Matthew van der Ven, a representative from OpenAI, showcases the capabilities of Spark, a feature within ChatGPT that allows users to build custom AI agents. This demonstration focuses on creating a sales agent designed to streamline lead research and outreach processes, highlighting how AI can be utilized to enhance productivity for sales teams.
Building a Sales Agent with Spark
The core of the demonstration revolves around Matthew's prompt to build a sales agent. He outlines the desired functions: to research each new lead, grade its quality, draft initial outreach emails, draft follow-up emails, and schedule calendar reminders for those follow-ups. This process emphasizes the natural language interface of Spark, where complex workflows can be described conversationally.
The full discussion can be found on OpenAI Youtube's YouTube channel.
Spark then translates this prompt into an actionable agent plan. The plan includes capabilities like researching new leads, scoring lead quality, drafting outreach emails, and scheduling follow-up reminders. It also specifies the tools the agent will use: Web Search, Gmail, and Google Calendar. The primary channel for interaction is noted as responding to messages in ChatGPT, with options to add other channels like Slack.
Automating the Sales Workflow
The video walks through the creation of the agent, showing how Spark integrates the selected tools and configures the necessary steps. Matthew explains that the agent will research leads using web searches, then use Gmail to draft the initial outreach and follow-up emails. Google Calendar is integrated to schedule reminders for subsequent actions.
After the agent plan is outlined, Matthew shows how Spark then begins building the agent in real-time. The process involves checking current OpenAI product guidance, verifying details from official sources, and drafting a concise, buyer-ready response. The agent's workflow is visualized, showing it pulling lead details, evaluating them, and initiating the outreach process.
Real-World Application: Lead Contact
To illustrate the agent's effectiveness, the video simulates a scenario where a new lead, Marc Reed from The AGI Company, contacts sales. Marc's inquiry is about the difference between ChatGPT Business and ChatGPT Enterprise and how his sales team can use the product. This is a typical sales qualification scenario.
Instead of a human salesperson immediately responding, Spark handles the submission in the background. The agent accesses the lead's information from the contact form. It then pulls up relevant OpenAI product guidance to formulate an accurate and tailored answer. The video shows Spark searching for information on OpenAI's website to compare ChatGPT Business and Enterprise.
Following the research, Spark drafts a personalized email to Marc Reed, explaining the differences and suitability of each product for different sales team needs. It also drafts a follow-up email to ensure Marc saw the initial message and suggests grabbing time on the calendar to discuss sales use cases. The entire process, from lead submission to drafting follow-up communications, occurs automatically.
Broader Applications of Workspace Agents
The demonstration concludes by highlighting that Spark is not limited to just sales lead engagement. The platform enables the creation of workspace agents for a variety of tasks. These include building account plans, managing RFP responses, and ultimately driving better customer experiences. The availability of these agents across ChatGPT Business, Enterprise, and Edu signifies OpenAI's push to integrate AI deeply into professional workflows.
This feature represents a significant step towards intelligent automation within business operations. By allowing users to define and build custom AI agents using natural language, Spark democratizes the creation of powerful AI tools, enabling teams to focus on strategic work rather than repetitive administrative tasks.
