Tech • AI • Robotics • Game
How OpenAI Puts ChatGPT to Work | DevDay 2026
9.4/10TL;DR
OpenAI showcased how employees are using ChatGPT, new collaborative Spaces, and autonomous Dots to hand off repetitive work, speed research, and make internal expertise available across teams.
KEY POINTS
A push to reduce workplace overload
The core idea is to move workers from reactive inbox-and-chat triage toward higher-value tasks. Instead of spending mornings checking email, Slack, dashboards, and scattered updates, teams are increasingly using ChatGPT work as an “agentic teammate” that can gather context, prepare answers, and maintain ongoing tasks.
Three internal use cases revealed a common pattern
The examples came from three very different functions at OpenAI: developer social, competitive research, and people operations. Despite the differences, each role used AI in the same way: define a repeatable responsibility, provide the right context and sources, and let the system handle monitoring, collection, and first-pass analysis so humans can focus on judgment.
Developer social monitoring was turned into a live reporting system
Dani, who leads social for OpenAI developers, used ChatGPT to combine launch-performance data and developer feedback into a live internal reporting site. A Dot continuously reads social feeds, adds relevant posts, and logs engagement metrics, giving teams an updated view of what content is working and how a product launch is being received.
Slack oversight was delegated to AI
The same setup was extended to internal communications. Because launch teams often need to follow many Slack channels at once, a Dot was assigned to monitor those channels, flag urgent developments, and alert Dani when direct input was needed. That shifted hours of manual scrolling into a background process and freed more time for actual community engagement.
A two-person research team expanded its reach
Adrian, who leads competitive research with a team of just two, turned his research method into a plugin that investigates competitors, finds patterns, and tests conclusions. That allowed the team to answer more requests from sales, product, go-to-market, and strategy groups without having to choose as often which questions would go unanswered.
AI-assisted investigations challenged assumptions with evidence
In one case, Adrian examined why a rival’s revenue was rising quickly by having the system search calls, newsletters, and podcasts. The evidence did not support the initial theory that the company was winning a new set of very large customers. In another case, the system reviewed 90 days of sales-call transcripts and uncovered reasons for lost deals that had not been captured in CRM summaries.
Competitive intelligence became self-serve
Adrian also built Compete Corner, an internal site organized around the questions colleagues ask most often. A Dot pulls competitive questions from Slack, researches them with Adrian’s method, drafts findings for review, and updates the site after approval. That turned one expert team into a scalable internal resource, while Adrian also developed Model Galaxy, an interactive tool comparing model performance, cost, and latency across benchmarks.
People operations used a shared AI knowledge base for coaching
Casey, OpenAI’s chief people officer, focused on how AI is changing team structures and work design. His team gathered meeting notes, papers, and articles into a shared knowledge base, then used that material to power a leadership coaching site with GPT Live. Leaders can use it to discuss where work is getting stuck and what tasks could be handed over to AI.
New products center on shared context and persistent agents
Spaces were introduced as collaborative work areas where pages, files, and research can be developed jointly by teammates and their AI assistants. Users can work on the same page at the same time, while individual chats remain private. Dots were presented as persistent agents with a cloud computer and browser, reachable through Slack, phone calls, or messages, able to continue tasks, send updates, and retain context over time.
CONCLUSION
The broader bet is that AI will not just answer questions but take ownership of recurring work. If that model holds, the main productivity gain may come less from faster drafting and more from freeing people to spend attention on judgment, strategy, and relationships.
Ask a question