OpenAI has put a former AWS partner-sales leader, Jason Adae, in charge of AWS partnerships across EMEA as it tries to turn Amazon Bedrock availability for its models and Codex into a services-and-integration business for the channel. The immediate practical opportunity is real: AWS customers can run OpenAI models and Codex through Bedrock, apply the spending to existing AWS commitments, and use AWS-native identity, network, logging, and governance controls. But the partner program announcement still leaves out the commercial terms that determine whether this becomes a durable resale and managed-services business or mainly a consulting-led implementation market.
CRN first reported Adae’s move from AWS, where he spent eight years in EMEA partner roles, to OpenAI as director of AWS partnerships in EMEA. In the interview, Adae positioned Bedrock as the point where OpenAI’s enterprise push meets a customer base already committed to AWS, and urged AWS partners to join the recently launched OpenAI Partner Network.
That framing is more consequential than another executive hire. OpenAI is trying to solve the enterprise distribution problem that follows a successful developer product: customers may want Codex and frontier models, but they also want those tools wired into existing AWS accounts, data controls, procurement commitments, internal applications, and change-management programs. Partners are the route through that work.
AWS first introduced OpenAI models, Codex, and OpenAI-powered managed agents on Amazon Bedrock in limited preview on April 28, then made GPT-5.5, GPT-5.4, and Codex generally available on June 1. In July, AWS added the GPT-5.6 Sol, Terra, and Luna family to Bedrock. AWS says OpenAI pricing on Bedrock matches OpenAI’s first-party rates, while the usage counts against a customer’s existing AWS commitments.
That last detail is the commercial hook. A large enterprise with a committed-spend agreement cannot treat an OpenAI rollout as merely another software subscription if it can consume the budget already allocated to AWS. For CIOs and procurement teams, that removes a common obstacle: a new AI project no longer necessarily requires a separate vendor contract and a new funding line.
For AWS partners, however, this is not automatically a margin opportunity on model consumption. Neither AWS’s Bedrock announcements nor OpenAI’s partner-network launch identifies resale discounts, referral fees, deal-registration protections, usage rebates, or a standard compensation model for partners bringing OpenAI-on-Bedrock deals to market. The published material describes co-selling, deployment support, enablement, and recognition tiers, but it does not state who owns the customer contract or how a partner is paid when a customer buys tokens through AWS.
That omission matters. A partner can build a profitable practice around architecture, security, data integration, agent design, application modernization, managed operations, and user adoption. It cannot responsibly forecast a software-margin business from OpenAI’s announcements alone. AWS partners should read Adae’s pitch as a call to build implementation capacity first, rather than evidence that OpenAI and AWS have opened a conventional cloud marketplace resale program.
The program is structured around an acknowledged enterprise bottleneck. OpenAI’s own launch post argues that model capability is no longer the limiting factor; organizations need help choosing use cases, redesigning workflows, connecting systems and data, handling responsible deployment, and making employees adopt new processes. That is familiar territory for global systems integrators, regional consultancies, managed service providers, and specialist AWS partners.
Adae told CRN that Accenture, McKinsey, and Slalom had already committed to investing in OpenAI skills. OpenAI’s published launch roster includes major consulting, systems-integration, data, and technology firms, so the program is entering the market with established enterprise implementers rather than waiting to recruit a channel from scratch.
The harder question is whether smaller AWS partners can turn certification into differentiated business before the largest firms absorb the early enterprise projects. OpenAI’s stated goal of 300,000 certified consultants is enormous. If achieved, it increases the available delivery capacity quickly; it also makes basic accreditation less scarce. The partner advantage will come from industry-specific assets—secure coding workflows for financial services, migration accelerators for Windows-heavy environments, approved data connectors, governance templates, and operating models for long-running agents—not from a generic “OpenAI certified” badge.
OpenAI has not published the allocation of that $150 million between training, partner incentives, technical support, co-selling resources, or customer deployment funding. It also has not said how many firms will qualify for each tier or which benefits are exclusive. The headline investment is meaningful, but partners evaluating it should separate the announced total from the benefits available to their particular practice.
The 10 million number should not be read as “Codex now has 10 million weekly active users.” It is a combined total for Codex and ChatGPT Work, the newer agent-oriented experience inside ChatGPT for multistep office tasks. OpenAI has not publicly provided a product-by-product split for that combined count in the material available so far.
