QueryStory came out of stealth on 26 August 2026 with an unfashionable pitch. Everyone else in enterprise AI is selling speed. This company is selling doubt — or rather, the machinery that resolves it. Its founders think the reason AI analytics has stalled inside big organisations is not that the answers are slow, but that nobody can tell whether they are true.
That is a narrow claim, and it is worth taking seriously, because the man making it spent six years at Google building the tooling that security analysts use to reconstruct what actually happened inside a network. Shapor Naghibzadeh’s route to QueryStory runs through the 2009 Operation Aurora intrusion, through Google’s X Labs, and through a cybersecurity company called Chronicle. The through-line is verified knowledge — knowing not just what the system says, but why it says it.
The company raised a $6 million seed round in late 2025 from Brightmind Ventures and New York Life Ventures at a $60 million valuation, and spent the months since piloting the product with customers before making any of it public. We have written before about what it costs to buy data analytics capability in the UK and how to prepare business data for AI; this launch is a bet that the missing piece was never the model.
What follows is what QueryStory actually announced, how the product works, who is behind it, where the trust argument holds up, and where a reasonable buyer should still push back.
Table of contents
- What QueryStory Announced on Launch Day
- From Operation Aurora to QueryStory
- The Trust Gap QueryStory Is Selling Against
- How QueryStory Works in Practice
- QueryStory Versus Frontier-Lab Co-Working Tools
- The Numbers Behind the QueryStory Seed Round
- What the QueryStory Test Drive Showed
- What QueryStory Means for Technology Buyers
- Risks and Open Questions Around QueryStory
- Frequently Asked Questions
- References and Further Reading
What QueryStory Announced on Launch Day
The stealth exit was a product reveal with a funding disclosure attached, not a funding announcement with a product bolted on — the round closed nine months earlier.
The timing
QueryStory emerged from stealth on 26 August 2026, in a TechCrunch piece by Tim Fernholz published at 6:00 AM Pacific. The seed round it disclosed had closed in late 2025, so the company had roughly three quarters of a year building and piloting before it said anything publicly.
The founding team
Naghibzadeh is CEO. He co-founded the company alongside CTO Stanley Yang, a former Google colleague who was lead engineer at EvolutionIQ, and CPO David Glusic, a veteran of Accenture. That is a deliberate spread: infrastructure, applied machine learning, and enterprise delivery.
The money
The seed was $6 million at a $60 million valuation, led into by Brightmind Ventures and New York Life Ventures. A $6 million raise against a $60 million post-money figure means roughly ten per cent of the company changed hands — an ordinary seed shape, not a hype-cycle outlier.
The product in one line
QueryStory is an agentic data platform that turns a sequence of questions asked of a database into a defensible narrative — with the SQL shown, a confidence indicator attached, and a human review step recorded in the system.
Launch facts at a glance
| Item | Detail |
|---|---|
| Company | QueryStory |
| Out of stealth | 26 August 2026 |
| Founded | Late 2025 |
| Seed round | $6 million |
| Valuation | $60 million |
| Investors | Brightmind Ventures, New York Life Ventures |
| CEO | Shapor Naghibzadeh |
| CTO | Stanley Yang |
| CPO | David Glusic |
| Target buyer | Large enterprises with big proprietary databases |
| Model strategy | Model-agnostic; currently uses frontier-lab models |
From Operation Aurora to QueryStory
The founder’s biography is not decoration here. The product is a direct port of a security workflow into a business-analytics setting, and the port explains most of the design decisions.
The 2009 war room
Naghibzadeh was a Google sysops engineer in 2009 when Chinese state-backed hackers targeted the company in the campaign that became known as Operation Aurora. He was pulled into a hastily assembled war room and asked to explain what was happening inside Google’s servers. The job was reconstruction under pressure.
What that taught him
Tracing an attack across disparate networks is expensive and slow, and the output is only worth anything if every step of it can be shown. That is the lesson the QueryStory pitch rests on: verified knowledge beats fast knowledge, and the verification has to be part of the artefact rather than a separate audit.
Six years, then Chronicle
He spent the following six years at the intersection of data and cybersecurity, using Google’s resources to build tools that let security analysts query genuinely complex data. In 2016 he co-founded Chronicle inside Google’s X Labs, a startup built to give other companies that same capability.
Why analytics, and why now
Last year, watching large language models take a bigger role in data analysis, Naghibzadeh concluded the security-investigation technique generalised. The same pattern — ask a series of questions, assemble the answers into something defensible — applies to a sales pipeline or a finance variance just as well as it applies to an intrusion.
