Written by Sabine VanderLinden

• Why 95% of enterprise AI projects fail to scale: AI Horizons 2030 summit reveals that despite 74% of insurance executives experimenting with AI, most struggle to move beyond pilot purgatory—with venture investors, incumbent carriers, and tech scale-ups sharing proven strategies to bridge the $1.8 trillion protection gap through hybrid build-buy-partner approaches.
• Insurance industry AI strategy 2030: Exclusive insights from 50 senior insurance executives, VCs from IAG Firemark Ventures, Assurant Ventures, FinTLV, and EOS Ventures reveal that winning insurance companies are combining internal AI capabilities with strategic startup/ scaleup partnerships and Big Tech platforms—not choosing between build, buy, or partner, but mastering all three simultaneously.
• Future of AI in insurance beyond 2025: From algorithmic underwriting to agentic AI claims automation, this ITC Vegas 2025 summit addresses the provocative questions facing insurance leaders—including whether AI will eliminate human underwriters, how to govern autonomous insurance systems, and which startups/ scaleups have the architectural differentiation to compete against OpenAI's 31x funding advantage.

Why AI Horizons 2030: Provocations and Possibilities?

I had the privilege of hosting AI Horizons 2030: Provocations and Possibilities at ITC Vegas on October 14, 2025 – an exclusive executive summit that brought together about 50 senior insurance leaders, investors, and innovators to step into the future of insurance. The theme was bold: we explored how ubiquitous AI, robotics, and human ingenuity could redefine risk, value, and trust in our industry over the next decade. AI is the backbone of innovation in modern computing, unlocking value for individuals and businesses. Artificial intelligence has its roots in computer science, drawing on programming, algorithms, and the study of computational systems to drive advancements. The sense of urgency was palpable – a recent stat revealed that 74% of senior industry leaders are already experimenting with AI and emerging tech, yet most insurers struggle to scale these innovations enterprise-wide. This summit was designed to bridge that gap between experiment and transformation, through a series of provocative discussions and interactive polls.

In my opening remarks, I invited the audience to imagine insurance in an “autonomous world” – one where AI handles everything from underwriting to claims. Would that world be a utopia of efficiency and personalization, or could it erode the human touch that insurance is built on?

These provocations set the stage for three deep-dive panels and executive discussions, each tackling the future from a different angle (investors, incumbent insurers, and tech scaleups), followed by an actionable conclusion. Throughout the afternoon, we also used real-time Slido polls and word clouds to gauge the room’s sentiment on key questions – adding a dynamic, interactive element that kept everyone engaged. The summit emphasized that AI is not only transforming insurance, but is also revolutionizing various industries, demonstrating its versatility and broad impact.

What follows is my recap of the summit’s highlights. I’ve distilled each panel’s most compelling insights and debates to give you a sense of being in the room with us, too.  

AI Fundamentals and Applications

Let's be real about AI. It's not just transforming how companies innovate—it's completely flipping the script on problem-solving and value creation. And if you're still thinking of AI as some futuristic concept? You're already behind. At its core, AI is this incredible suite of technologies that lets computers do things we once thought only humans could handle—seeing, understanding our messy language, crunching through data mountains, and actually making smart recommendations. In insurance and beyond, these AI systems aren't just unlocking opportunities—they're creating entirely new ways of doing business by automating the complex stuff and making our decisions way smarter.

Here's where it gets interesting. Machine learning is the foundation—think of it as teaching algorithms to spot patterns in data and make predictions that actually matter. But deep learning? That's where the magic happens. We're talking artificial neural networks with multiple layers that work surprisingly like your brain, processing massive amounts of messy, unstructured data—from images to text—stuff that used to make computers completely stumble. These deep neural networks are the reason we can now do image recognition and natural language processing that seemed impossible just a few years ago.

Now, you've got two flavors of AI to wrap your head around. There's narrow AI—the practical stuff that's designed for specific tasks like underwriting or claims processing. And then there's the holy grail: artificial general intelligence, which would basically match or beat human intelligence across everything we do. While that second one's still a long-term dream, today's narrow AI? It's already delivering real, tangible results that are changing how businesses operate.

And here's something cool that's happening right now. Venture client units—think BMW Startup Garage—are absolutely crushing it when it comes to leveraging AI and spotting promising startups. They're acting as early clients for these innovative solutions, which means large companies get access to cutting-edge tech before everyone else jumps on board. It's brilliant, really—this approach doesn't just speed up innovation, it helps startups polish their products for real-world use. In insurance specifically, venture clienting is helping companies stay way ahead of the curve by integrating AI-driven startup solutions that boost efficiency, accuracy, and make the customer experience actually enjoyable.

So, let’s delve now into the conclusions from our first panel discussion.

The Investor Perspective: Betting on Promising Startups for the Next Decade

The first discussion, “Betting on the Next Decade: Where Will Value Emerge in Insurance?”, was a venture investor-focused panel that I facilitated. Panelists included leading InsurTech venture capitalists from IAG Firemark Ventures, Assurant Ventures, FinTLV, and EOS Ventures – people who literally bet on the future through their investments. We explicitly contrasted during our session venture capital—which involves direct investment in startups for financial returns—with other innovation models like the venture client approach, which focuses on early-stage engagement and operational collaboration without significant equity investment. I opened with a trillion-dollar question: 

Given that an estimated 95% of AI projects fail in corporate settings according to MIT, how should insurers play it – build AI capabilities in-house, buy AI startups, or partner with Big Tech giants? 

The implication was clear: choosing the right innovation strategy could determine who survives into 2030. It is important to remember that the venture client model benefits corporates by providing quick access to new solutions while giving startups early revenue. Early collaboration with startups allows corporates to request customized iterations of solutions, ensuring that the innovations align closely with their specific needs and challenges.