That distinction changes how partners should plan. Codex may be expanding beyond developers into spreadsheet work, reporting, research, and lightweight internal tools, but its technical roots still shape the implementation challenge. A customer deploying Codex across Windows developer endpoints may need source-control policies, repository permissions, IDE configuration, terminal restrictions, secrets management, code-review gates, logging, and a clear process for human approval before changes reach production.
AWS says Codex can be configured to use Bedrock inference through the Codex app, CLI, and IDE integrations including Visual Studio Code, JetBrains, and Xcode. For Windows-focused IT teams, Visual Studio Code is the obvious on-ramp, but the important work is not installing an extension. It is deciding which repositories and environments can be exposed to an agent, how credentials are delegated, what data can leave a development environment, and how an organization audits the actions the agent takes.
AWS’s Bedrock pitch is tailored to those questions. Its OpenAI integration includes AWS Identity and Access Management, PrivateLink, encryption, CloudTrail logging, and Bedrock guardrails; the newer GPT-5.6 rollout also describes in-Region inference and data-perimeter policies. Those capabilities do not make an agent deployment safe by default. They give a partner the AWS control plane needed to create separate accounts, permissions, network paths, logging rules, and approval boundaries around it.
This consolidation will make informal experimentation easier. It also increases the risk of organizations treating very different workloads as though they have the same controls. A ChatGPT Work task that summarizes files, a Codex task operating on a local repository, and a managed agent connected to business systems each have different identity, tool-access, retention, approval, and incident-response requirements.
The services opportunity therefore sits in the boundary-setting that vendors often compress into the word governance. A credible AWS/OpenAI partner will need to establish which agent can access which systems, where execution happens, which actions require approval, what gets logged, who can review those logs, how model changes are tested, and how costs are attributed to teams or applications. Those are recurring operational services, not one-off prompts or workshops.
OpenAI has supplied the partner program, AWS has supplied a procurement and control-plane path, and Adae has been hired to connect the two in EMEA. What remains unpublished is the piece partners normally need before betting heavily on a new vendor route: the precise economics. Until OpenAI and AWS disclose partner compensation, deal protection, regional coverage, and support commitments, the prudent play is to build Bedrock-based delivery capability around real customer workloads—and avoid assuming that certification alone comes with a revenue stream.
That framing is more consequential than another executive hire. OpenAI is trying to solve the enterprise distribution problem that follows a successful developer product: customers may want Codex and frontier models, but they also want those tools wired into existing AWS accounts, data controls, procurement commitments, internal applications, and change-management programs. Partners are the route through that work.
Bedrock turns OpenAI usage into AWS consumption
AWS first introduced OpenAI models, Codex, and OpenAI-powered managed agents on Amazon Bedrock in limited preview on April 28, then made GPT-5.5, GPT-5.4, and Codex generally available on June 1. In July, AWS added the GPT-5.6 Sol, Terra, and Luna family to Bedrock. AWS says OpenAI pricing on Bedrock matches OpenAI’s first-party rates, while the usage counts against a customer’s existing AWS commitments.That last detail is the commercial hook. A large enterprise with a committed-spend agreement cannot treat an OpenAI rollout as merely another software subscription if it can consume the budget already allocated to AWS. For CIOs and procurement teams, that removes a common obstacle: a new AI project no longer necessarily requires a separate vendor contract and a new funding line.
For AWS partners, however, this is not automatically a margin opportunity on model consumption. Neither AWS’s Bedrock announcements nor OpenAI’s partner-network launch identifies resale discounts, referral fees, deal-registration protections, usage rebates, or a standard compensation model for partners bringing OpenAI-on-Bedrock deals to market. The published material describes co-selling, deployment support, enablement, and recognition tiers, but it does not state who owns the customer contract or how a partner is paid when a customer buys tokens through AWS.
That omission matters. A partner can build a profitable practice around architecture, security, data integration, agent design, application modernization, managed operations, and user adoption. It cannot responsibly forecast a software-margin business from OpenAI’s announcements alone. AWS partners should read Adae’s pitch as a call to build implementation capacity first, rather than evidence that OpenAI and AWS have opened a conventional cloud marketplace resale program.
