Biggest question in agentic commerce is: can an AI agent really capture what I want to buy? With Visa Intelligent Commerce (VIC), we call this the user’s intent. VIC Agent APIs require the intent , or purchase instruction, to prevent agent mistakes and avoid unnecessary disputes. But what if the agent miscaptures the user’s intent? How well can frontier models handle ambiguity with various sets of purchase instructions?
1/ We measured this with the Visa Agentic Commerce Benchmark: https://t.co/R8epQHkUg9. Turns out, we’ve got a long way to go, average success rate was 57.2% for 22 models and 15 test cases.
Huge shout out to our brilliant Visa AI Research team @mminaeib, @iamrachuri, and @ShimaaM_Ahmed that built this.
Anyone interested in pioneering research on the agentic economy send me a DM or email [email protected]
evals are so key! this will be instrumental for agents to hill climb on.
we are at nascent stages with agentic commerce but much like coding agents, i’m bullish that we will see drastic improvement.
as soon as we create the right incentives for every ecosystem participant in agentic commerce to benefit from increased agent delegated volume, enablement is just a matter of time.
kudos to @Visa!
We've been hard at work in all areas of agentic infrastructure. Building this benchmark was so fun!
Can't wait to share more as we continue to make progress with Visa Intelligent Commerce!
1/ Excited to introduce the Visa Agentic Commerce Benchmark, our first public benchmark designed by Visa Research to evaluate frontier model performance on shopping tasks.
https://t.co/2XFiTZQzf4
On Tuesday, Sierra announced Personal Agent Protocol with Meta, Genesys, Instinct, Rocket, Shopify, Stripe, and Walmart. The response has been incredible and today we’re publishing the first draft of the protocol at https://t.co/Qao9yEfzx4 and announcing 35 new design partners.
As personal agents take on more, our goal is simple: consumers have choice and control, and companies know when they’re dealing with an agent and who it represents.
This draft is a starting point. We’d love for companies, developers, and personal agent builders to read it, give feedback, and help us build an open standard together.
Great to see @openstandard launch OUSD today.
We are integrating OUSD into the Visa Stablecoin Platform and look forward to making it available to our partners.
Open Standard, $OUSD, is now live!
Open Standard is a new USD stablecoin with a very large network of participating partners (https://t.co/DLdavkVEL3). It's designed to be exceptionally secure, reliable, and developer-friendly. Importantly, most yield is shared directly with partners rather than being internalized by the stablecoin issuer.
Stripe is adopting OUSD in a number of core product flows.
Say hello to Apple Pay in India! 🇮🇳
We’re so excited to bring Apple Pay to India today, and make it easier than ever for our customers to make safe and secure purchases on Apple devices.
We're building the best stablecoin-backed card issuance platform in partnership with our friends at @Visa. Today we take one more step in that direction by launching the @Lightspark Cards Playground that allows you to design your card program. Try here: https://t.co/8b1JY4Mr5d
I’m leaving stripe to run Open Standard full time.
at bridge, we made stablecoins more accessible. but our impact was constrained by issuers' economics and incentives.
The world needs better money. we hope to remove these constraints.
AI can buy the couch, but do you trust it?
61% of U.S. consumers would when Visa handles the transaction - making Visa the #1 most trusted brand for agentic transactions.
Read more: https://t.co/LcPu5oEJ9Z
Excited for Visa to partner with Anthropic as they launch commerce blueprints to enable merchants to build and run agents on Claude.
Huge opportunity for next generation agent-operated merchants and excited to see what people build with it
https://t.co/P8p3fjx7qz