The Future of AI is Vertical

By Gaby Diamant, BridgeWise CEO

Earlier this week, there was a brief online and social media storm about how ChatGPT’s updated T&Cs now prohibited using the service to provide health, legal, or financial advice. That would indeed be a big change if it were the case, so I decided to check with ChatGPT itself.

In its response, it stated that the following language had been added to its terms, prohibiting the “provision of tailored advice that requires a license, such as legal or medical advice, without appropriate involvement by a licensed professional.” It also clarified that “the prohibition of giving tailored professional advice without a license has existed before.”

In an article covering the issue, The Verge reported that this was more of a clarification and unification of separate usage policies, and that the rules are the same as before.

The clarification from OpenAI is important and should be welcomed. While ChatGPT and other similar general-purpose AIs are remarkable tools with incredible capabilities, they absolutely should not be used for specific purposes: medical, legal, and financial advice chief among them. Essentially, if the advice provided by the AI comes with real implications and consequences for the life of the user, it’s not a use case for general AI.

Now, general AI chats are great for most types of queries, from general knowledge to more specialized ones such as recipes or travel advice. But again, the consequences of a less-than-perfect cake are dramatically different from those of poor medical or legal advice.

All that being said, it doesn’t mean that AI can’t provide solutions for those specialized queries. You just need the right AI. For some time, I have been a proponent of verticalized AI; that is, artificial intelligence that is highly specialized and targeted for a specific vertical. BridgeWise itself is a classic example of verticalized AI, with a highly focused and tailored solution for AI-driven investment advice.

The Four Pillars of Verticalized AI

Alongside my team at BridgeWise, we have identified four main pillars that underpin verticalized AI solutions. These come into play both in the process of identifying which subjects or industries require a verticalized AI solution and in how providers can think about addressing the needs of these markets.

Proprietary data

To serve a specialized vertical, you need specialized data. Having data that is not only targeted to a specific subject matter but is also validated is critical to verticalized AI. These spaces can’t be served by simply building a massive trove of data; rather, the question is more about quality and focus, ensuring that the data can provide both meaningful and reliable results.

Proprietary technology

Here again, there is a need to target the technology itself for the specific vertical. In the case of AI, this means developing targeted algorithms that can understand the subject matter and carefully curated language models that can provide accurate responses. In the case of BridgeWise, we have invested tremendous resources in creating our own Micro Language Model (MLM). This model, alongside more general-purpose language models, is the backbone of our generated content.

Any provider looking to offer a vertical AI solution must develop a subject-native layer that can deliver transparent, explainable results for that subject.

Regulation and compliance

This is one of the most significant obstacles that will block the entry of general AI providers into vertical spaces. While general AIs may be able to provide some useful information, the regulatory challenges of operating in these industries provide an entirely different layer that extends beyond the simple question of developing a sufficient technological solution.

At BridgeWise, we first had to establish our company as a licensed financial advisor and then navigate the regulatory challenges of using AI to provide financial advice that complied with our status.

This requires not just technological capabilities but significant operational capabilities to build the necessary infrastructure to manage the regulatory challenges.

Reputation and trust

In the spaces where verticalized AI will shine, trust is almost always a critical component. In the examples of legal, medical, and financial advice, users will only act on the advice if they can trust that it is accurate and free of “hallucinations”, false results that are frequently provided by general AI solutions.

If you are able to achieve the three other pillars, you can deliver on the promise of trust that is so critical in these targeted verticals.

Value Is Key for Verticalized AI

Beyond the basic structure outlined above, there is one other quality that is crucial for AI targeting specific markets – value. This may seem obvious; every solution needs to deliver value. But in the case of these highly specialized markets, it’s not just a question of the value being present, but of it going beyond what can be attained by general AI.

And for anyone who has used ChatGPT or other similar solutions, it’s clear that they offer a ton of value. That means a targeted AI solution needs to deliver something that is truly unique and can’t be provided by a general solution.

For us, our core value stems from the buy/sell recommendations we can provide on the broadest possible swath of global equities. This capability depends on layers of other problems we have solved across the four pillars I described previously.

This will be the key question as verticalized AI solutions roll out across the market: How much value do they provide above the general AI giants, and can they help companies navigate the unique challenges of the myriad verticals that exist in today’s market?