AI-assisted executive decision support system that consolidates meetings, performance data, and technical context into a weekly leadership workflow.
This system was developed for a Director of Product at a large technology startup, overseeing multiple product streams, teams, and areas of responsibility.
The client is already working with a range of AI-enabled tools, including meeting recordings, internal documents, analytics dashboards, and direct input from teams. Each week involves reviewing conversations, checking performance metrics for both product and company-wide revenue, preparing for leadership meetings, and following up on technical or operational queries.
This process is repeated on a weekly basis, requiring the client to revisit the same sources, interpret different formats, and assemble a clear view before being able to prepare for decisions and reporting.
Preparation for weekly leadership meetings requires reviewing several days of Slack messages, checking multiple dashboards, and following up on technical queries, often within a short time frame. This sits alongside the need to form a clear product view and make decisions based on what is changing across multiple streams.
The inputs come in different formats and levels of detail, from short Slack updates to detailed dashboards and technical discussions, each requiring manual interpretation. Understanding what changed often means connecting product behaviour to underlying technical work, which requires follow-ups with engineers. This process depends on one person reviewing and combining all of these inputs before a clear view can be formed.
The approach was to separate the different types of input based on how they are used in practice.
Operational conversations, performance data, and technical context were treated as distinct sources, each requiring a different way of being reviewed. Each input was handled independently first, allowing it to be processed in a way that matched how the client already worked with it.
The structure was organised around a fixed weekly review point, so that all relevant inputs are prepared ahead of leadership meetings rather than revisited continuously throughout the week.
Each part of the system serves a specific role:
The system is structured as three parallel workflows that run independently during the week and are reviewed together ahead of leadership meetings.
The first workflow is built in Notion, where meeting recordings and transcripts are stored in structured databases organised by product stream. Notion AI agents are used to process these records and generate weekly summaries for each stream.
The second workflow is an automated pipeline that processes data exported from internal business intelligence dashboards. Due to access restrictions, data is extracted manually as CSV files and uploaded into a controlled intake. A Make.com automation processes this data, compares weekly changes, and generates a structured summary for review.
The third workflow is a codebase query interface built using Cursor, allowing the client to retrieve technical context in plain language without relying on engineers for initial clarification.
These three workflows are reviewed together as part of the client’s preparation for weekly leadership discussions.
The current system supports a weekly preparation cycle. Further development is focused on tracking how decisions and issues evolve over time, including a record of recurring blockers and a view of dependencies between product streams. This would allow patterns to be reviewed across multiple weeks using the same structure.