the American office, one of the perks of being an executive was having an assistant to do all the boring administrative tasks, leaving you free to do serious, important work (and take three-hour lunches). The office as we knew it may be gone, but now everyone in workplaces large and small can have an assistant, or at least a suite of AI apps.

AI is especially helpful in sales and attracting clients, said Jacqui Jones, who owns a marketing firm in Birmingham, Alabama, during a Microsoft Digital Doors webinar this summer. According to Jones, by analyzing sales patterns, AI tools can help small businesses see 50% more qualified leads and produce 20-30% higher conversion rates.

Jones also uses AI tools to take notes during meetings, schedule appointments, send out mass emails and check her correspondence for tone and accuracy. This increases her efficiency at work, and she is not alone: She reports that 78% of business owners say that AI has helped them save time.

But Jones stresses that AI is a tool, and like all tools, it doesn’t run effectively without human oversight.

The Mastercard Center for Inclusive Growth and data.com's Artificial Intelligence to Accelerate Inclusion Challenge encouraged startups to develop and scale AI solutions that advance inclusion and economic empowerment.

One of the winners was Colombia's Quipu, which provides microloans for unbanked business owners. In Latin America, approximately one-third of business owners are part of the informal economy that operate  without bank accounts. Although they’re able to use digital wallets to keep track of their finances, their lack of a banking history makes it hard to build credit and get loans to expand.

“When you’re born poor,” observes Quipu’s founder, Mercedes Binart, “it’s hard to build credit.”

Quipu determines the creditworthiness of its loan applicants by using AI to search through more non-traditional but equally pertinent financial data, including social media posts, inventory and real-time transactions. Since 2021, Quipu has distributed $5 million in loans, helping 26,000 Colombian entrepreneurs, half of whom are women. Now Bidart hopes to license Qiupu’s AI technology to help small businesses across Latin America.

It’s a great example of how AI can make the world a better place. “AI holds extraordinary potential, but its true power is realized only when everyone can access and benefit from it,” says Uyi Stewart, the Center’s vice president for Inclusive Innovation and Analytics. “By breaking down digital and informational divides for underserved communities, we unlock a future where innovation is inclusive, economic growth is amplified and opportunities are shared by all.”

Another obvious question about AI is, if all these wonderful tools are available to everyone, doesn’t everyone also include hackers and scammers? For a business that relies on e-commerce and digital payments, a single breach can have long-term effects: loss of trust, loss of customers.

“When researchers find security vulnerabilities with AI, that’s great because we can fix them,” said researcher Mikko Hyppönen in a keynote address at the Black Hat cybersecurity conference in Las Vegas in August. “When attackers do the same, that's awful.”

AI is also, unfortunately, making cybercriminals more effective. AI-powered tools let them quickly and easily create convincing phishing emails, texts and social media posts as well as both audio and video deepfakes. In a Mastercard-commissioned poll of 13,000 consumers worldwide, AI-generated fake content was the collective number-one fear, and only 13% were confident in their ability to identify it.

The silver lining, according to experts, is that attackers have mostly been using AI to update old scams. Cybersecurity firms, meanwhile, have been working to get ahead by incorporating AI into their own products and making them more affordable and widely available. And the role of humans to support AI by providing it with real-life context and programming new rules can’t be underestimated.

“I’m continually surprised by the audacity and imagination of the fraudsters,” says Vince Haulotte, director of market delivery in Mastercard’s fraud and risk decisioning business. “There are always new fraud trends, so we have to keep building new solutions to stay ahead of them. Our work never stops.”