Data Infrastructure · Agent-Ready · MCP + CLI

Analytics infrastructure built for humans.Now optimized for AI agents.

AdaSight is the semantic data layer between enterprises and their AI agents — exposing analytics as structured tool calls, typed APIs, and citation-ready context.

$ adasight query --metric=activation --cohort=last_30d

→ MCP tool: analytics.activation_funnel

steps: [signup → first_event → aha_moment]

conversion: 41.2% (↑ 6.1pp WoW)

citation: events.amplitude #2024-Q4 · model: growth_states.v3

The problem

Data optimized for dashboards is wrong for AI agents.

For human analysts

  • 01

    60–80% of time spent cleaning data, not deriving insights

  • 02

    Raw data is messy, inconsistent, requires significant transformation

  • 03

    Existing tools produce visually digestible but semantically shallow outputs

For AI agents

  • 01

    LLMs cannot effectively consume dashboards, charts, or PDF reports

  • 02

    Outputs optimized for visual cognition, not programmatic reasoning

  • 03

    Agents need structured, context-rich, queryable data

The insight

The semantic layer we built for humans is exactly what AI agents need to reason over enterprise data.

In 2024–2025 we watched companies try to connect GPT-4, Claude, and internal agents to their analytics stacks. The results were poor — not because the models were weak, but because the data wasn't prepared for agent consumption.

Schema AmbiguityMissing ProvenanceFormat MismatchNo Action Surface

Phase II · Agent-ready infrastructure

Building the data layer for the agent era.

MCP Integration

Native Model Context Protocol server exposing analytics as structured tool calls.

  • Query, filter, aggregate without custom work
  • Built-in semantic descriptions
  • Agents understand what data represents

CLI for Agents

Headless interface for agent orchestration systems.

  • Composable commands for pipelines
  • Machine-readable JSON outputs
  • Optional human-readable summaries

Optimized Formats

Data structures designed for agent reasoning.

  • Contextual embeddings with business meaning
  • Typed APIs preventing hallucination
  • Citation-ready lineage metadata

Technical differentiation

Built for the agent era, not retrofitted.

CapabilityTraditional BIAdaSight Phase II
Primary consumerHuman analystsAI agents + humans
Output formatCharts, dashboardsStructured JSON, MCP tools, typed APIs
Context encodingVisual cues, legendsSemantic metadata, ontologies
Action surfaceView-onlyQuery, transform, validate, act
Integration modelEmbed / iframeProtocol-native (MCP, OpenAPI, CLI)

Traction

100+ engagements. Real customers. Real decisions unlocked.

100+

engagements shipped

$1M

in sales to date

10×

faster decisions reported

Tonic

Data Stack Audit

Audited Segment / Amplitude / Braze. Defined growth KPIs. Applied the Duolingo Growth States Model to surface engagement patterns.

"I love how you're correlating data, deriving potential causes, and describing the charts. Immensely helpful."

Analía Ibargoyen · Head of Product, Tonic

Seatfrog

Analytics Foundation

Reactivated dormant Product Analytics. Revalidated events. Turned a quiet tool into a live source of decisions.

"The team will be able to consume the data much more easily and make decisions 10× faster."

John Ritchie · Head of Product, Seatfrog

Past clients include · Personio · Candis · Notarize · Newscorp

Market opportunity

Defining a new layer in the enterprise AI stack.

Phase I TAM

$30B

Business intelligence and analytics market

Phase II TAM

$15–50B

AI agent infrastructure market by 2028

  • → Not fighting for share in a mature BI market
  • → Defining a new layer between data warehouses and agents
  • → Every company deploying agents will need this infrastructure

Competitive advantage

Our moat.

Proprietary Semantic Layer

Trained on patterns from 100+ growth-team deployments — Tonic, Seatfrog, Candis and many others.

First-Mover Advantage

Building agent-native analytics infrastructure before incumbents pivot.

Dual-Use Architecture

Same semantic layer serves both humans and agents — no separate systems.

Why not existing BI tools?

Tableau, Looker, Power BI are optimized for human visual cognition. Retrofitting them for agents requires fundamental architecture changes they're unlikely to make.

Why not raw data platforms?

Snowflake and Databricks provide storage and compute, not semantic preparation. Agents talking directly to data warehouses produce unreliable results.

Team

Operators who've built this before.

DM

Dayana Marin

Co-founder & CEO

Operator, ex-tech partnerships at Booking.com. Owns GTM. Bootstrapped and shipping.

GS

Gregor Spielmann

Co-founder & CTO

Former Solution Data Engineer at Amplitude and Optimizely. Consulting clients including Adidas.

15 additional team members from Booking.com · Make.com · Gojek · Sony · Optimizely · Canva · Adidas

Advised by Simon Jackson, PhD (ex-Booking.com, Meta, Canva) and Bruno Pais (Sr. Director AI Products, Sennder)

The ask

Building the foundation for human-agent collaboration.

The company that owns the data preparation layer between enterprises and their agents will be foundational infrastructure.

Goal 01

Ship MCP + CLI to production

Goal 02

Expand engineering team

Goal 03

10 design partners

Goal 04

Build the category

Become a design partner