AIdataNewsGlobal General Briefing: Hex 4.0, $100m, and the Move Toward Agent-Native Data Platforms
The latest Hex release turns a familiar question inside analytics product design — What would we build if we were starting from scratch? — into a concrete public example: rebuild a widely used data platform for AI agents rather than continuing to attach conversational features to existing software. A few months ago, after a best-ever quarter, the company said it had chosen that path. The result, Hex 4.0, is presented alongside a $100 million fundraise at a $1.2B valuation, and it gives observers a compact case study in how data work may be reorganized around code, controls, and agents.
What the announcement is and why it matters
The change can be read as a platform decision: Hex is not only adding another AI assistant to a BI workflow; it is trying to reset the foundation on which data applications are built. The product had been built for 6 years and used by over 2,500 customers, and the unusual choice was to reconstruct rather than keep retrofitting. That history matters because the claim is not simply that agents are useful, but that older architecture assumptions can limit how far agentic work can go.
After 4 years spent moving AI into data, the team concluded that a from-scratch approach could make agent-native workflows easier to create. The framing question, What would we build if we were starting from scratch?, was asked earlier this year. Founded in 2019, Hex began before the current wave of generative AI, so the rewrite is also an attempt to separate its present product from pre-AI assumptions.
Key data points from the story
| Fact | Why it matters |
|---|---|
| A few months ago, after its best-ever quarter | The company describes the decision as coming shortly after a best-ever quarter. |
| Product built for 6 years | Legacy assumptions are deeply embedded in a platform’s architecture and workflow logic. |
| Used by over 2,500 customers | The rebuild was undertaken with real adoption already in place. |
| After 4 years spent moving AI into data | AI integration has been an ongoing effort, not a single late feature launch. |
| Hex 4.0 and $100m financing narrative | The new version is presented as the base for future bets. |
| $100 million raised at $1.2B valuation | The company says the round adds runway for building and expansion. |
The limits of retrofitting AI onto an existing data product
The central argument is that AI had previously been integrated into a product whose data model, architecture, and assumptions about user behavior were set between 2020 to 2021. In that arrangement, the company was far from agent native, even if the phrase was attractive. The practical consequences became visible in several places.
First, the agent harness was becoming bloated with hundreds of tools. Those tools were needed to teach agents how to use Hex. Second, compatibility with external agents suffered because those products had to reach Hex through a keyhole via MCP rather than working with something more native. Third, the compute engine was not built for high-speed, high-concurrency, iterative workflows that agents seem to prefer.
The environment around data teams also broadened. The company described work happening in tools such as Cursor, Codex, Cognition, and Claude, as well as custom agents. It also pointed to users vibe coding dashboards and reports more than they used legacy BI tools, emailing HTML files without provenance, context, or permissions. The result was messy: a product could be useful, while still failing to create the trusted, governed environment required for AI-assisted analysis.
The AI track record before 4.0
The team also believes it did strong work adding AI over those years. It points to a text-to-SQL prototype using GPT-3 that arrived before ChatGPT, Magic features launched in February 2023, the first really good data agent shipped last year, and generative apps that changed the game again earlier this year.
What Hex 4.0 changes
Today, the company introduced Hex 4.0 as a from-scratch rewrite for the era of agents, accelerated by agents writing the code under the guidance of decades of combined expertise from teams that build scaled data platforms. The core assumption is that everything is code, and that teams need context, controls, and compounding to make it trusted and scalable. The product leans into the fact that data work happens in many places, and plain code, not a proprietary YAML format, is how that can make sense.
For readers asking what that means practically, the answer is less about a single interface than about a changed foundation. A detailed blog post explains the system, and the public framing can be summarized through a list.
- Code as the substrate: The rebuild treats code as the common layer for flexibility, reuse, and agent interaction.
- Agents need architecture, not accessories: The argument is that an old compute and metadata stack can be as constraining as an old dashboard layout.
- Trust is a control problem: Context, controls, guidance, and guardrails are presented as the conditions under which broad creation becomes usable.
- Future capability requires a new base: Some capabilities are described as impossible on the old architecture but simple on the new one.
The funding round behind the pivot
As the plan sharpened, more money was recognized as a way to move faster: scale the team, expand go-to-market, and keep producing great swag. The timing was not easy. During the first part of the year, investors thought all software was going to die and only two companies would remain. Shailendra and the team at Peak XV did not share that view. They were quite confident that model-independent platforms had a bright future, and they were not afraid of the company taking a bold first-principles route to build one.
The announced result is a $100 million financing round at a $1.2B valuation led by Peak XV and 137 Ventures, with participation from Amplify, Avra, a16z, Redpoint, Sequoia, and Zero Prime. That brings more runway to continue building and to place more bold bets in the future.
BI is not the end of software, but the shape may change
The company offers a jab at SaaS apocalypse thinking: a lot of software categories will effectively die over the next couple of years, but categories can also combine and be reborn into new shapes. The old BI paradigm is ending, and attaching a chatbot sidebar to legacy platforms will not change that. New opportunities may be rooted in code, with the context and controls needed to establish trust, governance, and guardrails.
The basic promise of BI is finally close, on this account: everyone can ask questions, and everyone can build. The present moment, however, is chaotic, and much of the trust has gone. That trade-off, broad creation versus reliable governance, is the problem the company says it is obsessed with solving.
A note on the AI-native label
The original narrative includes a wry aside. It finds it funny how many companies in the space that have been around for 8 years or more now call themselves AI native, even though that phrase is the opposite of what the name claims. Its threshold is blunt: if a company has not literally rebuilt its product in the last 24 months, the claim is hard to believe.
FAQ
Why did Hex decide to rebuild?
Because AI had been integrated into an existing product whose data model, architecture, and assumptions were set between 2020 to 2021. The company concluded that the agent harness was becoming bloated with hundreds of tools, external agent compatibility via MCP was limited, and the compute engine was not suited to iterative, high-concurrency work.
What are the most concrete facts in the announcement?
The company describes a product built for 6 years, used by over 2,500 customers, and a new $100 million round at $1.2B. It also lists earlier AI milestones: GPT-3 text-to-SQL work before ChatGPT, Magic features in February 2023, a data agent shipped last year, and generative apps introduced earlier this year.
Is this a statement that BI tools are obsolete?
Not exactly. The company argues that the old BI paradigm is ending and that a chatbot sidebar will not reverse that. It also notes that categories can combine and be reborn into new shapes over the next couple of years.
What does the funding add?
The $100m round is tied to scaling the team and expanding go-to-market. Peak XV and 137 Ventures led, with Amplify, Avra, a16z, Redpoint, Sequoia, and Zero Prime. The broader point is that investors still see room for platforms that are not dependent on a single model.
What should readers watch next?
Watch how the rebuilt 4.0 base handles trust, governance, and guardrails while giving more users the ability to build. The article’s AI-native note sets a rough test: if a company has not rebuilt in the last 24 months, labels that it has been around for 8 years or more as AI native are easier to question than to accept.
In short, the announcement is less a product version note than a market signal: if data platforms must serve agents, the debate shifts from dashboard features to architecture, controls, and trust. That is the problem Hex 4.0 is built to address, and the new capital gives the company space to make that case over time.