Where the name comes from
“You get this pattern of an investigation — you ask a bunch of questions of the data, and after you have been able to ask a number of questions, you assemble that together into a narrative,” Naghibzadeh told TechCrunch. “That became the genesis for the name QueryStory. It’s about telling stories with data, right? Putting a narrative together that’s grounded in truth.”
The founding team’s backgrounds
| Founder | Role | Background |
|---|---|---|
| Shapor Naghibzadeh | CEO | Google sysops engineer during Operation Aurora; six years on data and security tooling; co-founder of Chronicle in Google X Labs, 2016 |
| Stanley Yang | CTO | Former Google colleague; lead engineer at EvolutionIQ |
| David Glusic | CPO | Accenture veteran |
The Trust Gap QueryStory Is Selling Against
Every AI analytics vendor claims to solve accuracy. What makes this argument different is that it identifies an organisational failure, not a model failure.
Everybody gets their own version of the truth
Naghibzadeh’s description of the failure mode is precise. When companies connect their data to a chat interface, “you get hundreds or thousands of people within an organization all asking their questions and getting their version of the truth and putting that in a slide deck and sharing it — you just end up with this huge sprawl of content, and there’s no real place to hang that content that ties back to the data.”
Why that is worse than a wrong answer
A single wrong answer gets caught. A thousand mutually inconsistent answers, each embedded in a slide deck with no link back to the query that produced it, cannot be caught at all — there is nothing to check against. The sprawl is the problem QueryStory is built to prevent, and it is a governance problem before it is a modelling one.
The brittleness argument
“AI is more brittle than people realize when it comes to like building things that have to be durable and have large scale businesses relying upon them,” Tayler Sipperly, a partner at Brightmind Partners, told TechCrunch. That is an investor describing the risk in their own portfolio company’s category, which is unusually candid.
What “trust” means as a product
“What we’re doing is bridging that trust gap for AI to give enterprises answers that they can act on,” Naghibzadeh said. “Instead of, you know, like renting human judgment and armies of forward deployed engineers, we productized that.” The claim is that the consulting layer around enterprise AI — the humans who check the output — can be turned into software.
Where our own coverage lands
We have argued something adjacent for a while: that most failed analytics programmes fail on inputs and governance rather than on tooling. Our data governance framework guide and the data quality assessment checklist both make the same point from the other direction. QueryStory is attacking the output side of that same gap.
How QueryStory Works in Practice
Four mechanisms carry the product, and each one exists to make a step that is normally invisible into something a person can inspect.
The SQL surfaces automatically
When QueryStory answers a question, the SQL it wrote to get there is shown without being asked for. That sounds minor. It is the single biggest difference between this and a general-purpose chat tool, where the query is generated, executed and discarded unless the user explicitly requests it.
Confidence indicators
The platform breaks out a confidence indicator explaining why its agents believe an analysis is accurate. TechCrunch’s reviewer singled this out as the most notable feature of the product he tested. It is a calibration signal rather than a guarantee, but it tells a reader which parts of an output to interrogate first.
Human review, recorded
Users can flag an analysis for a human colleague to review, and that review is then recorded inside the platform. This is the piece that turns an answer into an artefact with provenance. The reviewer’s judgement lives next to the query and the data, not in a Slack thread that nobody can find in six months.
Narrative assembly
Individual answers get assembled into a story — decks, documents, dashboards and messages that refresh as the underlying data moves. The company frames this as the difference between a one-off answer and a decision that stays current. Static outputs go stale; the whole point of the narrative layer is that it does not.
Model-agnostic architecture
QueryStory is built to be model-agnostic, though in practice it currently runs on the latest models from the frontier labs. That is a hedge with a commercial motive attached, and the motive is worth spelling out, which the next section does.
The mechanism, feature by feature
| Mechanism | What it does | What it is meant to prevent |
|---|---|---|
| Automatic SQL display | Shows the query behind every answer | Unverifiable outputs nobody can reconstruct |
| Confidence indicator | States why the agents believe the analysis holds | Uniform confidence across strong and weak answers |
| Recorded human review | Captures a colleague’s sign-off in the platform | Review that happens but leaves no trace |
| Narrative assembly | Builds decks, docs and dashboards that refresh | Slide-deck sprawl disconnected from the data |
| Model-agnostic layer | Swaps the underlying model without a rebuild | Lock-in to one lab’s pricing and roadmap |
QueryStory Versus Frontier-Lab Co-Working Tools
The competitive question is obvious, and to its credit the company answers it directly rather than pretending the frontier labs are not in the room.
The manual loop today
TechCrunch describes an executive at a technology company querying a corporate database using Claude Cowork, then asking the model to show the SQL it had written, checking that the queries made sense, and forwarding them to a data analyst for human review. Every one of those steps is a deliberate human action.