The $150 million program is an enablement commitment, not a channel price sheet
OpenAI formally launched the OpenAI Partner Network in June with a stated $150 million investment and a target of training and enabling 300,000 certified consultants by the end of 2026. The company says participating firms can progress through Select, Advanced, and Elite tiers based on sales performance, technical capability, co-sell engagement, and deployment experience. It also says future specializations will cover areas including Codex, cybersecurity, and agents.The program is structured around an acknowledged enterprise bottleneck. OpenAI’s own launch post argues that model capability is no longer the limiting factor; organizations need help choosing use cases, redesigning workflows, connecting systems and data, handling responsible deployment, and making employees adopt new processes. That is familiar territory for global systems integrators, regional consultancies, managed service providers, and specialist AWS partners.
Adae told CRN that Accenture, McKinsey, and Slalom had already committed to investing in OpenAI skills. OpenAI’s published launch roster includes major consulting, systems-integration, data, and technology firms, so the program is entering the market with established enterprise implementers rather than waiting to recruit a channel from scratch.
The harder question is whether smaller AWS partners can turn certification into differentiated business before the largest firms absorb the early enterprise projects. OpenAI’s stated goal of 300,000 certified consultants is enormous. If achieved, it increases the available delivery capacity quickly; it also makes basic accreditation less scarce. The partner advantage will come from industry-specific assets—secure coding workflows for financial services, migration accelerators for Windows-heavy environments, approved data connectors, governance templates, and operating models for long-running agents—not from a generic “OpenAI certified” badge.
OpenAI has not published the allocation of that $150 million between training, partner incentives, technical support, co-selling resources, or customer deployment funding. It also has not said how many firms will qualify for each tier or which benefits are exclusive. The headline investment is meaningful, but partners evaluating it should separate the announced total from the benefits available to their particular practice.
Codex’s user figure is a combined metric, not a 10 million-user coding product
Adae told CRN that Codex and ChatGPT Work together had reached 10 million weekly active users, with knowledge workers representing about 20 percent of Codex users. The latter claim aligns with OpenAI’s June report on Codex: OpenAI said Codex had passed 5 million weekly active users and that knowledge workers accounted for roughly one-fifth of the product’s users.The 10 million number should not be read as “Codex now has 10 million weekly active users.” It is a combined total for Codex and ChatGPT Work, the newer agent-oriented experience inside ChatGPT for multistep office tasks. OpenAI has not publicly provided a product-by-product split for that combined count in the material available so far.
That distinction changes how partners should plan. Codex may be expanding beyond developers into spreadsheet work, reporting, research, and lightweight internal tools, but its technical roots still shape the implementation challenge. A customer deploying Codex across Windows developer endpoints may need source-control policies, repository permissions, IDE configuration, terminal restrictions, secrets management, code-review gates, logging, and a clear process for human approval before changes reach production.
AWS says Codex can be configured to use Bedrock inference through the Codex app, CLI, and IDE integrations including Visual Studio Code, JetBrains, and Xcode. For Windows-focused IT teams, Visual Studio Code is the obvious on-ramp, but the important work is not installing an extension. It is deciding which repositories and environments can be exposed to an agent, how credentials are delegated, what data can leave a development environment, and how an organization audits the actions the agent takes.
AWS’s Bedrock pitch is tailored to those questions. Its OpenAI integration includes AWS Identity and Access Management, PrivateLink, encryption, CloudTrail logging, and Bedrock guardrails; the newer GPT-5.6 rollout also describes in-Region inference and data-perimeter policies. Those capabilities do not make an agent deployment safe by default. They give a partner the AWS control plane needed to create separate accounts, permissions, network paths, logging rules, and approval boundaries around it.
The channel work is moving from pilots to operating controls
Adae’s central claim—that larger customers are moving from single-product use cases toward agentic workflow automation—is directionally supported by the products AWS and OpenAI are shipping. Bedrock’s OpenAI offering now spans models, Codex, and managed agents, while the updated ChatGPT desktop experience brings Chat, Work, and Codex into a single product surface on Windows and Mac.This consolidation will make informal experimentation easier. It also increases the risk of organizations treating very different workloads as though they have the same controls. A ChatGPT Work task that summarizes files, a Codex task operating on a local repository, and a managed agent connected to business systems each have different identity, tool-access, retention, approval, and incident-response requirements.
The services opportunity therefore sits in the boundary-setting that vendors often compress into the word governance. A credible AWS/OpenAI partner will need to establish which agent can access which systems, where execution happens, which actions require approval, what gets logged, who can review those logs, how model changes are tested, and how costs are attributed to teams or applications. Those are recurring operational services, not one-off prompts or workshops.
