Zendesk Staff Built About 600 Apps on ChatGPT Enterprise on Their Own
Zendesk rolled out ChatGPT Enterprise to all employees, and sales, marketing and operations built about 600 apps for their own tasks. One example is a sales AI agent that combines product usage data with the number of tickets per agent and suggests where a customer could buy more. Around 1,000 people use it every day. The takeaway: when business teams get access to data themselves, they find use cases that IT might not have considered.
AI-processed from YouTube: OpenAI; edited by Hamidun News
Enterprise AI usually works like this: the IT department picks one or two use cases, access requests take weeks to approve, and business users wait in line for analysts for years. Zendesk, a developer of customer support software, took a different approach and opened ChatGPT Enterprise to every employee. The results were presented by Daniel Hamble, the company's Vice President of Enterprise Data and Analytics, in a video published by OpenAI. According to him, the company already has about 600 apps "built by different people."
Six Hundred Apps Without a Single Client
Sales, marketing and operations are building tools for their own daily work. This is not a showcase project from one development team, but a set of working apps that came from the users themselves.
Business employees ask questions of company data directly. Previously, such a question turned into a request: an analyst wrote a query, built a report and sent a spreadsheet, and by then the question had often changed. Now the person who knows the task asks directly and gets an answer right away.
This approach has a cost. Six hundred apps from different authors mean varying quality, different data sources and possible duplicates. The video does not address this, but anyone who has rolled out internal tools knows that sooner or later they will need an app catalog, access controls and rules about which data can be connected.
How the Sales Agent Works
The most concrete example in the video is the AI agent in the sales department. It connects two datasets: how a customer uses the product and how many tickets each agent handles. By comparing them, the system shows where there are opportunities to expand the partnership. How exactly the agent evaluates these opportunities is not explained in the description.
The agent has a second job as well: it helps the salesperson explain to the customer what benefit they will get. For a manager, that is the difference between a dry "your workload has grown" and a ready argument for the conversation.
According to the video, about 1,000 people use this app every day. For an internal tool, that is a notable figure: it indicates the app is part of salespeople's working rhythm, not a demo sitting for management.
What to Take Away
First: enterprise AI is more useful when it is connected to real company data. Zendesk's example is built on a combination of product usage data and support data. Without access to them, the agent would be just a generic chatbot with general answers.
Second: delegation works. When people have access and clear tools, they find use cases that a central team might never reach. The role of IT and the data team shifts toward a different task: providing the platform, the data and the guardrails.
Third: it is worth tracking how many people return to a specific app each day. The figure of 1,000 daily users is the main evidence in this story that the tool has taken hold. Product managers should adopt this metric.
Limitations
The material is a video by OpenAI itself about its customer, so it is a showcase, not independent research. The description contains no data on how much additional revenue the sales agent generated, what it costs to support hundreds of apps, or how answer quality is verified. The figures "about 600" and "about 1,000" are approximate, with no counting methodology.
The scenario itself is clear and reproducible: take two data sources that previously lived apart, give them to an agent, and put the result directly into an employee's workflow. If you are rolling out AI in your company, start by asking which data questions your business teams are currently waiting weeks to get answered, and open access to that data first. In parallel, plan the app catalog and access rules in advance so that in six months you are not sorting through a hundred scattered homemade tools.
- Nov 21, 2026FDE job postings grew 4.6x in a year: what a forward deployed engineer actually does at the client's site
- Nov 21, 2026A swimming team captain handed ChatGPT three scheduled tasks and cut her routine work by a multiple
- Nov 21, 2026OpenAI GPT-6 Cyber Found About 35 Flaws in iPhone, Mac and Other Apple Devices
- Nov 21, 202622 Claude Agents Spent Six Weeks on Black Hole Problems — and Solved Two
Key Questions Answered
What are the key details on Six Hundred Apps Without a Single Client?
Sales, marketing and operations are building tools for their own daily work. This is not a showcase project from one development team, but a set of working apps that came from the users themselves. Business employees ask questions of company data directly. Previously, such a question turned into a request: an analyst wrote a query, built a report and sent a spreadsheet, and by then the question had often changed. Now the person who knows the task asks directly and gets an answer right away.
How the Sales Agent Works?
The most concrete example in the video is the AI agent in the sales department. It connects two datasets: how a customer uses the product and how many tickets each agent handles. By comparing them, the system shows where there are opportunities to expand the partnership. How exactly the agent evaluates these opportunities is not explained in the description. The agent has a second job as well: it helps the salesperson explain to the customer what benefit they will get.
What to Take Away?
First: enterprise AI is more useful when it is connected to real company data. Zendesk's example is built on a combination of product usage data and support data. Without access to them, the agent would be just a generic chatbot with general answers. Second: delegation works. When people have access and clear tools, they find use cases that a central team might never reach. The role of IT and the data team shifts toward a different task: providing the platform, the data and the guardrails.
Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews
Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.
