ChatGPT Work is gradually leaving the chat window
With GPT-6 Astra, OpenAI is no longer simply trying to improve the quality of ChatGPT’s answers. The company wants to bring its AI closer to the applications in which employees actually work. With the arrival of GPT-6 Astra on Amazon Bedrock, new enterprise plugins for ChatGPT Work extend its browser-use capabilities to business intelligence tools as well as to Workday, Navan and Avalara, with use cases focused on data analysis, operations and finance.¹ This is a significant change: ChatGPT Work can already gather information from applications and files, use the web and turn complex work into deliverables such as documents, spreadsheets, presentations or websites. With these additional connections, the AI is now moving directly closer to the software that structures a company’s day-to-day processes.
HR, travel, tax: AI moves closer to business processes
The choice of the first applications illustrates this shift well. Workday plays a central role in the HR and finance systems of many organisations, Navan handles business travel and expense management, while Avalara automates processes related to tax and compliance in particular. AWS has not yet provided an exhaustive list of actions specific to each of these plugins, so it would be premature to claim that ChatGPT Work can carry out any operation on these platforms. The announcement nonetheless confirms that their environments can now be used in workflows driven by Astra, alongside business intelligence tools.¹ The line between conversational assistant and business software is therefore becoming thinner: users can hand an objective to the AI, let it look for the information it needs in authorised environments and supervise the progress of the work rather than manually moving data from one application to another.
The browser becomes a gateway between AI and applications
This integration relies in particular on the advanced computer use and browser use capabilities of GPT-6 Astra. OpenAI presents its new model as particularly strong at complex, multi-step professional tasks, able to work across software interfaces and produce structured deliverables directly.² This approach is strategic because not all business software necessarily offers the APIs, connectors or automations that would let an agent carry out every step of a process. An AI capable of using a web interface can potentially pursue certain tasks by interacting with software in a way closer to that of a human user, which considerably widens the number of workflows that can be assisted or automated.
Automation does not mean handing AI the keys to the company
Connecting an agent to HR, financial or operational applications immediately raises the question of access rights. On this point, AWS specifies that the new plugins work through the user’s existing account and the permissions set by administrators, so they do not automatically give GPT-6 Astra additional privileges.¹ ChatGPT Work also makes it possible to control which applications and sites the agent can access, manage file transfers and require human confirmation before certain actions. Users can follow the progress of the work, change their instructions and approve important steps. This human-in-the-loop approach becomes essential as agents move beyond purely informational tasks to take part in processes that may affect the organisation’s data, spending or operations.
AI becomes a new layer of business software
This transformation brings Software and Application Management back to the fore. For a long time, administering business tools mainly meant deploying solutions, managing accounts, configuring access, keeping software up to date and supporting users. The arrival of agents able to move between several applications now adds an extra layer: deciding which data can be consulted, which actions can be carried out, which plugins should be authorised and when human validation remains mandatory. IT teams will also need to understand how these agents interact with ERPs, HR tools, financial solutions and collaboration platforms, so that a productivity gain does not turn into a proliferation of hard-to-control access. This development is therefore gradually bringing application administration closer to AI agent governance.
From copilot to workflow, enterprise software changes its logic
Above all, ChatGPT Work’s plugins illustrate a change of model. The first generation of generative AI mainly helped people write, summarise or search for information; the new generation seeks to complete an entire sequence of work by drawing on several sources and applications. GPT-6 Astra was specifically trained for multi-step professional tasks, and OpenAI says it can produce documents, presentations, spreadsheets and analyses tailored to an organisation’s templates and standards.² In this context, business software is no longer necessarily the place where users carry out each operation manually: it can become one of the resources an agent draws on to achieve a broader goal. The value of AI is thus gradually shifting from the answer it produces to the process it is able to orchestrate.
