Transform

the Way

You Work

with Ruya

Watch Video

Composable AI Agents

Multi-Agent Orchestration

Compose, chain, and orchestrate specialized AI agents that work together to solve complex data workflows — autonomously.

Composable pipelinesBranching logicReal-time execution

Start

Pipeline trigger received — dispatching agents

pipeline_input

Describe your data analysis task...

01 · INTERPRET

Context & Intent Analysis

Before dispatching agents, the planner decomposes your request — understanding intent, recalling prior analyses, and building a structured task plan.

Reasoning / Planner Agent— Task Interpretation

task_plan.yaml

analyst: "Data Analyst DF"

dataset: "Telco_Churn.csv"

intent_breakdown:

retention: "D1 / D14 anchored vs install"

attribution: "CAC + LTV + ROAS by channel"

signals: "behavioral drivers of retention"

agents_required: "search, sql, datascience, viz"

sub-tasks

data sources

142ms

plan latency

02 · DISPATCH

Parallel Agent Dispatch

The pipeline splits into specialized agents running concurrently — each one tackling a different data domain for maximum throughput.

Search Agent

Marketing spend data (CRM)

Product analytics events

Revenue and subscription tables

SQL Agent

fact_customer_lifecycle2.3M rows

dim_acquisition_channel5 channels

fact_product_events18.7M events

03 · PROCESS

Parallel Data Processing

Each dispatched agent works simultaneously — parsing documents, querying the warehouse, and transforming raw records into clean cohort data.

Parse Agent

→Parsed Telco_Churn.csv

→Identified 12 feature columns

→Mapped acquisition_channel field

Summarize Agent

→Summarized 6-week install cohorts

→Anchor: install date vs activation date

→Flagged 3 high-churn segments

Query DB Agent

→Fetched 2.3M lifecycle records

→Applied tenure_months <= 6 filter

→Joined with product event stream

Transform Agent

→Computed D1/D7/D14/D30 retention

→Normalized LTV to CAC ratio

→Encoded channel as categorical

04 · CONVERGE

Data Science & Execution Plan

All processed streams converge into a data science layer — statistical modelling, cohort analysis, and an executable plan for the visualization agents.

Data Science Agent

Anchored cohort retention: D14 +7pp vs install anchor

Organic best channel: 30% D30 retention vs 12% Paid Social

Feature usage in week 1 predicts 2.8x retention

Exec Plan Agent

charts: "retention_curve, cac_ltv, roas, churn"

format: "interactive dashboard"

delivery: "datagol.ai/dashboard/churn-analysis"

field_lineage

SourcesWarehouseDerivedTransform

05 · GENERATE

Analysis & Visualization

Specialized agents render seven distinct analyses and visualizations — each calibrated to answer a specific dimension of your original query.

Retention CurveD1–D30 by acquisition channel

Google SearchInfluencerOrganicPaid SocialReferral

CAC vs Estimated LTVBy acquisition channel

Google SearchInfluencerOrganicPaid SocialReferral

ROAS by ChannelReturn on ad spend comparison

Churn Rate by Channeltenure ≤ 6 months, sorted desc

Monthly Churn TrendChurned customers by month & channel

Google SearchInfluencerOrganicPaid SocialReferral

Top Churn ReasonsEarly-tenure customers (≤ 6 months)

Churn by Plan TypeChurn rate % by plan & channel

Google SearchInfluencerOrganicPaid SocialReferral

06 · DELIVER

Dashboard Published

Results are packaged, formatted, and published to your DataGOL dashboard — ready to share with stakeholders, export, or trigger downstream automations.

7 charts generated across 5 acquisition channels

Retention anchored on activation improves D14 by +7pp

Organic shows 18.4x ROAS — highest across all channels

Paid Social churn at 42% — flag for growth review

Publishing to Dashboarddelivered

6/6

datagol.ai/dashboard/churn-analysis

12.5K

3.9K

31.2%

Organic

Ready for next analysis

The full pipeline ran in 142ms end-to-end. All agents have been returned to standby. Your dashboard is live at datagol.ai.