The two most valuable AI companies on earth shipped products for enterprise deployment within eight days of each other this month.
Anthropic launched Ode on July 15, a consulting organization built around forward-deployed engineers who embed inside companies to make Claude work in production. OpenAI launched Presence on July 22, an enterprise platform for deploying and managing AI agents across customer-facing and internal workflows, also staffed by forward-deployed engineers who lead every implementation.
Both companies are addressing the work of getting AI to function inside the mess of policies, permissions, legacy systems, and human workflows that make up a real business.
That convergence makes deployment support worth a closer look.
What Presence actually does
Presence starts with a deployment engagement. You cannot sign up for it online. OpenAI has not published pricing. Every deployment starts with OpenAI’s own engineers sitting down with your team, identifying a specific workflow, connecting the agent to your internal systems, defining what it can and cannot do, testing it against edge cases, and then monitoring it after launch.
Each agent gets a defined job. Resolve a billing dispute. Handle an insurance claim. Process an employee IT request. The agent receives only the data and system access required for that task. Your team decides which actions the agent handles alone, which require approval, and when a human takes over.
Before anything reaches production, teams run simulations. The system tests how the agent handles routine requests, unusual scenarios, and policy changes. Graders evaluate whether it followed the rules, used the right tools, and escalated when it should have. After launch, the system keeps monitoring. When something drifts, it flags the issue and proposes a fix that gets tested against the current version before rolling out.
OpenAI already runs this system on its own customer support line at 1-888-GPT-0090. According to the company, it resolves 75 percent of inbound issues without a human ever getting involved. Over a 10-day period, their automated improvement loop cut human handoffs by another 15 percentage points (VentureBeat, July 22, 2026).
Those numbers come from OpenAI. Independent verification is still lacking. The company describes testing the system on its own support work before selling it.
The Palantir delivery model
If the delivery model sounds familiar, it should. Forward-deployed engineers embedding inside client organizations to make proprietary software work in complex environments is the model Palantir pioneered years ago for government and defense contracts.
OpenAI and Anthropic both adopted a playbook suited to a similar problem. Enterprise software that touches real operations, real compliance requirements, and real data requires people who understand the existing workflow and can wire things together with the customer.
For years, the narrative was that intelligence would scale through APIs: build a smarter model, expose an endpoint, let customers figure out the rest. Cisco data shows how much work remains after access: 85 percent of enterprises are now piloting AI agents, but only 5 percent have shipped them to production. Presence and Ode address the operational work involved in closing that gap.
What this means if you run a team
Three things to take from this convergence.
First, assess deployment readiness alongside model capability. OpenAI’s investment in forward-deployed engineers gives your team a reason to examine the implementation work needed around an API.
Second, ask who owns deployment in your organization beyond vendor selection and the pilot. Name the person responsible for workflow redesign, permissions, escalation rules, monitoring after launch, and updates when policies change. Assign that responsibility before moving the demo into production.
Third, budget for the boring part. The exciting part of AI is the capability. The expensive part is making it work inside your business. Presence and Ode both exist because companies kept buying API access and then discovering that nobody on the team knew how to turn it into a working system. The forward-deployed engineer model is a direct response to that pattern.
The convergence is the signal
Both competitors are investing in deployment support within days of each other. That is a reason to examine what support your own team needs.
OpenAI and Anthropic are shipping operations products to help enterprise customers get through deployment. Their investments address the practical work customers have struggled to complete.
BBVA in Mexico, SoftBank in Japan, and IAG in Australia are among the first organizations evaluating Presence. They are evaluating a way to get existing models into the workflows their teams actually run.
If you are a business leader planning AI adoption, budget for mapping workflows, defining permissions, testing edge cases, monitoring performance, and updating the system when the business changes. The companies that make the technology are investing in that work too.
That work determines whether the deployment can function reliably inside your business.