Companies will have to learn to govern digital co-workers
The arrival of Workday, Navan, Avalara and BI tools in the ChatGPT Work ecosystem ultimately offers a glimpse of what the agentic workstation could become. Instead of opening several pieces of software in turn, extracting information and then assembling it manually, employees will be able to entrust more cross-functional tasks to an AI capable of working across these environments. But the more agents can act, the more strategic their supervision becomes: permissions, identity, traceability, validation of operations and accountability in the event of error will need to be considered as carefully as their performance. The real change, then, is not simply that ChatGPT can connect to more applications; it is that AI is beginning to become an orchestration layer between the software that keeps the business running.
360° IT News — decoding technology to understand the jobs of tomorrow.
Further reading
This shift had already begun in collaboration tools. Read our CCCLX article “ Claude Tag turns Slack into a genuine agentic workspace ”, which looks at how Anthropic is building an agent directly into Slack to track projects, find information and support teams without leaving their working environment.
References
1. Amazon Web Services. (2026). Take on your most ambitious work with GPT-6 Astra on Amazon Bedrock.
https://aws.amazon.com/blogs/machine-learning/take-on-your-most-ambitious-work-with-gpt-6-astra-on-amazon-bedrock/
2. OpenAI. (2026). GPT-6 Astra: A new generation of intelligence.
https://openai.com/index/gpt-6-astra/
3. OpenAI. (2026). ChatGPT Release Notes.
https://help.openai.com/en/articles/6825453-chatgpt-release-notes
ServiceNow wants to turn process improvement into a continuous AI-driven loop
With AI Workflow Factory, ServiceNow aims to link the identification of inefficient processes with building and running their improvements. The announcement notably includes Autonomous Engineer, a tool designed to automate more of the implementation work.[1] For...
Mistral Large 4: what its trillion-parameter model could change for businesses
Mistral has opened the public preview of Mistral Large 4, nicknamed “Le Chonk”, and plans to release its weights by the end of the month.[1] This European model targets agentic use cases in particular, as well as cybersecurity, finance and industry. For businesses,...
IBM Bob: developing with AI without your code leaving the company
IIBM has announced a self-hosted version of Bob, its agentic development platform, for internal infrastructure, private or sovereign clouds and environments isolated from the internet.[1] This option addresses a concrete obstacle: some applications contain code, data...
HPE is preparing a network that anticipates incidents and automates certain interventions
HPE plans to deploy a Self-Driving Network for the 2027 Ryder Cup, with Wi-Fi 7 and an operations centre bringing together data and AI tools.[1] The stated aim is to monitor usage, adapt the network and move towards autonomous interventions. For businesses, the...
Can your AI agent really defend your company? Hack The Box puts its skills to the test
Hack The Box is launching AI Range Enterprise Edition to let security teams assess their own AI agents on the roles assigned to them.[1] For businesses, the stakes are concrete: having evidence before delegating a task, then checking that this evidence remains valid...
VAST Data wants AI to work with sensitive data in a protected environment
VAST Data has unveiled DataEnclave, a confidential execution capability designed to bring artificial intelligence models closer to companies’ sensitive data. Presented as a preview, it is due to go on sale in the first quarter of 2027.[1] The aim is to make certain...
Databricks buys Row Zero to bring spreadsheets closer to governed data
Databricks has announced the acquisition of Row Zero and plans to integrate its spreadsheet experience into Genie.[1] The aim is to let business teams explore data in a familiar interface, with greater control over access and usage. For companies, the challenge is to...
Google Cloud takes its X5 instances to 48 TB for large SAP databases
Google Cloud has announced the general availability of the X5 series, which raises the maximum advertised memory capacity on a single node to 48 TB, compared with 32 TB for X4.[1] Aimed in particular at very large SAP HANA databases, the offering is based on bare...
Cloud Storage: Google automates the detection of anomalies
A rise in reads, cross-region transfers or stored volume can inflate the cloud bill and disrupt applications. With Storage Intelligence advisor, Google offers a view of unusual changes in Cloud Storage, along with leads for investigating them. For businesses, the...
AI voice agents: Oracle details a LiveKit architecture on OCI
Oracle has described an architecture that pairs the LiveKit Agents framework with Oracle Cloud Infrastructure to automate real-time voice calls. The design links telephony, AI models and business services, while placing internal components in private subnets. For...