To get a pulse from the audience, we ran a Slido poll on this very question. Insurers had four choices: Build In-House, Buy Startups, Partner with Big Tech, or Hybrid (combine all three). 

The results were telling – a strong plurality voted for a Hybrid approach, reflecting a recognition that no single strategy is sufficient. In follow-up discussions, many executives shared that they’re investing internally in AI talent while also partnering with startups and tech giants. Only a small minority believed in a pure “build it ourselves” path or putting all bets on Big Tech. This aligns with what I’ve been hearing anecdotally: insurers feel they must do it all – cultivate internal AI capabilities, and tap into external innovation – to keep up with the breakneck pace of change. As large corporations, insurers are leveraging their extensive resources, expertise, and internal programs to foster innovation, support startups, and promote an entrepreneurial culture. For instance, the shift to autonomous vehicles is expected to significantly reduce accident frequency, affecting auto insurance premiums by 30% to 50% over the coming decades. This means we will need to invest in autotech as much as risk-mitigating solutions. The venture panelists had nuanced takes on this. One cited that even though InsurTech funding has matured with “substantially higher-quality startups” after 2022, many young companies struggle to achieve profitability while pouring money into AI.  

This means investors are scrutinizing startups not just for shiny innovation, but for sustainable business models. Gil Arazi of FinTLV noted that InsurTech as a buzzword has cooled down: “InsurTech is practically dead; incumbents only want scale-ups now.” What he meant by this is that insurers are no longer interested in innovation for its own sake; they want proven solutions that can plug into their business and deliver results. 

Sam Evans from EOS Ventures echoed this, explaining how his fund now evaluates startups’ architectural differentiation – for example, does a startup have an AI architecture or data advantage that incumbents can’t easily replicate? 

A key theme was that value in 2030 will concentrate around those who control critical data and distribution or have a unique tech architecture, not just those with flashy front-ends. AI can personalize marketing campaigns and improve customer experiences by analyzing customer data and preferences. The venture client model has been particularly successful in manufacturing, automotive, healthcare, energy, and financial services, demonstrating its versatility across industries. Within insurers, individual business units play a crucial role in identifying and prioritizing challenges, then seeking out the right startup solution to address their specific needs.

I was particularly struck by our discussion on the funding gap between Big Tech and insurance. We noted that OpenAI’s war chest for AI far exceeds the entire InsurTech sector’s funding (i.e., OpenAI raised 31 times more than all InsurTechs combined last year!). Today, they are seeking a $1 trillion IPO valuation.

This led to a frank question: Can insurance-focused startups truly compete in AI, or should we expect the likes of Google, Amazon, and Microsoft to dominate with horizontal AI platforms? The consensus was that insurance problems are unique enough to require specialized solutions, but partnerships with Big Tech will be crucial to access their AI platforms. 

In fact, partnering with tech giants was seen as a likely path for many insurers (hence the popularity of the hybrid strategy). This also means that insurers might adopt venture client models – essentially co-financing or deeply collaborating with startups – to ensure the solutions they rely on remain viable. 

Venture clienting requires less upfront investment compared to other corporate venturing tools, making it an attractive option for insurers. By matching the right startup solution to a business unit’s challenge, insurers can accelerate innovation and operational improvements. This approach not only helps address immediate needs but also opens up opportunities for both insurers and startups to create new revenue streams. Ultimately, the financial motivation for collaboration is clear: generating additional revenue streams is a key driver for both corporates and startups seeking sustainable growth.

Investor Panel Highlights:

• Build, Buy, or Partner? We polled the room on whether to develop AI in-house, acquire startups, or partner with Big Tech. A hybrid strategy won out, signaling that insurers see no one-size-fits-all answer and are hedging their bets across all three options. When collaborating with startups, panelists emphasized that building a long-term relationship is crucial for sustained innovation and long-term success.
• Show Me the Money: Investors stressed profitability + innovation as the winning combo. Post-2022, there’s been a shift from growth-at-all-costs to sustainable growth. Startups must demonstrate not only cool tech, but a clear path to financial viability (e.g., solid unit economics, paying customers).
• Distribution in 2030: We debated whether AI will eliminate brokers/agents or empower them. The panel’s view: hybrid distribution models will prevail. AI will automate simple tasks and match customers to coverage, but human advisors will play a higher-level, advisory role aided by AI. In other words, agents won’t vanish – they’ll evolve (perhaps using AI copilots to serve clients better). The overall premium pool for traditional auto insurance could shrink by over 50% in the long term due to reduced accident frequency.

This underscored a theme: the unsung but critical work of modernizing core systems and processes could unlock massive value, even if it doesn’t make headlines.

During our segment “Meet the Disruptors” three startup founders pitched their visions, injecting fresh energy into the room. Suril Kantaria from Adaptional, provocatively claimed “The Death of the Underwriter (As We Know Them).” In 90 seconds, he explained how their intelligent workflow platform processed 100,000+ submissions with 98% accuracy, essentially automating the rote work of underwriting. He was quick to clarify they’re not replacing human underwriters but “giving them superpowers.” I loved that framing: it resonated with the audience, who know how labor-intensive underwriting can be. 

The idea is that by 2030, underwriters focus on complex, high-level risk judgments while AI handles the grunt work. Scott Cronin from 7Analytics, went with an equally bold provocation: “Betting Against Mother Nature (and Winning).” Scott tackled the climate risk challenge, describing how traditional CAT models have been blindsided by events like record hurricanes and floods. 7Analytics uses 1-meter resolution geospatial data and real claims data (not just theoretical models) to predict losses more accurately. Essentially, he argued that better data and AI can shrink the $1.8 trillion protection gap (the gap between insured losses and total economic losses) by pricing and covering risks that were previously deemed uninsurable.

