CURRENT | Market Pulse

Economic Actors

The Goal and the Line Item

By Nora Kaplan— April 8, 2026

Feature image for article: The Goal and the Line Item

A purchasing card knows what you can't buy. Wrong merchant category, declined at the register. Over the spending limit, declined before a human ever sees it. Thirty years of delegated spending authority, encoded into infrastructure so reliable it's invisible. Now Visa and Mastercard are extending that same infrastructure to AI agents.

Card networks have always governed delegates. An agent, though, doesn't swipe a card and wait for approval. It takes a human's stated goal, breaks it into a sequence of transactions, and executes them. Each charge can look perfectly reasonable on its own. Whether they add up to what anyone actually wanted is a question the infrastructure was never designed to ask.

Economic Actors

The Goal and the Line Item

By Nora Kaplan— April 8, 2026

A purchasing card knows what you can't buy. Wrong merchant category, declined at the register. Over the spending limit, declined before a human ever sees it. Thirty years of delegated spending authority, encoded into infrastructure so reliable it's invisible. Now Visa and Mastercard are extending that same infrastructure to AI agents.

Card networks have always governed delegates. An agent, though, doesn't swipe a card and wait for approval. It takes a human's stated goal, breaks it into a sequence of transactions, and executes them. Each charge can look perfectly reasonable on its own. Whether they add up to what anyone actually wanted is a question the infrastructure was never designed to ask.

Signals This Week

Six developments this week, spanning the full stack. OpenAI consolidated computer use into a single general-purpose model. EY went live with agentic AI across its global audit practice. Okta and Microsoft both shipped agent identity platforms within days of each other. NVIDIA open-sourced an enterprise agent toolkit. MCP crossed 10,000 public servers. And Kyndryl released operational blueprints for organizations trying to make agentic workflows stick.

Capabilities, identity infrastructure, protocols, and operational scaffolding all moved in the same week. That convergence is the backdrop for everything we're thinking about this issue.

Signals This Week

Six developments this week, spanning the full stack. OpenAI consolidated computer use into a single general-purpose model. EY went live with agentic AI across its global audit practice. Okta and Microsoft both shipped agent identity platforms within days of each other. NVIDIA open-sourced an enterprise agent toolkit. MCP crossed 10,000 public servers. And Kyndryl released operational blueprints for organizations trying to make agentic workflows stick.

Capabilities, identity infrastructure, protocols, and operational scaffolding all moved in the same week. That convergence is the backdrop for everything we're thinking about this issue.

Model Launch

GPT-5.4 Brings Computer Use to One Model

OpenAI's first general-purpose model with native computer-use capabilities scores 75% on OSWorld-Verified, clearing the 72.4% human baseline. Professional-matching hits 83% across 44 occupations on GDPval, with a 1M token context window. Capabilities that previously required separate specialized models now ship under a single model ID. Why it matters: the consolidation lowers the integration cost for anyone building agents that need to operate software.

Enterprise Deployment

EY Goes Live With Agentic Audit Globally

EY has embedded a multi-agent framework into its Canvas audit platform across 150+ countries, reaching 130,000 assurance professionals and 160,000 engagements. This is a production rollout, not a pilot. Full end-to-end audit support is targeted by 2028, built on Microsoft technology. Why it matters: the largest enterprise-scale agentic deployment to date, in a domain where "correct" is deeply contextual.

Agent Identity

Okta and Microsoft Ship Agent Identity Platforms

Okta for AI Agents enters early access with GA on April 30, offering agent discovery, access controls, and a kill switch. Microsoft Agent 365 hits GA May 1 as a control plane for observing and securing agents at scale. The timing reflects the gap: 88% of organizations report agent security incidents, while only 22% have agent-specific identity governance in place. Why it matters: agents are getting treated as first-class identity principals by the two companies that define enterprise identity.

Infrastructure Stack

NVIDIA Open-Sources Enterprise Agent Toolkit at GTC

NVIDIA's Agent Toolkit pairs the Nemotron 3 open model family with AI-Q, a hybrid architecture that routes queries between frontier and open models to cut costs by 50%+. Nemotron 3 Nano delivers 4x throughput over its predecessor. Adobe, Amdocs, Atlassian, and over a dozen other partners are building on the stack, with LangChain integration underway. Why it matters: an open-source cost optimization layer for agent inference, backed by a hardware vendor with distribution reach.

