The adoption of artificial intelligence is having a profound impact on many industries, including insurance. As it becomes more common, AI will also allow insurers to introduce new products, significantly reduce their risks and dramatically increase their services, efficiencies and returns.

The promise of AI has unsurprisingly prompted programmed aggregators and InsurTech startups to jump into the market. Legacy carriers that have been traditionally slow to adopt new technologies are taking notice.

The big question is whether legacy carriers can act quickly enough. According to our own research with Harvard Business Review, most companies are just starting their AI journeys. Surprisingly, only a few have cracked the code. The rest are vulnerable to market changes and disruption-minded challengers.

We found that 90 percent of businesses are interested in leveraging AI. That said, 60 percent are still in planning or pilot stages, and only 18 percent of those surveyed report the achievement of significant business outcomes. This gap between leaders and laggards provides a tremendous window for disrupters, such as Lemonade.

Why, if AI is so beneficial and so disruptive, are so many traditional companies so slow to unleash it?

Every company will have a unique AI journey. The highest hurdle is orchestrating AI with human capital. Simply put, many companies simply don’t know where to start or how to scale what they’ve started.

Nonetheless, the survey data is instructive. A majority of companies (57 percent) cite “change management” as a challenge, and about half (48 percent) report struggling to find talent to build AI or work alongside it. Fewer than one in five (18 percent) have transformed processes to integrate AI.

Despite this, leaders are emerging. AI-driven bots handle FNOL. AI-driven estimatics platforms can capture damage and estimate claims costs. On the casualty side, companies are dipping their toes into platforms that leverage AI-powered bill review, content extraction and summarization, and BI tools, while AI-driven telematics platforms can analyze driver behavior and make usage-based insurance possible.

Carriers have a significant advantage over InsurTechs because of their deep domain expertise and understanding of the regulatory framework. AI systems are built on data and keep improving as they get exposed to more and more data. For example, an AI system developed to assess vehicle damage can expedite claims processing. To develop such a system, the AI algorithms should be exposed to hundreds of thousands of damaged vehicle images along with associated claims costs—all of which is usually available with legacy carriers.

Those who can combine AI with a deep understanding of the industry can achieve significant outcomes—one such attempt is Blackboard by AIG—but for the most part getting to that point can be a challenge, which underscores the need for an AI-driven ecosystem and supply chain.

However, the hurdle to become a leader in AI is real. Transforming processes with AI is not as simple as saying, “Hey, Alexa, turn on the television.” AI systems are built on data. With Alexa, the data is valid and tested, which is critical to quality.

Though legacy companies have collected lot of data, the data may be incomplete, incompatible, corrupted or otherwise difficult to use. Creating additional challenges for insurers are the silos of data stored in various policy, claims and other systems.

Having the ability to parse and cleanse data creates arguably a bigger challenge than running and scaling AI. Modern AI algorithms can also result in “proxy discrimination.” For example, ZIP codes could become a proxy for race or ethnicity and thus penalize a subset of community without the insurance company even realizing it.

AI can be an imperfect solution when improperly deployed. Knowing the realities of bias and other data flaws, just 38 percent of industry leaders said they had concerns with bias in AI, and only 19 percent reported having “an effective governance program for AI.”

In a traditionally risk-averse industry, these figures are surprising. To chart a successful future, carriers must understand the inherent risks of bad data, as well as how to model good data. With the exodus of so many industry professionals, now is the time to leverage talent that can orchestrate domain talent with AI capability.

It is not a question of if AI will become critical to a carriers’ brand, revenue and profitability but when it will reach such scale that the laggards will never catch up. Many who don’t embrace this change will be replaced by nimble upstarts that can pit clean AI against legacy resources.

The good news is that the amount of companies still in the early stages and the relatively small amount of AI leaders present real opportunities for those struggling to start or scale—if they move quickly. As Steve Jobs said, “Innovation distinguishes between a leader and a follower.” Leaders and disrupters have figured this out; for laggards the clock is ticking.