What the product automates
In QueryStory that loop is the default path rather than a discipline the user has to remember. The SQL appears without a prompt; the review is a flag rather than an email; the outcome is stored rather than narrated. The tools do the same work — one of them just does not depend on the user knowing to ask.
Steps that need a human hand-off
Where the frontier labs still win
Breadth. A general-purpose agent handles the request that does not fit the schema, the ad-hoc document, the question nobody anticipated. A purpose-built analytics tool is narrower by construction. Naghibzadeh’s counter-argument is that the narrowness is the feature — a tool that understands and preserves context can be more efficient and more accurate than a generalist.
The pricing argument
This is the sharpest part of the pitch. “We have a lot of things going for us here in not being one of those companies that built their business around this consumption model of compute or storage or tokens,” Naghibzadeh said. A vendor paid per token has no incentive to make the answer cheap. A vendor selling trust in the answer does.
What the CFO is actually buying
“The thing that we are selling is the trust in the answers, right?” he said. “The thing that we’re selling them is the value that we’re adding to the business, and our whole goal is giving the CFO the ability to understand ‘what is this thing going to cost?'” Predictable pricing is being positioned as a trust feature in its own right.
Head to head
| Dimension | Frontier-lab co-working tool | QueryStory |
|---|---|---|
| Query visibility | On request | Automatic |
| Review workflow | Outside the tool | In-platform and recorded |
| Confidence signal | Not broken out | Explicit indicator |
| Scope | General purpose, very broad | Purpose-built for data analysis |
| Commercial model | Consumption of tokens and compute | Positioned away from consumption pricing |
| Model dependency | Single lab | Model-agnostic by design |
The Numbers Behind the QueryStory Seed Round
Three figures were disclosed, and a little arithmetic on them says more about the company’s position than the press language does.
Ten per cent of the company
A $6 million raise at a $60 million valuation is exactly one tenth. That is a conventional seed structure, and it implies the founders kept the great majority of the equity through a round closed before the product was public.
The strategic investor
New York Life Ventures is a corporate venture arm, and its partner Tim Del Bello did not just invest — he uses the platform. He has been running a quarterly business review through it, replacing work previously done by several people, and hopes to convert that review into a real-time dashboard.
The buyer described in his own words
“The product was built for people like me: decision-makers seeking the ground truth who need to work with complex, disparate data sources but don’t have a data science or BI team at their disposal, especially when operating in a highly regulated industry,” Del Bello told TechCrunch. That is the target customer, stated precisely.
Why a regulated investor matters
Insurance is one of the most heavily supervised industries there is. An investor from that world adopting the tool internally is a stronger signal than a logo on a slide, because a regulated firm has to be able to explain any number it acts on. That is exactly the property QueryStory claims to supply.
The seed round in proportion
What the QueryStory Test Drive Showed
The launch coverage included a hands-on test, which is more than most stealth exits offer, and the result is the most concrete evidence in the story.
The space-activity database
TechCrunch’s reporter handed the platform a database of space activity — the kind of dataset used to understand what operators like SpaceX are doing in orbit. It is proprietary, messy, and not something a general model has memorised.
Weeks compressed into hours
QueryStory produced a visualisation of that data in a few hours. The reporter had previously done the same project with a developer, and it took several weeks. It also produced dashboards and analysis, alongside the confidence indicator that explained the agents’ reasoning.
The order-of-magnitude claim, plotted
What a single test does not settle
One dataset, one reviewer, one task. The comparison is against a bespoke development project rather than against a modern BI tool, which is a generous baseline. It demonstrates capability; it does not establish that QueryStory beats the alternatives a buyer would actually shortlist.
The part that is genuinely novel
The confidence indicator is the detail worth holding onto. Producing a chart quickly is now table stakes. Explaining, in the same artefact, why the system thinks the chart is right is not, and it is the feature most directly descended from the security-investigation heritage.
What QueryStory Means for Technology Buyers
You do not have to buy this product to take the argument from it. Four things follow for anyone evaluating AI analytics in the next year.
Make query visibility a requirement, not a nice-to-have
If a tool cannot show you the query it ran, you cannot audit the answer. Put it in the requirements document. This applies whether you end up with QueryStory, a frontier-lab agent, or the data visualization stack you already own.
Ask where the review is recorded
Human review almost always happens somewhere. The question is whether it leaves a durable record attached to the analysis, or evaporates into a chat message. A review with no artefact is not a control, and it will not survive an audit.
