Agentic Commerce, 18 Months In. Check out my latest research on the current state of agentic commerce: hivemind.capital/content/age… An honest take on where the market stands today, who is positioned to win, where value is accruing, and the whitespace opportunities that still exist. What’s been most revealing from conversations with agentic commerce leaders the past few months is the disparity between the buzzy headlines vs the actual strategies that companies are taking to build defensible moats in this space. There’s a real gap between noise vs action. Agentic commerce feels like less of a greenfield opportunity and more of an extension of the existing commerce/payments market today, with most value still expected to accrue to the existing leaders who already own the end-user relationship, the merchant network, or the settlement rails. I'd encourage builders to look at the bottlenecks that are still hindering growth in this space: - KYA, agent identity, authorization - Agent governance, policies, and controls are still quite limited and simple vs trad corporate card controls that offer far greater customization and granular policy. Agentic payments products need to first reach feature parity with what exists in trad commerce for users to consider adopting. - Merchant readiness and acceptance - Many digital storefronts are not machine-readable, and a meaningful share of merchants block agents outright. Merchants rightfully have concerns about agent fraud protection, security, and liability (i.e. do they incur the chargeback liability when an agent buys something it shouldn’t have?). Fragmented protocols, standards, and identity solutions also give merchants reasons to wait until some of the dust settles and there’s a clearer path forward for them to engage with agents. Startups can build merchant solutions that alleviate some of these risks and headaches for merchants. - B2B - A lot of the data that B2B agents need to build powerful enterprise solutions sits trapped in ERP custom fields, procurement email threads, and PDF contracts. Tackling the B2B opportunity requires first building the pipes to collect this data at scale before existing workflows can be automated. Whoever does this effectively will benefit from the B2B flywheel: more transactions, better accuracy, more spend authority, higher switching costs. Opportunity exists for corp spend mgmt fintechs, vertical-specific startups, and enterprise agentic policy/security/knowledge graph startups to all compete for market share. Agents will only transact at scale once consumers, merchants, and finance teams trust them with real spending authority. The rules, controls, and data behind that trust are still being developed, and the companies focused on this will shape how agentic commerce works once the volume arrives. If you’re building or investing in agentic commerce, please reach out. I’d love to hear from you!

Building off my last couple posts, here is a breakout of the agentic commerce stack by use case - B2C, M2M, B2B. B2C is furthest along, M2M is the 0-1 opportunity, and B2B has the most concrete path to adoption and the most whitespace. B2C - Currently driving 63% of total AC revenue ($3.74B). But this is mostly AI-driven discovery, not actual agent-executed checkout yet. Interestingly, AI-referred traffic has a 42% higher purchase conversion than trad search. Incumbents dominate at the consumer app layer; the current bottleneck for agent checkout is getting merchants agent-ready. Startups @nekuda_ai and @Nevermined_ai are tackling this. M2M - New, early, and mostly experimental today. x402 and MPP processed ~$1M TPV last month with $0.06 avg tx size. The M2M business model is still being determined. x402 (Coinbase/Cloudflare) currently leads, routing transactions to Cloudflare for metering (L2) and Base for settlement (L1), among other providers. But x402 settles every payment individually, which strains micro-transaction economics. MPP addresses this with streaming "sessions" and Nevermined batches payments through prepaid credits. It’s still too early to say Coinbase and Cloudflare have won the high-value layers of the M2M stack. B2B - Clearest path to near-term revenue with massive TAM. 40% of B2B buyers are already using agents in purchasing flows with $15T in agent-led B2B purchases projected by 2028. Whitespace exists across the app layer, agent discovery, identity/auth, and commerce protocols. B2B moats will form through a data flywheel, and the market will verticalize. An agent tied to an employee, role, and policy engine generates company-specific reconciliation and invoice-matching data with every purchase. That data improves accuracy, earning the agent more spend authority over time and raising switching costs. Because B2B workflows are bespoke, we’ll likely see 1-2 leaders emerge per vertical, like we’ve seen in vertical AI and SaaS. Currently: - @Visa @Mastercard are prioritizing B2C, but also building for M2M and B2B. - @coinbase @Cloudflare are focused on M2M. - @tryramp @brexHQ @airwallex are targeting B2B. - @stripe is building across all 3. Looking ahead: - Card networks will prioritize all 3 use cases, becoming universal card & stablecoin settlement networks. - Stripe will further define and bundle its stack by use case. - B2B whitespace will verticalize and thin out as competition heightens between 1) fintechs that already hold corporate spend relationships, 2) startups building vertical niches, and 3) startups with stronger agent policy and identity products.

My last post mapped the agentic commerce stack into 6 layers. Across these layers, value accrues in a barbell: the app layer at the top, merchant acceptance & settlement at the bottom. This post covers how the major players are scaling across those layers to compete for market share, with each one already dominating a high-value layer and racing to control more of the end-to-end transaction flow to gain an early-mover advantage. Some takeaways: