Amplitude has launched Wave, a proactive product agent that significantly accelerates the product improvement cycle, reducing it from approximately one month to just one to two days. This innovative tool leverages OpenAI’s advanced models, integrating analytics, session replays, customer feedback, and code to identify, prioritize, and draft solutions for product enhancements. The agent then hands off technical plans to coding agents, such as OpenAI’s Codex, for implementation.
Wave Agent Integrates Multiple Data Sources for Rapid Iteration
The Wave agent is designed to bridge the gap between understanding how customers use a product and implementing necessary changes. Teams provide Wave with a specific product area and target metrics, and the agent collates data from Amplitude’s analytics, session recordings, customer feedback, and source code. This comprehensive analysis allows Wave to surface and rank opportunities for improvement, draft potential fixes, and prepare the groundwork for coding agents. Post-implementation, Wave monitors the results and feeds the learnings back into the system for continuous iteration.
Specialized OpenAI Models Enhance Efficiency and Accuracy
Amplitude utilizes a trio of OpenAI models within Wave, each selected for specific roles to optimize performance and cost. GPT-6.1 Sol acts as the Opportunity Manager, handling deep investigations and demonstrating strong reasoning capabilities. GPT-6 Luna is employed for high-volume, low-cost tasks like scanning code and running routine checks, while GPT-Image-2.5 Flare assists with generating mockups. Internal evaluations have validated the effectiveness of these models, citing their performance in reasoning quality, cost-efficiency, and latency.
Humans Remain in the Loop for Strategic Decisions
Despite the advanced automation, Amplitude emphasizes that its design keeps humans involved in the decision-making process. Customers using Wave in beta can set their preferred levels of autonomy. While many bug fixes are routed directly to coding agents, proposals for new features and larger strategic initiatives still undergo human review. Early results from beta clients show significant progress, with one in four changes flagged by Wave being launched the same day, and two-thirds launched within a week.
Demonstrated Success in Real-World Applications
Amplitude has showcased Wave’s capabilities through internal experiments. On its own homepage, Wave personalized suggestion cards, leading to a 160% increase in card-driven sends. In one test, Wave identified that users were largely bypassing generic suggestion cards on Amplitude’s homepage. By analyzing session replays, it proposed replacing them with personalized suggestions based on recent user activity, such as charts or dashboards viewed. This led to a significant increase in engagement through these cards after just one day.
Transforming Software Development Workflows
The introduction of Wave represents a significant shift in how software companies can approach product development. By automating much of the discovery, planning, and initial implementation phases, Amplitude aims to enable self-improving products as a standard for software development. The company believes that as AI models become more adept at reasoning and coding, more complex tasks can be delegated, ultimately allowing better products to reach users faster. Amplitude’s vision with OpenAI is to create a future where software actively collaborates with human teams to enhance product quality and user experience.
The Chenab Times News Desk