The third disruptor, Armilla, pitched an equally daring provocation: “Insuring the Insurers (Before They Kill Themselves).” CEO  Karthik Ramakrishnan from Armilla reminded the room that as enterprises race to deploy generative AI, they’re exposing themselves to new forms of liability. Armilla’s solution—developed in partnership with Chaucer Group—offers stand-alone third-party liability coverage for AI systems, addressing the evolving risks of mechanical underperformance in algorithms such as hallucinations (false or misleading outputs), model drift and other deviations from expected behaviour. In other words, Armilla is insuring the AI itself: if an autonomous underwriting model denies a valid claim because of a drifted model or latent bias, Armilla’s policy provides legal defence and covers the resulting liability. Karthik argued that deploying AI without insurance is “corporate suicide” because regulators, plaintiffs and investors will not forgive companies that let rogue algorithms harm customers. His provocation resonated because it reframed trust as a balance-sheet item: risk transfer for AI errors could unlock faster adoption of generative and agentic models. By turning AI governance into an insurable asset, Armilla highlighted that responsible innovation must include a financial backstop

These venture provocations sparked dozens of questions in the Q&A – from data privacy concerns to how to integrate such tools into legacy systems. For me, it was a glimpse into the “art of the possible” – a term Magda Ramada of WTW used in her subsequent talk. Magda took the stage to round out Panel 1 with a brief presentation, showcasing emerging AI applications in insurance. She highlighted examples like parametric insurance triggered by IoT sensors (paying claims instantly for flight delays or crop losses) and using AI agents to continuously monitor portfolios for emerging risks. Her message was optimistic: 

By the end of this first panel, the vibe was set. I could see eyes lighting up in the room, not just from the hype of cool tech, but from the realization that the choices leaders make today (where to invest, whom to partner with, which risks to take) will shape their company’s fate by 2030. As one investor panelist summarized,

“In this industry, we can’t afford to sit and wait. The horizon is 2030, but the decisions are now.” 

The Incumbent Perspective: Rewiring the Core with Venture Client Units

The second discussion shifted to the viewpoint of incumbent insurers. Titled “Rewiring the Core: How Will Incumbents Survive and Thrive in an AI-First World?”, this panel was moderated by Denise Garth (Chief Strategy Officer at Majesco) and featured senior executives from a US insurer (State Farm), a European insurer (MAPFRE), an Australian insurer (IAG) and a global broker and risk management consultancy (WTW). If Panel 1 was about where the money is going, Panel 2 was about how the big ships (established insurance carriers) can turn course in time for the AI revolution. Widespread automation in underwriting and claims processing will displace many traditional roles in the insurance industry, making it imperative for incumbents to rethink workforce strategies and adapt to this transformation. Increasingly, AI and automation are being used to handle repetitive tasks, freeing up human workers to focus on higher-value activities.

A provocative scenario kicked off the debate: Imagine a major disaster – say a hurricane or cyberattack – in 2030 where claims are handled entirely by AI systems, with no human intervention. What could go wrong? 

The initial reaction from panelists was a mix of excitement and caution. On one hand, they acknowledged that AI-driven claims (from first notice to payment) could be lightning-fast and efficient for customers, as AI systems can autonomously perform tasks that previously required human intervention. 

On the other hand, they rattled off a list of nightmare scenarios: algorithms denying legitimate claims due to unseen biases, fraudsters tricking the AI, or simply the lack of empathy when an AI tells a disaster victim, “Your claim is declined.”

We collectively acknowledged that while automation can dramatically speed up processes, insurance is ultimately a promise of trust – and trust can be broken if AI makes decisions that people don’t understand or accept. 

“In a world of self-learning algorithms, what is the irreplaceable value of human expertise?” one panelist asked rhetorically, echoing the panel’s key question. 

The answer that emerged: humans are still critical as the conscience and fail-safe of insurance. By 2030, underwriters, claims adjusters, and agents might look different (more data analysts and relationship managers than paperwork-pushers), but their role in oversight, exception handling, and ethical judgment will be more important than ever, especially as AI augments problem-solving capabilities within organizations. The shift in liability for accidents will move from human drivers to vehicle manufacturers and software developers, challenging the traditional personal auto insurance model. Automated claims processing by AI will allow routine claims to be processed in minutes rather than weeks, which could significantly enhance customer satisfaction. Over 90% of current accidents are caused by human error, and autonomous vehicles are expected to drastically reduce such accidents, further influencing the insurance landscape.

This panel also confronted the gritty operational challenges of becoming an AI-driven insurer. A major point of discussion was the need to modernize core systems with cloud-native and AI-native solutions. Many large insurers still run on decades-old policy admin, billing, and claims systems that were never designed for real-time data or AI integration. 

Over the past years, most big carriers have been on a modernization journey – migrating to cloud-based core platforms, opening up core systems via APIs and ConnectedTech. Magda Ramada noted that event-driven architecture and microservices are becoming the norm for new insurance systems, allowing easier plug-and-play of AI modules. However, this transformation comes with risk and cost. We referenced the example of Duck Creek (a core system vendor) reportedly needing a $900M refinancing – a cautionary tale that even legacy tech partners can face instability and inability to invest in the technology. Speakers on the panel also explained that the way they mitigate this is to ensure that they have a “plan B”: whenever they implement a new AI vendor or platform, they ensure there’s an exit strategy or alternative (either another vendor or an internal capability to take over) in case that partner falters. 

