QueryStory
by QueryStory
Agentic data platform that turns plain-English questions into auditable, decision-ready business narratives
QueryStory is an agentic data platform for enterprises that need answers they can defend, not just answers they can read. It connects to existing warehouses, CRMs and unstructured sources under the permissions already in place, answers questions in plain English, and assembles a series of investigations into a single business narrative published as a deck, doc or dashboard that keeps itself up to date — with the SQL, sources and assumptions shown.
QueryStory came out of stealth in San Francisco on 26 August 2026, founded by Shapor Naghibzadeh, who co-founded Chronicle inside Google's X Labs and led AI for Google SecOps, with CTO Stanley Yang (a former Google colleague and lead engineer at EvolutionIQ) and CPO David Glusic (an Accenture veteran). The product applies techniques the team built for security data — extracting trustworthy answers from messy, multi-vendor telemetry — to general business analytics. Its thesis is that speed without traceability is just a faster way to be wrong, so the platform is organised around the audit trail rather than the chat window. Users ask questions in plain English against 15+ connected sources spanning data warehouses (Snowflake, Databricks, BigQuery), CRMs, and unstructured material such as call recordings, presentations and old spreadsheets, under the permissions and governance already in force. The SQL the agent writes surfaces automatically rather than staying hidden, analyses can be flagged for a human colleague to review and those reviews are recorded in the platform, and every output shows its sources, definitions and assumptions. Individual investigations are then assembled into what the company calls a unified business narrative and published as living outputs — decks, documents, dashboards or messages that refresh as the underlying data changes — with decision rationale captured as institutional knowledge. It is model-agnostic in design while currently relying mainly on frontier-lab models, and deliberately owns none of the storage or compute: data stays in the customer's own cloud and region, and the AI management system is independently audited against SOC 2, ISO 27001, ISO 42001 and GDPR. The company raised a $6 million seed round in late 2025 from Brightmind Partners and New York Life Ventures at a reported $60 million valuation, and says it is running with Fortune 500 customers in regulated sectors, though none are named publicly and TechCrunch describes the deployments as pilots.
The head of data or analytics at a regulated enterprise whose team is the bottleneck — business users wait days on ticketed analysis requests — and who cannot hand those users a general-purpose LLM because the answers would be unverifiable.
Business users get answers in minutes with the SQL, sources and assumptions visible and a human review recorded, so the analytics team reviews decisions instead of writing every query.
At a Glance
- Category
- Data & Analytics
- Pricing
- Contact for pricing
- Target Market
- CDOs, Heads of Data and Analytics, CFOs, RevOps Leaders, Data Scientists, Business Analysts
- Deployment
- Cloud-first, Hybrid
- Founded
- 2025
- Headquarters
- San Francisco, United States
- Team Size
- 11-50
Key Features
- ✓Unified business narrative
Assembles a series of individual questions into one coherent investigation rather than leaving users with disconnected one-off answers.
- ✓Visible SQL and assumptions
Surfaces the generated SQL, sources, definitions and assumptions on every answer so the reasoning can be checked rather than trusted blindly.
- ✓Human review workflow
Analyses can be flagged for a colleague to review, and those reviews are recorded in the platform as a durable decision audit trail.
- ✓Living outputs
Publishes results as decks, documents, dashboards or messages that automatically update as the underlying data changes.
- ✓Cross-source querying
Connects 15+ sources including Snowflake, Databricks and BigQuery plus CRMs, call recordings, presentations and spreadsheets in one investigation.
- ✓Permission-preserving access
Respects the access controls and governance already in place rather than requiring a parallel permission model to be built.
- ✓Data stays in your cloud
Runs against data in the customer's own cloud infrastructure and region, with no customer data used for model training.
- ✓Project-based context
Organises work around team goals with shared metric definitions and governance so different teams stop producing different versions of truth.
Capabilities
Use Cases
- •Quarterly business review preparation
One investor replaced work previously done by several people preparing QBRs, according to TechCrunch's reporting on early deployments.
- •Finance variance analysis
Produces board-ready metrics with an audit trail showing the definitions and assumptions behind each figure that was reported.
- •RevOps pipeline and forecast review
Sales and revenue operations teams interrogate pipeline health and forecast movement without waiting on an analyst ticket queue.
- •Metric governance for data teams
Shared definitions and recorded human reviews stop different teams arriving at different numbers for the same business question.
