By Rahul Sachdev

AI is great at automating complex tasks, and product compliance has historically been a highly complex, manual task. At first glance, this might make AI seem like the perfect solution to modern product compliance challenges. And it can be, when approached in the right way.

The problem, though, is that AI is not all that good at generating trust, and product compliance requires trust and efficiency in equal measure. AI systems that suffer from information gaps or hallucination risks just can’t generate the degree of trust that compliance teams require to operate with confidence.

This doesn’t mean, however, that AI has no role to play in product compliance. On the contrary, AI is becoming an increasingly critical resource for compliance teams striving to keep pace with the ever-increasing complexity of modern product regulations.

But to balance the speed and efficiency that AI unlocks with the need for trust, compliance teams must approach AI in the right way, leveraging capabilities not available from generic AI models. This is the critical differentiator separating businesses that leverage AI effectively for product compliance from those that adopt AI but fall short of reaching their goals.

AI’s essential role in modern product compliance

Product compliance—meaning the process of ensuring that products comply with the various regulations governing their manufacture, sales and operation—was historically a process that required tremendous manual effort. Compliance teams, composed of technical and legal experts, had to identify relevant regulatory requirements, then determine which product changes to make to remain in compliance.

That approach worked in a world where regulations were relatively few and unchanging. But it doesn’t scale well, and it has become increasingly impractical for businesses operating in modern product compliance environments.

Consider, for instance, that research by my company has found that, on average, about 4,500 new or updated product regulations appear each year across the ten industries we follow. Trying to keep track of relevant regulations, let alone interpret them, is just not feasible at this scale when product compliance teams rely on a fully manual approach.

Hence the critical role that AI now plays in modern product compliance. Through capabilities like assessing regulations and mapping requirements onto individual products, AI can substantially accelerate the product compliance process.

The potential shortcomings of AI-assisted product compliance

But again, this hardly means that AI—or at least, not generic AI models and chatbots like ChatGPT and Claude—offers a simple, drop-in solution that can solve product compliance teams’ woes overnight. Teams that attempt to outsource product compliance to generic AI models subject themselves to two key risks:

• Information gaps: Generic models often don’t have up-to-date access to all relevant regulations, limiting their ability to provide reliable compliance guidance. (Indeed, according to research conducted by my company, generic models cover only about 30%–40% of relevant regulatory knowledge for most industries.)
• Inconsistency: Generic AI systems are non-deterministic. This means that, by design, they give different responses each time, even when presented with the same information. That’s great if you’re using AI models to produce things like creative works, but it can be a real problem in a compliance context because it makes it more challenging to validate AI-generated results or ensure consistency in AI-guided compliance processes.

Put together, these limitations undercut trust. They make it difficult for compliance teams to place a high degree of faith in the compliance recommendations or guidance provided by generic AI systems.

Bringing AI up to speed with product compliance

Fortunately, it’s possible to apply AI effectively to product compliance, but doing so requires purpose-built AI systems capable of delivering the following key characteristics:

• Comprehensive regulatory coverage: AI-driven product compliance tools must have real-time access to all regulations that are relevant for the industries and jurisdictions in which a company operates.
• Deterministic responses: Rather than generating different results, the systems must be deterministic such that when presented with the same data, they will produce consistent outcomes.
• Expert validation: Expert assessment of AI-generated compliance insights should be built into the system. This is possible when compliance experts actively validate AI outputs, ensuring that they are sensible and relevant.
• Traceability: In a compliance context, insight into how decisions are made, factors like which data they are based on and which possibilities were evaluated, is critical. To this end, AI tools for product compliance must be traceable, meaning teams can map compliance inputs onto outputs.
• Human-in-the-loop actionability: AI tools can accelerate and augment compliance workflows, but they’re never a replacement for human expertise. Ultimately, humans must sign off on the recommendations generated by AI systems before implementation.

Building AI compliance solutions that meet these criteria is eminently possible, but it requires more than simply training a model and throwing product compliance questions at it. It takes a mix of models, each tailored to different aspects of the compliance process, combined with deterministic algorithms that inject consistency into non-deterministic AI workflows. Just as important is manual testing of the systems by compliance experts to ensure they actually do what they’re supposed to.

With this type of solution, compliance teams no longer have to choose between speed and trust. They can have them both, and in turn, they can achieve the scalability necessary to transform product compliance from a slow, manual operation into a source of valuable insight and guidance for the business.

About the author

Rahul Sachdev is the CEO of Adherent.