Many corporations create venture client units that act as intermediaries between the company and startups, facilitating smoother collaboration and innovation. It was a sobering insight: resilience now includes your vendor ecosystem. You must stress-test not only your algorithms but also the companies providing them. 

The talent angle was another hot topic. As insurers automate routine tasks, the skills needed in their workforce are changing. Willem Paling, the analytics head from IAG, mentioned they are upskilling their staff – training adjusters to become data-savvy and hiring new profiles like prompt engineers and AI model auditors. But everyone agreed there’s a talent crunch in AI. 

Insurers are competing with tech firms for data scientists, and it’s not easy to attract that talent to a traditionally conservative industry. One idea floated was industry-wide collaboration on certain AI initiatives – for example, sharing non-sensitive data to improve fraud detection models, or even jointly funding an “insurance AI utility” for common functions (much like credit card companies share fraud intel). “Some problems are bigger than any one company – cyber risk, climate risk – we might need to tackle those collectively,” the panel noted. This hinted at a future where coopetition (collaborative competition) becomes more common: insurers pooling resources on foundational AI research or infrastructure, then competing on products and customer experience.

Incumbent Panel Highlights:

• How Far Will Automation Go? By 2030, a significant chunk of underwriting, billing, and claims could be fully automated. One panelist estimated that underwriting might be 80% AI-driven for personal lines, for instance. However, no one on the panel believed the human element would disappear entirely. The consensus: humans + AI together outperform either alone. AI will handle high-volume, straightforward cases, while humans focus on complex or sensitive cases. Crucially, humans will provide the empathy and ethical oversight that machines lack.
• Trust and Ethics: With algorithms taking more decisions, incumbents are focusing heavily on AI governance. We discussed upcoming regulations (e.g., the EU AI Act, California’s AI law in 2026) and the fact that GDPR fines in financial services hit €1.2B+ in 2024. The takeaway: Insurers must ensure transparency, fairness, and accountability in AI models. Several panelists mentioned setting up AI ethics committees internally. I shared that 79% of insurance executives feel a moral obligation to use AI to close protection gaps (e.g., offer coverage to more people) – so there’s ethical pressure both to innovate and to do it responsibly.
• Rewiring Tech Foundations: The panel stressed that becoming “AI-first” isn’t just about apps and models – it demands core system transformation. Cloud migration, data lakes, API integration, and real-time processing are being pursued aggressively now so that by 2030, insurers have a flexible infrastructure for AI. One exec shared that his company is investing in an “AI-ready core” – essentially rebuilding modules of their policy administration system to ingest streaming IoT data and trigger actions automatically. It’s expensive and multi-year, but they see it as survival. “If your core can’t talk to AI, you’ll be stuck in the slow lane,” as was shared. Insurers will need to invest heavily in data protection due to the reliance on data from connected devices, which raises data privacy risks.
• New Skills for a New Era: A live word cloud (shared below) at the beginning of the summit asked: “What’s the biggest risk of full AI autonomy in insurance?” The crowd’s responses floated on the screen in real time. The largest words were “Bias”, “Security”, “Accountability”, and “Outage” – highlighting fears of unfair AI decisions, cyber vulnerabilities, lack of clear responsibility, and tech failures. This led to a series of great group discussions on what new leadership and internal skills will be needed. Responses included: data literacy (every leader needs to understand AI basics), agility (to continuously adapt processes), and crisis management (for when AI misfires). We turned those into a quick list of priorities for the audience to consider at the end of the session. Many heads nodded when someone said:

“In 2030, we might need a Chief AI Ethics Officer just as much as a Chief Marketing Officer.” 

AI-powered chatbots and virtual assistants will provide 24/7 customer support, allowing human agents to focus on complex tasks, which will also require new training and skillsets for employees.

Overall, the incumbent panel exuded cautious optimism. Yes, big insurers have legacy baggage and must move carefully, but they also have scale, data, and customer trust that startups lack. The key message was that:

Regulatory Environment and Compliance

As artificial intelligence becomes more deeply embedded in business operations, the regulatory environment isn't just evolving—it's demanding that we step up. AI systems need to be transparent, explainable, and fair, period. Companies can't afford to treat AI ethics and governance as an afterthought if they want to minimize risks and avoid those unintended consequences that keep executives awake at night. And here's what I've seen time and again: venture client units are often the ones on the front lines, responsible for bringing these new AI solutions into the organization. The question isn't whether you should prioritize this. It's whether you're ready to lead the charge.

Here's the reality check nobody wants to hear: effective risk management starts with robust due diligence, and I'm not just talking about checking boxes. You need to evaluate not just the technical capabilities of AI systems, but their compliance with relevant regulations and industry standards. Frameworks like GDPR aren't suggestions—they set strict requirements for processing personal data, mandating privacy by design and clear accountability for data use. I've watched too many venture client units stumble because they didn't work closely enough with legal, compliance, and IT teams from day one. The truth? Any AI development or deployment that doesn't align with these principles is a ticking time bomb.

So what's the game-changer? Establishing clear governance structures isn't just essential. It's your competitive advantage. This means defining roles and responsibilities for AI oversight, setting up processes for ongoing monitoring, and ensuring that risk management strategies like diversification and regular audits are actually in place and working. The companies that get this right? They're the ones embedding compliance and ethical considerations into every single stage of the innovation process. They're building trust with customers, regulators, and partners while unlocking the full potential of AI-driven solutions. The question is: are you ready to be one of them?

The Tech-Innovator Perspective: Artificial Intelligence Disruptors or Enablers?

