SituationThe Data Fog and Friction Points
For a major OTT client, the success of a ‘Tentpole’ title hinges entirely on strategic intervention during its initial launch phase — specifically when anticipation and word-of-mouth peak. To capitalize on this critical momentum, marketing leaders must execute rapid, tactical pivots — such as reallocating spend by geo, gender, or cohort — based on immediate, precise feedback from awareness and demand metrics (e.g., Unaided Top of Mind, Customer Intent).
However, the foundational data needed for these rapid tactical pivots was difficult to access and use effectively in time. The client’s process tracked approximately seven key metrics, which were siloed across two different external agencies and managed via a series of disconnected Excel and Google Sheets.
This fragmented environment created a difficult reality for marketing managers:
- Dependency Latency: Managers spent significant time chasing multiple teams and data sheets for updates.
- Historical Blind Spots: Digging up older files to understand comparative benchmarks of past titles added significant mess and hassle, leading to strategic delays.
The compounded effort, which required approximately two hours per week of manual aggregation, resulted in delayed decision-making — causing decisions in marketing pivots to miss crucial, high-impact windows.
TaskAutomate Alignment and Enable Real-Time Action
The mandate for Madspek was to eliminate the manual aggregation effort, provide a single source of truth, and — crucially — integrate the solution directly into the marketing manager’s daily workflow, enabling instant data-driven decisions.
The goal wasn’t just to build a dashboard. It was to put intelligence directly into the hands of decision-makers, exactly where they already work — Google Chat.
ActionMulti-Agent Orchestration on Vertex AI
Madspek began by collaborating closely with the marketing managers to understand the what, how, and why of their exercise. This foundational work informed the creation of the Streaming Sentinel — an autonomous, multi-agent ecosystem built entirely on Google Cloud’s Vertex AI and deployed through a specialized bot in Google Chat.
| Component | Operational Role | Google AI/Cloud Equivalent |
|---|---|---|
| Ingestion & Mapping Agents | Performs fuzzy and semantic matching (Name Mapper logic) across the various Google Sheets and Excel files, collating and refining the data. | Vertex AI Pipelines orchestrating Gemini 1.5 Pro Function Calling |
| Temporal Logic Engine | Programmatically translates raw dates into the client’s internal Phase-Relative format for direct launch momentum comparison. | Cloud Functions / Cloud Run |
| Data Store & Refresh | Manages automated data refresh and serves as the Single Source of Truth for the Chatbot’s grounding. | BigQuery & Google Cloud Storage (GCS) |
| Human-in-the-Loop (HITL) | A verification layer used to audit semantic matching and train the end Chatbot on desired outputs (e.g., tabular format, concise analysis). | Vertex AI Grounding & Custom UIs |
Core Infrastructure: Vertex AI and Gemini 1.5 Pro
The core system is powered by Gemini 1.5 Pro, selected for its large context window and powerful reasoning capabilities, serving as the Chatbot’s Grounding Engine. The infrastructure is a multi-agent flow that automates complex data joining and temporal alignment.
The Action Layer Integration
The Streaming Sentinel is designed to take action like sending emails directly from the chat interface, a feature enabled by Gemini’s Function Calling on Vertex AI. For example, a manager interacts with Google Chat, reviews the data generated by the Streaming Sentinel, and then asks in the same chat window:
- "This looks great, email it to the Social Team."
- The Gemini model calls a pre-defined function that interfaces with the Gmail API to draft and send the stakeholder email instantly, removing friction from the reporting process.
Token Optimization and Economics
Madspek optimized the agentic flow to maximize efficiency and minimize the cost of generating data for user queries, leveraging the cost-effectiveness of Gemini 1.5 Pro.
- Initial Data Ingestion: Establishing the agentic ecosystem required processing and semantically mapping more than 100M tokens across the client’s fragmented Excel files, Google Sheets, and operational datasets.
- Cost Efficiency: Madspek implemented a Retrieval-Augmented Generation (RAG) architecture that transformed the disconnected datasets into a structured Single Source of Truth. Instead of repeatedly processing entire datasets during live interactions, the system retrieves only small, highly relevant data snippets for each user inquiry.
- This significantly reduced token consumption and lowered the cost of typical production queries to fractions of a cent to a few cents per interaction, depending on query complexity and output size.
ResultsFaster Decisions During Critical Launch Windows
The implementation transformed a fragmented reporting process into a centralized launch intelligence system, enabling marketing teams to respond faster and more confidently during the most commercially important phases of a title launch.
Significantly Reduced Reporting Time
Manual aggregation and QA workflows that previously required hours of coordination across agencies, Excel files, and Google Sheets were reduced to minutes, dramatically lowering the operational overhead involved in preparing and validating launch data.
Faster, More Confident Marketing Decisions
Marketing managers gained immediate visibility into how different audiences and geographies were responding during critical launch windows, making it easier to spot where awareness and demand were building strongest. This allowed teams to quickly shift budgets, messaging, and focus toward high-momentum segments while there was still time to influence launch outcomes.
Contextual Performance Visibility
Teams could also compare current title performance against historical launches, giving managers additional context to calibrate expectations and guide decision-making throughout the campaign lifecycle.
Ready to Transform Your Data Strategy?
Contact Madspek Consulting today to discuss how our multi-agent solutions can automate your complex data alignment and operational workflows.
