The seed of Vellum traces back to challenges Julianna Muir, the company’s founder and CEO, identified consistently throughout her career.
She began as a reinsurance broker at Guy Carpenter, later joined McKinsey’s insurance practice, and then moved into the startup world, including helping build an MGA focused on small business insurance. Across those different roles and vantage points, she kept encountering the same underlying problem: despite major advances in technology, remarkably little had changed in the way the industry sends, receives and analyzes data, which has ‘huge costs’ for everyone in the value chain.
“As a consequence, the back of the value chain is forced to become extremely reactive,” said Muir. “It can take the reinsurer 90-plus days to receive data on risks that are underwritten today, and the level of fidelity associated with that data can decrease by over 90% creating huge costs alongside a lack of transparency and modern connectivity.”
Vellum was built to bridge this gap, enabling the back of the value chain to engage proactively with the front end to drive both mutual profitability and earlier identification of emerging risks.
Muir stresses that it is often the MGA/coverholder market that drives innovation. If firms do not enable the back of the value chain to provide guidance and input as that innovation evolves, she says, a stagnant cycle can emerge: “Everybody needs top line growth, and the coverholder space is exciting, but when the loss ratios start coming in, the coverholders and MGAs become the villain.” For Muir, genuine collaboration is the solution— and she believes it can only happen if everyone has the right data and information.
What AI-Ready Data Looks Like
Fixing data across the value chain is critical, but what does AI-ready data actually look like in insurance?
In Muir’s view, it starts with developing a shared language.
“What one person calls a ‘product’, another calls a ‘line of business’, another calls a ‘coverage’,” she said. “ When you think about being AI-ready, you need a level of standardization to be able to have a clean, trusted data layer for that AI to work off of.”
She added, “You need to solve for a consistent data hierarchy and naming conventions that the AI can draw from regardless of where or how that data was ingested.”
The Data-Quality Challenge
Data quality remains a major challenge for carriers and reinsurers. In Muir’s view, it may be the biggest problem facing the industry.
“One of the biggest challenges is that the organizations ultimately underwriting the risk often receive too little data, too late. If you only learn about a risk 120 days after it was written, that information cannot influence the decisions being made today, and it takes even longer to incorporate those insights into your view of the broader portfolio.”
Beyond timeliness, there are also structural problems and inconsistencies in data across the insurance value chain. For Muir, that distinction is critical: the organizations that move fastest to deploy AI will not necessarily be the ones that ultimately get the most value from it.
“It’s not going to be the organization that launches a hundred agentic workflows first that wins the AI race,” Muir emphasized. “It will be the organizations investing today in making their data clean, consistent and trustworthy enough for those agents to operate on.”
The Risk of Poor Data
Muir also believes that poor data is now a bigger barrier to artificial intelligence, than the technology itself.
“AI is not perfect, you can ask a model the same question twice and get two different answers,” she noted. “Despite this, it’s powerful and we’re already seeing where its driving efficiency and insight that would have been close to impossible pre-AI.”
However, Vellum is seeing instances of what Muir labels ‘pilot purgatory’ where many initiatives get started, but few bear fruit.
She continued, “You have two dynamics at play. One, you have a change management challenge that’s separate from the technology. Second, organizations can’t trust the results because it’s built on bad data.”
To break out of pilot purgatory and make these AI initiatives work, the Vellum CEO is adamant that the data has to be structured first, a step that cannot be taken for granted.
Unlocking Legacy Infrastructure
When asked how insurers can unlock legacy data without embarking on costly multi-year system overhauls, Muir highlights another area where AI can solve a key challenge.
Muir explained, “What AI is allowing us to do is to be able to take any data, in any format, in any naming convention, and conform that to a single trusted data layer, therefore unlocking the value of the data without having to rebuild the data’s infrastructure.”
She explained that if you were to look inside a carrier today, you may find 10 different policy admin systems. Instead of having to build everything around these systems, Muir outlines that data from multiple systems can be brought into one trusted layer.
However, despite AI’s capabilities, Muir makes clear that insurers can’t fully surrender to the probabilistic processes of I, without having the right human expertise in the loop.
The Vellum CEO added, “In insurance, a single word or term can have several different meanings, which is why industry expertise is so important. Combining AI’s ability to consolidate unstructured and semi-structured data from legacy systems with the insurance expertise needed to interpret it is what ultimately unlocks the value of that data.”
Positioned as the agentic data layer for delegated authority and reinsurance, Vellum addresses this bottleneck headon by turning fragmented, delayed data into trusted, actionable intelligence. The platform enables insurers to act on portfolio insights earlier—delivering 90% faster trend identification, 50% operational efficiency gains, and 80% faster financial close.
The Accessibility-Governance Balance
How can insurers balance AI accessibility with data governance and security? From the perspective of Muir, this comes down to the belief that AI has to be done within a firm’s own security boundaries.
Vellum has also accelerated its efforts to educate employees on emerging security threats, recognizing that while AI offers significant potential, it also introduces new risks that must be carefully managed.
“We’ve always taken a security-first approach,” said Muir. “This is our clients’ data, and it is one of their most valuable assets. We have a responsibility to protect it.”
The AI Differentiator
For insurers, the dividing line is increasingly whether they are making their data AI-ready or simply bolting AI onto legacy systems.
Muir advises leaders to adopt a first-principles mindset to separate hype from true capability.
“Organizations have to get their data organized before they can really make meaningful progress with AI. It’s about going slow to go fast. Get your data ready and organized before you apply AI on top,” she concluded.
In the end, AI readiness won’t be defined by the increased efficiency of the tools that insurers adopt, but by the data foundation they build for what comes next.
The full InsurTech100, including profiles on each company, can be found here.
Copyright © 2026 FinTech Global