The final panel of the day flipped the script to the tech innovators’ perspective. I handed moderation over to Franklin Manchester of SAS, who led a discussion titled “Disruptors or Enablers: How Will Scaleups Shape the Insurance Value Chain by 2030?” This was a lively session featuring founders/execs from four high-growth ventures: Federato (AI-powered underwriting), Charlee.ai (AI for claims), DataHaven (insurance data infrastructure), and Majesco (representing the new breed of core platforms). These are exactly the types of companies that incumbents and investors from the earlier panels are watching closely. The core question: Are these startups friends or foes to traditional insurers? Will they redefine the insurance value chain or simply slot into it? The answer, unsurprisingly, was “a bit of both.” 

Franklin kicked off by asking each panelist to pitch their vision for 2030 in 90 seconds – a mini lightning pitch round. It was electrifying. Megan Bock from Federato painted a picture of underwriting in 2030 where AI copilots help underwriters instantly analyze risk data from hundreds of sources – “Underwriters will go from data hunters to data curators,” she said, meaning AI will gather the data, and humans will make nuanced decisions. Today’s AI models can analyze data from sensors, images, language, and sequences to extract insights, make predictions, and drive automation in insurance workflows. 

Sri Ramaswamy from Charlee.ai described a near future where claims might be settled before the customer even knows there’s an issue – for example, an IoT sensor detects a pipe leak and an AI triggers a proactive payout, all in minutes. 

Yandy Plasencia from DataHaven took a more behind-the-scenes angle: he argued the biggest disruptor will be those who manage to aggregate and pipe high-quality data (from vehicles, homes, wearables, open finance, etc.) into insurance workflows in real time. “The one who controls the data pipelines will have an outsized influence,” he noted, suggesting that a new kind of data utility could emerge in insurance. 

Finally, Manish Shah from Majesco spoke to the platform approach: Manish envisioned incumbent insurers increasingly operating as orchestrators, assembling capabilities from various tech partners (AI providers, data services, customer-facing apps) into a cohesive offering. In that scenario, companies like Majesco aim to be the “platform of platforms” that stitches everything together. Dynamic pricing models will allow auto insurers to adjust premiums based on real-time data from telematics and connected vehicles. Most active venture client units engage with 25-50 startups per year through pilot projects, which helps them identify and integrate innovative solutions effectively. As part of open innovation strategies, some insurers are adopting approaches like BMW Startup Garage and open Bosch, which proactively scout and collaborate with startups to accelerate innovation.

A big debate in this panel was around business models – essentially, which type of insurtech model will produce the next unicorn? Franklin posed it directly: “Will the next insurance unicorn be a data juggernaut, a platform orchestrator, or an AI agent provider?” 

Each panelist, naturally, made a case for their corner of the world. Megan at Federato (underwriting focus) said that vertical AI platforms (those deeply solving a specific insurance function) have the edge, because they can achieve real ROI for insurers faster. Sri Ramaswamy from Charlee.ai countered that an AI agent ecosystem could be revolutionary. For instance, swarms of specialized AI agents handling everything from risk assessment to customer service might dramatically lower costs. Yandy Plasencia at DataHaven jumped in to assert that without the right data infrastructure, none of those AI solutions can work properly: a point that got a lot of nods given data quality is a known pain point. What was enlightening was that they all agreed some convergence is likely: by 2030, the successful ventures might blur these categories, e.g., a platform orchestrator that also controls key data flows and provides AI agents. 

From the audience perspective, it wasn’t so important who was “right,” but rather understanding these emerging roles in the ecosystem. The question for incumbents is how to engage with each: when do you partner versus when do you build a capability in-house versus when do you perhaps acquire a startup outright. (One audience question during Q&A was literally this: “How do you decide whether to compete with or collaborate with a startup?” – which yielded the answer that it’s case-by-case, but early engagement through pilot projects helps gauge whether a startup is a long-term partner or a potential competitor.)

The “Valley of Death” challenge came up next... That is, the notoriously high failure rate of pilots in insurance. Franklin cited a sobering statistic: 73% of InsurTech proof-of-concepts never make it to full production, and those that do can take 14+ months to scale beyond a single department.

 All the founders and scaleup executives had scars and lessons to share on this. Megan, COO of Federato, said one strategy is to align with insurers’ existing workflows, not demand an overhaul. Her company initially positioned itself not as a rip-and-replace underwriting system, but as a layer on top of underwriters’ tools that provides AI insights – essentially “augment, don’t disrupt (too much)”.  

Sri, CEO of Charlee.ai, added that demonstrating quick wins within 3-6 months is crucial. This is because otherwise projects lose executive sponsorship. Her tactic: pick one pain point (for example, automating a particular claims task that’s backlogged) and nail it, rather than selling a grand vision of AI handling all of the corporation's claims from day one. The panel also agreed that co-creation with insurers is a recipe for success, building trust by letting the insurer guide the roadmap. This certainly ensures the solution actually fits their needs and that there are internal champions fighting for the pilot to succeed. 

The venture client model allows corporations to purchase solutions from startups at an early stage before the startup’s product is fully developed. As a result of these approaches, some startups are overcoming the pilot purgatory. It was encouraging to hear that each of those founders had at least one large insurer client fully deploying their solution across an enterprise, not just in innovation theatre talks. The mood in the room among the incumbents listening was hopeful. If these scaleups have found ways to break through the pilot barrier, it means the industry is learning how to adopt innovation faster.

The discussion then turned toward the technologies that will have the biggest impact by 2030. Franklin posed a rapid-fire question: “Which tech will reshape insurance the most in the next 5-10 years –agentic AI (autonomous AI agents), edge computing (IoT and on-device intelligence), quantum computing for risk modeling, or something else?” 

