What Is Golden Analytics and How to Get Started

Golden Analytics is an AI-native business intelligence platform that went public in April 2026 after emerging from stealth with $7 million in seed funding. Built by Francois Ajenstat, Tableau's former Chief Product Officer, the tool is designed to take a user from a raw dataset to a shared dashboard in two clicks. Early access is available at goldenanalytics.com, with general availability expected within weeks of launch.

What Golden Analytics Does

Most business intelligence tools were designed around the assumption that the person doing the analysis knows SQL, understands data modeling, and is comfortable spending an afternoon building a report. That assumption has always excluded a large share of the people who actually need answers from data.

Golden Analytics takes a different approach. Upload a file or connect a data source and the platform automatically interprets the data, identifies relevant metrics, and generates a set of visualizations without the user configuring anything first. The company describes the experience as going from a raw dataset to a shared dashboard in two clicks.

In a demo published at launch, Ajenstat walked through a raw e-commerce dataset. The platform surfaced a complete dashboard covering sales trends, regional breakdowns, and profitability comparisons within seconds. A storytelling agent generated a written narrative identifying which regions showed the widest profit margins and where performance was declining. No SQL was written. No axes were dragged into place. No visualization type was manually selected.

The output is interactive. Users can ask follow-up questions in natural language: add a filter, change the date range, or drill into a specific segment. The AI can also surface additional questions worth investigating, acting as an exploration partner rather than just a report generator.

Under the hood, Golden routes approximately 120 different LLM calls through an orchestration layer that assigns each task to the best-fit model. Claude handles data analysis; Gemini handles visual design decisions. This architecture, which Ajenstat describes as a platform of AI specialists rather than a single generalist agent, limits the blast radius of any one model failure. When a generalist model handles both analytical reasoning and visual layout simultaneously, errors in one layer compound errors in the other. Separating the tasks produces more consistent results across different kinds of datasets and different kinds of questions.

The Slider of Autonomy

Golden's central feature is a control the company calls the Slider of Autonomy. It lets users set how much of the analytical work the AI handles versus how much they handle themselves, and that setting can change from task to task.

At the fully automated end: upload data, receive a finished dashboard, review it, and share. No intermediate steps. This is useful for business users who work with data regularly but spend their energy on interpreting results rather than producing them.

At the manual end: use the AI as an assistant that suggests approaches, explains its reasoning, and handles repetitive formatting, while the user controls every analytical decision. Experienced analysts who need to understand and justify every step of their methodology will work here.

The middle of the slider is where most users will spend most of their time. A weekly revenue report might run fully automated. An analysis going to a board or an auditor might require hands-on review at each stage. The slider is a runtime control that changes with the stakes of the task, not a one-time setup choice.

This design addresses a specific criticism that has followed AI analytics tools since their early versions: that full automation removes the analyst from the process and makes results hard to audit or defend. When an automated analysis produces a number that a regulator or a CFO challenges, someone has to explain where it came from. A platform that lets the user stay in the loop at variable depth gives that person a defensible record of what happened at each stage.

As Ajenstat puts it: "Analytics tools have spent decades asking humans to adapt to software. We built Golden to flip that. The software adapts to you, so you can focus on the insight, not the mechanics."

Who Built Golden Analytics

Francois Ajenstat spent three decades tracking the business intelligence industry through each of its major transitions. He started at Cognos during the first generation of enterprise BI, when reports ran overnight and were reviewed by senior staff in quarterly meetings. He moved to Microsoft and spent a decade in product roles across SQL Server and Office before joining Tableau in its early years.

At Tableau, he served as Chief Product Officer for more than seven years, guiding the product through its IPO and its $15.7 billion acquisition by Salesforce in 2019. After Tableau, he spent nearly two years as a venture advisor at NEA and served as CPO at Amplitude, a product analytics company based in San Francisco.

Golden Analytics is his first role as a sole founder and CEO. The team at launch is five full-time employees and a fractional CTO, with engineers drawn from Tableau, Snowflake, Apple, and Microsoft. The company is based in Seattle.

The $7 million seed round was co-led by NEA and Madrona. NEA was an early investor in Tableau. Madrona venture partner Mark Nelson served as Tableau's president and CEO from 2021 to 2022. The investor overlap with Tableau is deliberate: the people backing Golden have direct experience with the last major generational shift in BI, and they are betting a similar shift is underway now.

Ajenstat's stated view of the competitive landscape: the current BI market leaders, Tableau, Power BI, and Looker, are adding AI to products built for a different technical architecture, and the results feel grafted on. Golden is built from the start with AI at the core. Whether that architecture delivers a meaningfully better outcome at enterprise scale is the open question. The founding team's track record is the primary argument in its favor.

How to Get Access

Golden Analytics is in early access as of April 2026. General availability was described as weeks away at launch.

To request access: go to goldenanalytics.com and sign up. Pricing has not been published. The go-to-market model is product-led growth, meaning individual users adopt first and the tool spreads within organizations, consistent with how Cursor and Slack were initially adopted. There is no enterprise sales process at this stage.

Early access includes the core product: dataset upload, automated dashboard generation, the Slider of Autonomy, and the AI narrative analysis feature. The team reports roughly a dozen users providing active feedback during this window, which is typical for a pre-GA phase focused on rapid iteration over scale.

If you need to analyze data today without waiting for access approval, tools like VSLZ let you upload a file and get charts, statistical summaries, and plain-English breakdowns from a single prompt with no setup required.

What to Test in Early Access

Run your own data through it. The two-click demo works on a clean dataset prepared for the walkthrough. The useful test is whether the platform handles your actual data, which almost always contains missing values, inconsistent category labels, mixed date formats, and columns that require domain knowledge to interpret correctly.

Three things worth testing:

First, upload a dataset you already understand well. If you know the expected answer before you run the analysis, you can judge whether Golden's output catches what actually matters versus what looks analytically complete but is contextually irrelevant to your business.

Second, test the Slider at both ends. At full automation, how much of the output is immediately usable without modification? At the manual end, how much control does the interface actually provide, or does the AI still make choices you cannot override?

Third, check the narrative analysis on a domain-specific dataset. AI-generated narratives are generally accurate at identifying statistical patterns but can miss domain context. A regional sales breakdown might correctly flag that the northeast underperformed in Q1 but miss that this was expected due to a known product launch delay. Testing the narrative on data you know well reveals how much editing it requires before it can go to a stakeholder.

Teams operating with a mix of technical analysts and non-technical business users will find the slider concept most directly useful. Whether it serves both audiences well without compromising either is only verifiable on real work, not a demo dataset.