Anthropic and OpenAI Race to Embed AI Agents on Wall Street

In a span of just three days in early May 2026, the two most valuable AI companies on the planet fired their opening salvos in a new war — not over model benchmarks, but over who gets to embed AI agents inside Wall Street's most critical workflows.

On May 4, Anthropic announced a $1.5 billion joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs to create a forward‑deployed engineering firm, as reported by Fortune. The next day, the company shipped 10 ready‑to‑run finance agent templates alongside Claude Opus 4.7, its most capable financial model yet. Twenty‑four hours later, OpenAI unveiled its $4 billion Deployment Company and an expanded partnership with PwC to build AI agents for the CFO's office.

"Enterprise demand for Claude is significantly outpacing any single delivery model," Anthropic CFO Krishna Rao said, according to InvestmentNews. The message was unmistakable: the next phase of frontier AI isn't about model capabilities — it's about deployment at scale.

Anthropic's announcement, detailed on the company's blog, covers the full sweep of financial services work. The 10 agent templates span research functions — pitch building, meeting preparation, earnings review, financial model building, and market research — alongside operations roles like general ledger reconciliation, month‑end close, statement auditing, and KYC screening.

Each template ships as a plugin in Claude Cowork or Claude Code, and as a cookbook for Claude Managed Agents that can run autonomous, long‑running sessions with full audit trails. Claude also now integrates directly with Microsoft Excel, PowerPoint, Word, and Outlook, carrying context across all four applications simultaneously.

The data ecosystem is equally ambitious. New connectors to Dun & Bradstreet, Verisk, Third Bridge, SS&C Intralinks, and Moody's give agents governed, real‑time access to market data on over 600 million companies. Claude Opus 4.7 leads the Vals AI Finance Agent benchmark at 64.37%, and is already in production at JPMorganChase, Goldman Sachs, Citi, AIG, and Visa, per.¹

OpenAI's countermove came through two channels. First, the OpenAI Deployment Company — backed by more than $4 billion from 19 investment firms led by TPG — will embed Forward Deployed Engineers directly inside client organizations. OpenAI also agreed to acquire Tomoro, an applied AI consulting firm bringing roughly 150 experienced FDEs from day one, whose clients include Tesco and Virgin Atlantic.

Second, the expanded PwC collaboration targets the CFO's office specifically, building AI agents for planning, forecasting, reporting, procurement, payments, treasury, tax, and accounting close. OpenAI's own finance team serves as "customer zero" — testing the procurement agent internally before broader rollout.

"Finance is at an inflection point, where organizations are moving from process efficiency to intelligent, decision‑centric operations," said Tyson Cornell, US Advisory Leader at PwC, in the CFO Dive report. OpenAI CFO Sarah Friar added that the opportunity "is far bigger than efficiency — it's about giving finance leaders the tools to operate with greater foresight, agility, and strategic impact."

Why the sudden urgency? Both labs have identified the same structural problem: AI models are improving faster than companies can deploy them. "If we look at the models that have come out the last two years — the AI revolution — the quality of these models is going up and up. The ability for companies to deploy them is not keeping up. That gap is increasing," Sanjay Subramanian, PwC Partner and US & Global Anthropic Alliance Lead, told.⁴

Jason Cutler, SVP of Anthropic Consulting and Engineering at Caylent, described a three‑phase enterprise maturity model in the same report: most customers remain stuck at phase one — basic training and enablement. "You can bring AI into an existing process, but you're not getting the full advantage of AI yet. You really have to recreate the process alongside AI and your employees to really benefit," Cutler said. The forward‑deployed engineering model — embedding AI engineers directly inside client organizations — is designed to push companies past this bottleneck.

Not every financial workflow is ready for agentic AI, and the companies involved are candid about the boundaries. The sweet spot is deterministic, back‑testable tasks. In one insurance underwriting deployment documented by,⁴ the cycle was compressed from 10 weeks to 10 days through a three‑phase deployment: backtesting against historical outcomes, co‑delivery with human oversight, and finally agents providing first‑pass deliverables with underwriters reviewing at checkpoints. Liability remains with humans — "it's too early to change that process," Sanjay Subramanian, PwC Partner and US & Global Anthropic Alliance Lead, told.⁴

Where agentic AI struggles: open‑ended, high‑variance tasks where the question space is unbounded. Subramanian offered a specific example to:⁴ "A supply chain company where they've got lots of parts that need to get fixed — if those parts are so diverse, the questions are so diverse, there's less precision around that outcome." Early numbers from OpenAI's internal deployment are promising — 5x contract processing throughput with the same headcount using Codex, and an IR‑GPT tool that managed 200+ investor interactions during a fundraise, per.⁴

The forward‑deployed engineering model creates an entirely new talent pipeline. OpenAI's acquisition of Tomoro alone brings 150 experienced FDEs, and Anthropic's joint venture with Blackstone and Goldman Sachs signals that private equity firms see AI deployment engineers as a scarce, high‑value resource. For developers with AI/ML skills, this means a new career track: embedded AI engineer, working inside banks, insurers, and asset managers to redesign workflows around agentic systems.

But the tooling gap is real. Brad Shimmin of Futurum Group noted to ⁴ that "even within highly regulated industries, where inaccuracies are not tolerated, generative and agentic AI promises to reinvent the way both consumers and financial professionals work with data." The challenge for builders is that financial services demand precision that general‑purpose AI models still struggle to deliver consistently — creating opportunities for developers who can build the middle layer of verification, governance, and domain‑specific tooling.

Both companies are racing toward IPOs that could reshape the technology landscape. Anthropic is valued at approximately $380 billion and OpenAI at $852 billion as of their most recent funding rounds, InvestmentNews reports. Both are among the companies pursuing trillion‑dollar‑plus public debuts in 2026.

"Finance took off relatively later, but we believe we're at the inflection point right now, about six months to one year behind coding," Lisa Crofoot, research product management leader at Anthropic, told.¹ She noted that less than a year ago Claude "could barely format a table without ref errors," and today it's doing senior‑analyst‑level work. The race to embed AI agents on Wall Street isn't just about winning financial services — it's about proving that agentic AI can work in the most demanding, regulated environments on earth before either company rings the opening bell.