We even polled the audience on this (because why not leverage the collective wisdom in the room). The results weren’t too shocking: AI (especially advanced AI agents) got the top vote as the leading enabler by 2030, validating the general theme of the event. Edge computing was second, which makes sense because of the IoT explosion. Insurers anticipate needing to process data at the source (cars, homes, wearables) for speed and privacy. A smaller group bet on quantum computing (perhaps thinking of its potential to revolutionize risk analytics, though most agree that’s beyond 2030 for mainstream use).

Interestingly, about 10% of respondents chose “something else”, which in discussions people clarified could mean genomics, climate engineering impacts, or other currently niche tech that could suddenly intersect with insurance.

The panelists commented on these results: none of them are explicitly building quantum solutions today (too early) or edge hardware, but they all see their products as having to fit into a tech ecosystem that includes those things. For example, Yandy from DataHaven said their data platform is being designed to ingest and analyze streaming IoT data (edge) in real-time, and Megan at Federato mentioned they’re watching quantum computing progress because one day it could make currently intractable risk calculations feasible.

Many startups are leveraging artificial neural networks—machine learning models inspired by the human brain—to tackle complex insurance tasks. An artificial neural network consists of interconnected layers of nodes that process data, enabling pattern recognition, image analysis, and natural language processing. These neural networks are essential for training foundation models and are widely used for tasks such as claims automation and risk assessment. Generative AI models, often built on neural network architectures, can create complex original content across various media, including text, images, and audio, in response to prompts. This means AI not only analyzes data but also creates new images, data, or responses based on its training and input. Reinforcement learning is another key machine learning technique discussed, where AI agents learn by trial and error, receiving feedback in the form of rewards or penalties to improve their performance, such as training a robotic hand to pick up objects or optimizing claims workflows.

When it comes to training AI models, startups use both supervised and unsupervised learning. Supervised learning involves training models with labeled datasets to enable accurate classification or prediction tasks, while unsupervised learning allows models to identify patterns in unlabeled data without explicit human guidance. Unlabeled data is a key resource for training large AI models, especially in unsupervised and semi-supervised learning, enabling AI to autonomously extract features and generate insights from vast, unstructured datasets.

The takeaway was convergence – that by 2030, the winners will be those who can orchestrate multiple technologies seamlessly (AI, IoT, cloud, etc.), not just focus on one in isolation.

One of my favorite parts of this panel was the conversation on trust and transparency. In the earlier incumbent panel, we talked about ethics from a big-company standpoint; here the growth ventures discussed it from an innovator standpoint. Sri from Charlee.ai gave a frank example:

"Years ago, our AI once flagged a legitimate claim as fraud because of a bias in the training data. We only caught it because a human claims expert reviewed the AI’s decision and said, ‘This doesn’t smell right.’ That taught us to build in explanations for every flag the AI raises to ease the integration of that human-in-the-loop.” 

This kind of introspection was great to hear. 

The panel agreed that explainable AI is not just a regulatory box to tick, but a market differentiator. The startups that can say to an insurer “our AI’s decisions can be audited and understood” will win trust faster.

We referenced how California’s upcoming AI regulations (taking effect in 2026) and similar laws will likely require proof of fairness and transparency. I also mentioned the broader context: financial services in general have been hit with big privacy and AI-related fines, like Meta’s €1.2B GDPR fine. Startups are aware that if they want to work with enterprise clients, they need to anticipate tomorrow’s regulations today. The venture client model benefits corporates by providing quick access to new solutions while giving startups early revenue. So there was a sense that ethics and compliance can’t be an afterthought. It must be baked into the product from early on, even if startups typically like to “move fast and break things.” One panelist wryly noted:

“In insurance, if you break things, people lose trust and you’re done. So we actually move fast and fix things like fixing the inefficiencies and gaps in insurance, but carefully.” That earned a chuckle from the audience, many of whom deal daily with the cautious culture of insurance.

Finally, we tackled the ever-important topic of financial sustainability for these innovators. Franklin brought up a stat that raised eyebrows:

And only 15% of insurtech scaleups have positive unit economics within 5 years. Those numbers underscore why so many startups flame out.

So how will today’s scaleups make it to 2030? The executive founders answered candidly. They said the vanity metrics (like “we have 100 customers on PoCs”) don’t impress anymore. What matters is proving that using their product actually improves the insurer’s bottom line or growth (meaning value creation 1-0-1!) 

For example, Megan at Federato pointed out that one of their carrier clients saw a measurable improvement in loss ratio after using their AI recommendations in underwriting – that kind of metric (real $$ impact) is what will justify renewals and expansions. Sri at Charlee.ai mentioned her team's focus on customer retention and referenceability as key metrics: if their early customers stick with them and are willing to vouch to peers about the value, that’s golden. Essentially, the advice to fellow startups was: chase quality over quantity: better to have 10 deeply integrated, happy clients than 30 pilots that go nowhere.

From the investor side (tying back to Panel 1), that’s exactly what VCs want to see too. So there was a neat alignment in the narrative across the day: innovation is exciting, but disciplined innovation wins the long game.

As a closing flourish to Panel 3, Franklin did a rapid-fire round with the panelists, asking each to name one standard tech industry mantra that does NOT work in insurance. The answers were both insightful and fun:

• “Just ship it” (implying you can’t deploy half-baked solutions in insurance and iterate later – reliability is paramount),
• “Growth at all costs” (a no-go when you must mind solvency and regulations), and
• “One size fits all” (insurance is highly regional and segmented; a product that works in one market might flop in another due to different regulations, customer expectations, etc.).

This light-hearted end underscored an important point. Insurance may learn from Silicon Valley, but it isn’t Silicon Valley. Startups and insurers have to forge their own path, balancing innovation with the trust and stability that customers and regulators expect.

