Why vertical AI beats general-purpose AI in Financial Services

The rise of the AI agent

AI agents are everywhere. You can hire an SDR agent from Artisan, a data analyst agent from Julius, a recruiter agent from Ribbon, and even a “do-everything” AI employee from Lindy. Just the other week, I came across Tracelight, another London-based startup, developing an agent for spreadsheet tasks.

We’re living in a new world, but for AI to truly be valuable in customer operations, it needs the proper context. It must understand the business model, regulations, and edge cases - and not just handle basic support questions from customers.

X post from Aaron Levie

That’s why, when Neal, Danai, and I started building at Gradient Labs, we quickly moved away from a general support AI agent and instead focused on financial services - a space we all knew very well from our time at Monzo, where we navigated the realities of fraud investigations, scam prevention, compliance reviews, and vulnerable customer handling.

Why general-purpose AI is not sufficient for financial services

When we speak to financial services companies, we hear the same frustrations about general-purpose AI tools: 

• They can’t be trusted unsupervised. Hallucinations, verbose and inaccurate answers, and missed compliance checks mean human review is required for most queries.

• They start from zero. You have to train them from scratch, even for basic support - slowing time to value.

• They don’t understand the business. Without domain expertise, even simpler queries can break down (e.g., “Why is my deposit late?”)

• Compliance is an afterthought. In regulated industries, a wrong answer isn’t just a bad experience - it’s a potential fine, brand damage, or most likely, both.

• Complex ops are out of reach. Tasks like dispute handling, document processing, fraud screening, or support for vulnerable customers are beyond their capabilities.

The result? They can handle only basic questions, e.g. “Where’s my order?” but not most customer operations use cases (which are also the hardest to automate). 

The case for vertical AI in financial services

Ycombinator recently discussed vertical AI agents on their Lightcone podcast, and Jason Lemkin also wrote about vertical AI on X. At Gradient Labs, we’re building an AI agent purpose-built for financial services. That means it:

• Understands financial services-specific use cases from the start. From KYC verification to payment disputes to scam investigations, AI agents can follow complex SOPs, use integrated APIs, and make decisions that align with both business rules and regulations

• Built in compliance. Finance-specific guardrails are built into the product, helping prevent hallucinations, detecting vulnerable customers, avoiding financial advice, and routing sensitive topics to human agents instantly.

• A team with domain experience. Our AI delivery team, with a background in financial services, understands the customers’ business and can ensure a successful implementation in weeks.

• Delivers results from day one. We resolve 40–60% of tickets on day one without custom training — rising to 80%+ as procedures are added.

• Minimal engineering lift. We integrate with support platforms like Zendesk, Intercom, Freshworks, and Salesforce in just a few clicks.

Real-world impact and future outlook

Our customers are already seeing transformative results with our AI agent: Zego’s AI agent is regarded as the best ‘human’ on the team. Sling Money has called it a “game-changer” for customer support, resolving complex queries that once needed specialist intervention. Plum achieved an 80% CSAT and a 52% resolution rate right out of the box.

The shift from general-purpose to vertical AI is already happening. Fintech, insurtech, and other regulated companies are realizing that automation rates and CSAT scores only climb when an AI is built for their exact workflows, compliance needs, and customer scenarios.

With our AI agent, we’re not just building a better bot; we’re building the AI engine for financial services customer operations. And we’re only getting started.

If you’re interested in seeing how this agent can work for your company, let’s chat!