- •Campaign attribution and customer journey analysis
Marketing traces journeys and attributes campaigns across warehouse data and unstructured sources in a single governed investigation.
Ideal For
Best For
- ✓Regulated enterprises that need every AI-produced number traceable to its source, definition and assumption before anyone acts on it
- ✓Unblocking a data team that has become a ticket queue for routine business questions from RevOps, finance and sales
- ✓Recurring decision artefacts — quarterly business reviews, board metrics, variance analysis — that must refresh automatically as data changes
- ✓Querying across structured warehouse data and unstructured material (call recordings, decks, spreadsheets) in one investigation
- ✓Organisations that need answers without moving data: it stays in the customer's own cloud and region
- ✓Capturing why a decision was made, not just what the number was, as reusable institutional knowledge
Not Ideal For
- ✗Risk-averse buyers who need a proven vendor: the company emerged from stealth on 26 August 2026 with a $6M seed, no named customers, and deployments TechCrunch describes as pilots
- ✗Teams wanting a self-serve tool — there is no published pricing, no free tier and no trial; the only path in is a demo request
- ✗Organisations without a governed semantic layer or clean warehouse, since the platform reasons over existing data and does not provide storage, compute or data cleanup
- ✗Buyers standardising on their warehouse vendor's native agent (Databricks Genie, Snowflake Cortex Analyst), where an additional vendor is hard to justify
Deployment
Market Analysis
Pros
- ✓Traceability is architectural rather than bolted on: the generated SQL, sources, definitions and assumptions are all surfaced and human reviews are recorded
- ✓Unusually strong compliance posture for a seed-stage company — SOC 2, ISO 27001, GDPR and an ISO 42001-audited AI management system
- ✓Data never leaves the customer's cloud or region, and is not used for model training, which removes the usual blocker for regulated buyers
- ✓Founding team has directly relevant pedigree: Chronicle (Google X Labs, acquired by Google Cloud), Google SecOps AI, EvolutionIQ and Accenture
- ✓Connects unstructured sources such as call recordings and old decks alongside the warehouse, which warehouse-native agents generally do not
Cons
- ✗Three days old as a public company at time of writing — no named customers, no case studies, no reference architecture, and TechCrunch characterises current deployments as pilots
- ✗Zero independent validation: no listing on G2, Capterra or TrustRadius, no Product Hunt launch, and a Hacker News search returns nothing about the company at all
- ✗$6M seed and a roughly 12-person team is thin cover for an enterprise deployment in a regulated sector — vendor-continuity risk is real and should be priced in
- ✗Model-agnostic in design but currently dependent on frontier-lab models, so answer quality, latency and cost track a third party the customer does not control
- ✗No published pricing, no trial and no self-serve path, so evaluation cost is entirely front-loaded into a sales cycle
- ✗Competing directly with the native agents its own connectors depend on — Databricks Genie and Snowflake Cortex Analyst ship inside the warehouse the customer already pays for
Pricing
Enterprise (demo required)
Contact for pricing
- ✓15+ data source connectors
- ✓Plain-language querying with visible SQL
- ✓Human review workflow and decision audit trail
- ✓Living decks, docs and dashboards
- ✓Data stays in customer cloud and region
- ✓SOC 2, ISO 27001, ISO 42001 and GDPR compliance
No pricing is published anywhere — the homepage, about page and blog all route to a demo request, and there is no free tier, no trial and no self-serve sign-up, which is normal for a company three days out of stealth. Worth noting for cost modelling: the founder has explicitly positioned the company against consumption pricing, arguing customers prefer a vendor not incentivised to sell as much compute, storage or tokens as possible, which implies a seat or platform subscription rather than usage metering. Since QueryStory provides no storage or compute and runs against data in your own cloud, warehouse query costs remain on your existing bill.
Security & Compliance
Sources
This page was written from 5 sources, 2 on domains other than querystory.ai.
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
turbopuffer
Vector and full-text search built object-storage-first, at roughly a tenth the cost of RAM-resident vector databases
Ellis
AI-native operations platform that reconciles private credit fund data and runs the close, reporting and monitoring
Monte Carlo
Data and AI agent observability for enterprises running production AI
DataBahn
Agentic data control plane that cuts security telemetry 40-70% before it reaches your SIEM