AI Talent and Skills Development

Building a rock-solid AI talent foundation isn't just nice to have—it's make-or-break for companies serious about leading innovation and crushing the competition. This starts with getting your people trained in the fundamentals that actually matter: machine learning, deep learning, natural language processing. And here's the truth nobody wants to admit: as AI technologies evolve at breakneck speed, your workforce skills better evolve just as fast, or you're already behind. Manish Shah, President, CPO at Majesco, explains this well:

Now, venture client units? These teams are absolutely unique, differentiating and game-changers in this whole process. But here's what separates the winners from the wannabes: their people need a killer blend of capabilities that most companies completely underestimate. We're talking about the ability to frame messy business problems so AI can actually solve them, evaluate and pick the right startups (not just the flashy ones), negotiate partnerships that don't suck, manage pilot projects that actually deliver, and navigate those nightmare corporate processes we all know and hate. When you foster these skills internally—and I mean really invest in them—you're not just adopting new technologies. You're building a sustainable competitive advantage that your competitors will be scrambling to copy.

But here's where it gets interesting: collaboration isn't just key—it's everything. Partnering with startups and academic institutions gives you direct access to the latest research, cutting-edge tools, and expertise you simply can't build in-house fast enough. The smartest organizations I've seen? They're going all-in on ongoing education programs, AI academies, and cross-functional training that actually prepares their teams to harness real innovation power. When you prioritize talent development like this—when you make it non-negotiable—you create something magical: a culture of continuous learning and experimentation. And that's not just nice corporate speak. That's the secret sauce for long-term growth in the age of AI.

What's Next?... From Insight to Action

After three panels packed with ideas, it fell to me to close the summit by tying everything together. Standing on the stage, I felt a strong sense of optimism in the room – but also a realization that the clock is ticking. We summarized the day’s key insights and issued a call to action for our executive attendees: 

• Embrace Provocation: I encouraged everyone to take these provocative questions back to their teams. Ask your organization, “What if we dared to do X?” – whether that’s automating a whole product line, partnering with an unlikely player, or reinventing your customer experience with AI. The goal is not to predict the future with certainty, but to stretch our thinking. As we saw, those willing to challenge the status quo, like the startups automating underwriting or the insurers overhauling their core systems, are already reaping rewards.
• Focus on Purpose and Trust: A recurring theme was that technology must serve a greater purpose. Whether it’s closing the protection gap, improving customer well-being through prevention, or acting ethically with AI, we have a responsibility to deploy tech in Service-of-People. Denise underscored this by revisiting the stat that 79% of execs feel an obligation to close the $1.8T protection gap. That’s huge. This means most leaders (80%) want to use innovation to make insurance more inclusive and fair. Our challenge is turning that intent into concrete projects. So we urged attendees: identify one societal or customer problem your company can help solve with AI (be it climate risk, health, inequality in coverage) and make it a priority. Solving real problems is not only noble, it’s likely to be profitable in the long run, because it opens up new markets and strengthens customer trust.
• Collaborate and Experiment: One takeaway I personally had is that no one can do this alone. Insurers will need to partner (commercially) with startups, with each other, and even with regulators to navigate the next decade. I made a point that came from one of the panels: consider forming or joining coalitions – for example, a balanced AI risk consortium (like ours!) to share data on fraud or a joint venture to test new insurance products for emerging risks (like autonomous vehicles or cyber-physical threats). And internally, create structures that let you experiment safely... Sandboxes, venture client programs, pilot budgets that aren’t tied to immediate ROI. The faster we can test new ideas, the faster we learn what works. As one audience member said during a break, “I’ve got pages of notes – now I need to go home and convince my board to let us try some of these things!” If even a handful of the 50 executives in the room go back and launch a bold experiment, I’d consider the summit a resounding success.
• Prepare Your People: Finally, we reminded everyone that technology change is ultimately human change. The tools we’ve discussed – AI, IoT, analytics – are powerful, but it’s our people who will make or break their impact. Denise gave a great closing line:

“In the AI age, your competitive advantage is your talent. Train them, empower them, and they’ll drive your transformation.”

This resonated, especially with the insurers in the room. We highlighted examples like companies that have created internal “AI academies” to upskill employees, or those that have revamped performance metrics to reward innovation efforts even if they fail (to combat the culture of fear of failure). By 2030, the most successful insurance organizations will likely be those that managed to blend new tech skills with deep insurance expertise across their workforce.

Wrapping up the event, I shared one more personal thought. Over the past few years, I’ve seen the insurance industry’s attitude towards technology shift dramatically. Early on, there was a lot of hype and fear – people worried about disruption or thought AI was just a shiny object. Now, coming out of this summit, I sense a much more mature outlook: a mix of urgency and agency.

• Urgency, because we know the world is changing fast (climate change, digital-native customers, you name it,) and we can’t afford to lag.
• Agency, because we also know we have the power to shape that change – through collaboration, smart investment, and a willingness to act. The horizons of 2030 we explored aren’t fixed; they’re ours to co-create.

As attendees filtered out to our networking happy hour (no better way to end a long summit than with a drink and informal chat!), I heard conversations continuing enthusiastically.

• One CEO said, “I’m fired up to start a pilot with that startup who presented.” 
• An investor mentioned, “I got three new ideas for where to focus our next fund.” 
• And an innovation lead from a carrier told me, “This gave me the courage to push our exec team harder on AI adoption – everyone’s doing it, we can’t wait.” 

These comments made me smile. They showed that provocations and possibilities, when combined, can inspire real action. 