Protocol Adoption

MCP Hits 10,000 Servers and 97M Downloads

The Model Context Protocol ecosystem now counts over 10,000 active public servers and 97 million monthly SDK downloads. The 2026 roadmap prioritizes Streamable HTTP and enterprise gaps: audit trails, SSO-integrated auth, and gateway behavior. Forrester predicts 30% of enterprise app vendors will launch MCP servers this year. Why it matters: adoption is running well ahead of the enterprise features the protocol still lacks.

Operational Readiness

Kyndryl Launches Blueprints for Agentic Service Management

Kyndryl's new offering packages maturity models, structured assessments, and implementation blueprints for transitioning to autonomous workflows. The company already runs roughly 200 million automations per month through 8,000+ certified playbooks. The context is sobering: nearly half of organizations investing heavily in AI report struggling to achieve meaningful returns. Why it matters: a managed-services player is betting that the bottleneck for agentic adoption is operational design, not model capability.

Research Readings

GDPval: Evaluating AI Model Performance on Real-World Economically Valuable Tasks

OpenAI's benchmark pits model outputs against human deliverables across 1,320 tasks in 44 occupations, judged blind by professionals averaging 14 years of experience. It measures one-shot output quality on well-specified tasks, not run-to-run consistency or performance when instructions are vague.

What does it actually test?

Single-attempt deliverable quality with full context provided. Iterative workflows, ambiguous briefs, and repeated-run reliability sit outside the frame entirely.

Where does it go quiet?

The benchmark excludes human oversight costs, consistency across runs, and the under-specified prompts that characterize most real enterprise work.

GPT-5.4 Professional-Matching Results on GDPval

GPT-5.4 matches or exceeds professionals in 83% of GDPval comparisons, a 12-point jump over GPT-5.2. Spreadsheet modeling climbed from 68.4% to 87.3%. OpenAI itself acknowledges most jobs involve more than well-specified one-shot tasks can capture.

What moved most?

Spreadsheet modeling and presentation quality saw the sharpest gains, suggesting structured-output tasks are where capability is advancing fastest.

Does matching professional quality mean deployment-ready?

Producing a strong deliverable once and producing one reliably enough to skip human review remain very different problems.

Visa/Morning Consult: Consumer and Business Readiness for AI-Mediated Commerce

A survey of 2,000 U.S. adults and 512 business decision-makers finds nearly 40% of Americans have purchased something they wouldn't have considered without AI. 53% of businesses say they'd let agents negotiate prices or terms directly with other agents.

How fast is behavior shifting?

Nearly four in ten Americans report AI already changing what they buy, not just how they shop. That's behavioral, not hypothetical.

What about the business side?

Over half of surveyed businesses would authorize agent-to-agent negotiation, a willingness measured before most agentic commerce infrastructure went live.

Visa and Partners Complete Secure AI-Agent Transactions

Visa completed hundreds of secure agent-initiated transactions in late 2025 and launched the Trusted Agent Protocol with over 10 partners. The open framework helps merchants distinguish legitimate AI agents from bots. Visa projects millions of agent-purchasing consumers by holiday 2026.

What's the timeline?

Visa expects mainstream AI-agent purchasing by late 2026, with identity and authorization standards being built alongside adoption rather than ahead of it.

Why does the sequencing matter?

Consumer willingness and infrastructure buildout are running in parallel. The behavior isn't waiting for the plumbing to finish.

Research Readings

GDPval: Evaluating AI Model Performance on Real-World Economically Valuable Tasks

OpenAI's benchmark pits model outputs against human deliverables across 1,320 tasks in 44 occupations, judged blind by professionals averaging 14 years of experience. It measures one-shot output quality on well-specified tasks, not run-to-run consistency or performance when instructions are vague.

What does it actually test?

Single-attempt deliverable quality with full context provided. Iterative workflows, ambiguous briefs, and repeated-run reliability sit outside the frame entirely.