OpenAI has supplied the partner program, AWS has supplied a procurement and control-plane path, and Adae has been hired to connect the two in EMEA. What remains unpublished is the piece partners normally need before betting heavily on a new vendor route: the precise economics. Until OpenAI and AWS disclose partner compensation, deal protection, regional coverage, and support commitments, the prudent play is to build Bedrock-based delivery capability around real customer workloads—and avoid assuming that certification alone comes with a revenue stream.
References
- Primary source: crn.com
Published: Wed, 05 Aug 2026 14:50:00 GMT
OpenAI’s New AWS Director On ‘Significant’ Codex, ChatGPT And Bedrock Opportunity For Partners
OpenAI’s AWS partnership and AI innovation via Amazon Bedrock, Codex, ChatGPT and the OpenAI Partner Network is a win for partners, OpenAI’s director of AWS partnerships, Jason Adae, says.www.crn.com - Related coverage: openai.com
Introducing the OpenAI Partner Network | OpenAI
OpenAI launches the Partner Network, investing $150M to help global partners accelerate enterprise AI adoption, deployment, and transformation.openai.com - Related coverage: aws.amazon.com
GPT-5.5, GPT-5.4, and Codex from OpenAI are now generally available on Amazon Bedrock - AWS
Discover more about what's new at AWS with GPT-5.5, GPT-5.4, and Codex from OpenAI are now generally available on Amazon Bedrockaws.amazon.com
- Related coverage: aws.amazon.com
Amazon Bedrock now offers OpenAI models, Codex, and Managed Agents (Limited Preview) - AWS
Discover more about what's new at AWS with Amazon Bedrock now offers OpenAI models, Codex, and Managed Agents (Limited Preview) - Related coverage: docs.aws.amazon.com
OpenAI - Amazon Bedrock
The following OpenAI models are available in Amazon Bedrock:docs.aws.amazon.com
- Related coverage: linkedin.com
I’m excited to share that I’ve joined OpenAI as Director, AWS Partnerships in EMEA. As I begin this next chapter, I’ve been reflecting on 8 incredible years at Amazon Web Services (AWS), a defining… | Jason Adae | 727 comments
I’m excited to share that I’ve joined OpenAI as Director, AWS Partnerships in EMEA. As I begin this next chapter, I’ve been reflecting on 8 incredible years at Amazon Web Services (AWS), a defining part of my career and an opportunity to work alongside exceptional customers, partners, and...www.linkedin.com
- Related coverage: edgen.dev.edgen.tech
- Related coverage: stephenvantran.com
OpenAI Puts Full-Duplex Voice Into Codex • The Weighted Average
OpenAI brought GPT-Live's full-duplex voice to Codex and ChatGPT Work as the pair crossed 10M weekly users, pushing agentic coding hands-free.
stephenvantran.com
- Related coverage: docs.aws.amazon.com
Supported foundation models in Amazon Bedrock - Amazon Bedrock
Information previously on this page, including model IDs, supported Regions, modalities, and inference parameters, has moved to models at a glance . Select a model on that page to view its details. - Related coverage: openai.com
Codex is becoming a productivity tool for everyone | OpenAI
The Next Era of Knowledge Work report explores how Codex is transforming productivity through AI-powered research, data analysis, workflow automation, and content creation.openai.com - Related coverage: community.openai.com
GPT-5.2 and GPT-5.3-Codex have been sunset in Codex with ChatGPT subscriptions - #52 by chriskuta - Codex - OpenAI Developer Community
Just wonder if this is expected? As I’ve been using 5.3 Codex even after 5.4 and 5.5 got released. I personally find it gives me the best outcome, for me even better experience than gpt 5.5. If expected, are we going to…
community.openai.com
- Related coverage: academy.openai.com
Workflow clip: Proactively monitor accounts with Codex - Video | OpenAI Academy
Yash Pahade (AI Success Engineer, OpenAI) demonstrates a Codex-powered account briefing workflow that brings daily account context, recent activity, next touches, approved sources, and metrics into one reviewable page for the team. The clip shows both the...
academy.openai.com
- Related coverage: cdn.openai.com
- Related coverage: help.openai.com
ChatGPT — Release Notes | OpenAI Help Center
A changelog of the latest updates and release notes for ChatGPT
help.openai.com
- Related coverage: cdn.openai.com