In closing, AI Horizons 2030 accomplished what we set out to do: spark dialogue, provoke new thinking, and illuminate the possibilities ahead. The insurance industry is often called traditional – some even say resistant to change – but the energy in that room proved otherwise. We saw an industry imagining its future in real-time, wrestling with big questions, and collectively brainstorming solutions.

I left with a notebook full of insights (And I shared some of these with you above). More importantly, a renewed confidence that by 2030, insurance will not only have survived the age of AI and autonomy, but will be thriving in it – more resilient, more customer-centric, and more collaborative than ever. 

Stay tuned for the upcoming trio of articles that I would like to unpack in 2026 across the three stakeholder-groups. We’ll delve deeper into each perspective (investor, incumbent, and innovator) with tailored insights and guidance.

The journey to 2030 is just beginning, and we’re excited to continue exploring it with you. 

 Want to know more -> Contact us here. 

Sources

Fortune - MIT report: 95% of generative AI pilots at companies are failing

BCG - AI Adoption in 2024: 74% of Companies Struggle to Achieve and Scale Value

PitchBook - Duck Creek Technologies launches $890M bank-led loan to take out private credit

Root.ai - Insurance AI Trends & Highlights” (July 2025) – 77 % of U.S. insurers use AI, but scaling remains uneven

SAS - Data & AI Impact Report – The Trust Imperative – 46 % of organisations face a trust dilemma and trust-worthy AI drives 1.6× ROI

Digital Insurance article “Do insurers have an ethical obligation to close the global protection gap?"– $1.8 trillion protection gap; 79 % of executives see closing it as an ethical duty

Munich Re article “Balancing the promise and peril of AI in insurance” – 74 % of companies use AI, but 60 % of CEOs are hesitant to invest further

Insurance Thought Leadership article “AI Needs a Strong Foundation” – between 70–80 % of U.S. insurers have implemented generative AI; about 74 % still rely on legacy systems

FAQ: AI Horizons 2030 – Provocations and Possibilities

1. How often do AI pilots in companies fail?

Answer: According to an MIT/BCG report summarised by Fortune, only about 5 % of generative-AI pilot programmes achieve rapid revenue acceleration; the other 95 % stall or fail. This high failure rate highlights why insurers need clear use-cases, rigorous project governance and trusted partners when experimenting with AI.

2. How widespread is AI adoption in insurance today?

Answer: A Roots.ai survey found that 77 % of U.S. insurers now use AI in areas like claims and underwriting. While adoption is high, insurers still struggle to scale pilots enterprise-wide because of legacy systems and fragmented efforts.

3. What proportion of organisations experience an AI trust dilemma?

Answer: The SAS Data & AI Impact Report (the “Trust Imperative”) found that nearly half of organisations (46 %) worldwide fall into a “trust dilemma,” meaning there is misalignment between organisational trust in AI and the actual trustworthiness of their models. The report stresses investment in data governance, skilled talent and infrastructure to bridge this gap.

4. What is the global protection gap and do executives see closing it as an ethical duty?

Answer: Digital Insurance reports that the global protection gap—the difference between economic losses and insured losses—stood at about US$1.8 trillion. In the same SAS survey, 79 % of insurance executives said closing this gap is an ethical obligation, and 76 % viewed it as a significant business opportunity.

5. How many companies are already using AI and what holds CEOs back from further investment?

Answer: A survey referenced by Munich Re found that 74 % of companies had already deployed some form of AI, yet about 60 % of CEOs remain hesitant to invest further because of concerns over liability and unclear return on investment. This juxtaposition underscores why AI adoption is both widespread and cautious.

6. Why is investing in trustworthy AI critical for return on investment?

Answer: The SAS Trust Imperative report indicates that organisations that embed robust data governance and ethical AI practices are 1.6× more likely to achieve double-digit ROI on AI initiatives compared with those that do not. In other words, trustworthy AI isn’t just a compliance requirement—it drives better financial outcomes.

7. What percentage of insurers still rely on legacy systems while pursuing AI?

Answer: According to industry analysis cited by Insurance Thought Leadership, between 70 % and 80 % of U.S. insurers have implemented generative AI in at least one business function, yet about 74 % still depend on outdated legacy systems. This gap underscores the need to modernise core infrastructure before scaling AI initiatives (see “AI Needs a Strong Foundation” by Insurance Thought Leadership). 

8. What is the recommended innovation strategy—build, buy or partner?

Answer: At the AI Horizons 2030 summit, a live poll asked attendees whether insurers should build AI internally, buy AI solutions, partner with big tech, or pursue a hybrid strategy. The overwhelming response favoured a hybrid approach, reflecting recognition that no single strategy is sufficient for success in AI. Although this insight comes from the summit discussion rather than an external study, it echoes investor guidance that insurers should combine internal capability building with strategic partnerships and acquisitions.

9. Why are humans still critical in an AI-driven insurance future?

Answer: Panel discussions and industry research agree that even as underwriting and claims become heavily automated, human expertise remains vital for oversight, empathy and ethical judgment. AI can handle routine tasks, but complex or sensitive cases still require underwriters and claims adjusters to interpret context and ensure fairness. Industry executives emphasise that an AI-first world must still be human-centred to maintain trust and regulatory compliance. 

10. How does the talent gap affect AI adoption in insurance?

Answer: Many insurers face a shortage of AI-savvy talent even as their veteran workforce retires. This talent gap slows AI implementation and creates anxiety among employees about job security. To address it, leading organisations invest in upskilling programmes and promote a “hybrid workforce” where humans and AI agents work together. The AI Horizons 2030 summit underscored that ethical, governance-centred AI programmes and continuous learning are essential to overcome the talent barrier.