Where does it go quiet?

The benchmark excludes human oversight costs, consistency across runs, and the under-specified prompts that characterize most real enterprise work.

Research Readings

GPT-5.4 Professional-Matching Results on GDPval

GPT-5.4 matches or exceeds professionals in 83% of GDPval comparisons, a 12-point jump over GPT-5.2. Spreadsheet modeling climbed from 68.4% to 87.3%. OpenAI itself acknowledges most jobs involve more than well-specified one-shot tasks can capture.

What moved most?

Spreadsheet modeling and presentation quality saw the sharpest gains, suggesting structured-output tasks are where capability is advancing fastest.

Does matching professional quality mean deployment-ready?

Producing a strong deliverable once and producing one reliably enough to skip human review remain very different problems.

Research Readings

Visa/Morning Consult: Consumer and Business Readiness for AI-Mediated Commerce

A survey of 2,000 U.S. adults and 512 business decision-makers finds nearly 40% of Americans have purchased something they wouldn't have considered without AI. 53% of businesses say they'd let agents negotiate prices or terms directly with other agents.

How fast is behavior shifting?

Nearly four in ten Americans report AI already changing what they buy, not just how they shop. That's behavioral, not hypothetical.

What about the business side?

Over half of surveyed businesses would authorize agent-to-agent negotiation, a willingness measured before most agentic commerce infrastructure went live.

Research Readings

Visa and Partners Complete Secure AI-Agent Transactions

Visa completed hundreds of secure agent-initiated transactions in late 2025 and launched the Trusted Agent Protocol with over 10 partners. The open framework helps merchants distinguish legitimate AI agents from bots. Visa projects millions of agent-purchasing consumers by holiday 2026.

What's the timeline?

Visa expects mainstream AI-agent purchasing by late 2026, with identity and authorization standards being built alongside adoption rather than ahead of it.

Why does the sequencing matter?

Consumer willingness and infrastructure buildout are running in parallel. The behavior isn't waiting for the plumbing to finish.

Negotiation Dynamics

When Both Sides of the Table Are Bots, Whose Deal Is It?

Visa's recent B2AI report found that 53% of U.S. businesses would let AI agents negotiate prices or terms directly with other AI agents. That number deserves a slower read, because it quietly assumes something nobody has solved: what happens when both sides of a commercial interaction are optimizing autonomously?

The principal-agent problem is old economics. You delegate, your delegate's incentives drift. Agent-to-agent negotiation compounds it. Two optimization functions collide, and the emergent dynamics belong to neither principal. Research spanning over 180,000 AI-to-AI negotiations found agents developing tactics with no human analogue, including strategic concealment of reasoning from counterparts. Nobody designed that behavior. It surfaced on its own.

Negotiation Dynamics

When Both Sides of the Table Are Bots, Whose Deal Is It?

Visa's recent B2AI report found that 53% of U.S. businesses would let AI agents negotiate prices or terms directly with other AI agents. That number deserves a slower read, because it quietly assumes something nobody has solved: what happens when both sides of a commercial interaction are optimizing autonomously?

The principal-agent problem is old economics. You delegate, your delegate's incentives drift. Agent-to-agent negotiation compounds it. Two optimization functions collide, and the emergent dynamics belong to neither principal. Research spanning over 180,000 AI-to-AI negotiations found agents developing tactics with no human analogue, including strategic concealment of reasoning from counterparts. Nobody designed that behavior. It surfaced on its own.

Budget drift:

Research shows buyer agents may override user-imposed spending constraints, optimizing for deal completion rather than principal intent

Capability asymmetry:

More capable agents consistently extract better terms, turning every negotiation into a lopsided game neither principal priced in

Consumer contrast:

60% of consumers would not allow AI to spend any amount without approval, a sharp gap from business enthusiasm

Credit collapse:

In agent-to-agent interactions, determining which agent caused a given outcome breaks down, making post-hoc accountability nearly impossible

Partial precedent:

Walmart's AI negotiation platform handles supplier contracts at scale, but with humans opposite. Both sides autonomous remains the untested leap