Model the cost you will actually pay
Consumption pricing is not automatically bad, but it makes the annual bill a function of adoption, which is precisely the variable you are trying to grow. Whichever way you go, model it. Our Power BI implementation cost breakdown shows how quickly the non-licence line items dominate.
Fix the inputs first
A platform that shows its working will show you, faithfully, that your data is inconsistent. That is useful, and it is also a reason to do the groundwork before procurement rather than after. The AI-ready data checklist covers the ground; sound AI strategy work starts there too.
A buyer’s due-diligence table
| Question to ask | Why it matters | A weak answer sounds like |
|---|---|---|
| Can I see every query the system ran? | Without it there is no audit trail | “You can ask the model to show you” |
| Where does a human sign-off get stored? | Review with no record is not a control | “Teams usually handle that in Slack” |
| What does the confidence score mean? | A number with no definition is decoration | “It reflects the model’s certainty” |
| What happens when the model changes? | Outputs can move under a stable question | “We always use the latest model” |
| Where does our data physically sit? | Residency drives most compliance answers | “It’s all in the cloud” |
| What is the bill at three times the usage? | Adoption is the thing you want to grow | “It scales with your consumption” |
Risks and Open Questions Around QueryStory
Four things are genuinely unresolved, and none of them is settled by a launch article or a demo.
No named production customers
The company says it piloted with customers through 2026, but the only named user in the coverage is a partner at one of its own investors. That is not disqualifying for a stealth exit — it is normal — but it means there is no independent reference account in public yet.
A confidence indicator is not a correctness proof
Showing why a system believes an analysis is accurate is a real improvement on showing nothing. It is still the system’s own account of itself. Calibration of these indicators, across question types and data shapes, has not been published, and it is the number that would matter most.
Model-agnostic today, dependent in practice
Being architecturally model-agnostic while running on frontier-lab models means the economics still route through those labs. If the competitive claim is that a specialist beats a generalist, the specialist is currently building on the generalist’s substrate — and paying for it.
The category is filling up fast
Trust, lineage and governance layers over enterprise data are attracting a lot of capital right now, and the frontier labs are moving into the same workflow from above. A $6 million seed buys a head start, not a moat. The AI models and tools hub tracks how quickly this ground shifts.
Frequently Asked Questions
What is QueryStory?
QueryStory is an agentic data analysis platform for large enterprises, founded in late 2025 and out of stealth on 26 August 2026. It answers questions across proprietary databases, shows the SQL it used, attaches a confidence indicator, supports recorded human review, and assembles the results into narratives, dashboards and documents.
Who founded it?
CEO Shapor Naghibzadeh, a former Google sysops engineer who co-founded Chronicle inside Google X Labs in 2016; CTO Stanley Yang, a former Google colleague and lead engineer at EvolutionIQ; and CPO David Glusic, an Accenture veteran.
How much funding has QueryStory raised?
A $6 million seed round in late 2025 at a $60 million valuation, from Brightmind Ventures and New York Life Ventures. No later round has been disclosed.
What problem is it trying to solve?
Answer sprawl. When thousands of people query a company’s data through a chat interface, each gets their own version of the truth and pastes it into a deck with no link back to the underlying data. The platform is built to keep every answer tied to its query, its confidence and its reviewer.
How is it different from Claude Cowork or similar tools?
The verification steps are the default rather than something a user has to request. SQL surfaces automatically, analyses can be flagged for human review, and those reviews are recorded in the platform instead of happening over email.
What does model-agnostic mean here?
The platform is not architecturally tied to one lab’s model and can swap the underlying model out. In practice it currently runs on the latest frontier-lab models, so the independence is a design property rather than a present-day fact.
Is QueryStory cheaper than a consumption-priced tool?
No pricing has been published. The company positions itself against per-token and per-compute pricing and says its goal is letting a CFO understand what the system will cost, but the actual commercial terms are not public.
Has anyone tested the product independently?
TechCrunch ran a hands-on test with a space-activity database and reported a visualisation produced in a few hours against several weeks for an equivalent project built with a developer. That is a single reviewer and a single dataset, not a benchmark.
What kind of data sources does it connect to?
The company describes connections to structured and unstructured sources — warehouses such as Snowflake, Databricks and BigQuery, CRM systems, call recordings, presentations and spreadsheets — while respecting the permissions already in place.
Should a UK mid-market business be looking at this?
Probably not yet as a purchase, since QueryStory is aimed at large enterprises with big proprietary databases and has no public pricing. The requirements it implies — visible queries, recorded review, defined confidence, predictable cost — are worth applying to whatever you are evaluating today.
References and Further Reading
More AI coverage: explore Progressive Robot's AI Models, Tools & Releases hub — hands-on reviews, setup guides and benchmarks in one place.