On May 15th, OpenAI launched a personal finance experience inside ChatGPT for Pro subscribers in the US. You can now connect your bank accounts through Plaid, see a spending dashboard, and ask ChatGPT questions grounded in your real financial data. It’s genuinely interesting — and about three years behind where dedicated AI financial platforms already are.

That’s not a dig. It’s just context.

ChatGPT scored 82.5 out of 100 on OpenAI’s internal personal finance benchmark, which they built themselves with 50 finance professionals. Not to mog them too hard or anything, but…Rockport’s AI Advisor scored 98.3% on the actual CFP® sample exam — the actual industry standard test for human financial planners, administered across 6,000 unique questions over 432 hours. The average human CFP® scores 79.5%. GPT-5 scores 93.8%. Rockport scores 98.3%, with variance held between 95–97% across multiple runs.

That gap isn’t a coincidence. It’s architectural.

What ChatGPT’s finance feature actually does

To be fair about what OpenAI shipped: it’s a meaningful step. Connect your accounts through Plaid, get a dashboard of spending and portfolio performance, and ask questions based on your actual transaction data. Over 12,000 institutions supported. Read-only by design. Data is removed within 30 days if you disconnect. That’s a solid foundation for a general-purpose chatbot adding financial awareness.

But OpenAI was also clear about what it isn’t: the feature explicitly is not a replacement for professional financial advice. It’s a conversational layer on top of your account data, currently limited to Pro subscribers at $100/month, US-only, with a phased rollout to Plus users still to come.

What it can do: summarize your spending, answer questions like “what did my vacation actually cost me,” and help you think through a rough savings plan. What it can’t do: run a Monte Carlo simulation on your retirement timeline, flag that your RSU grant structure doesn’t match what you described and ask if you meant stock options instead, or route your query through a specialized investment agent, a tax planning agent, and a compliance gateway before the answer reaches you.

Why a feature inside a chatbot isn’t the same thing as a purpose-built system

Rockport’s AI Advisor was built from scratch for one job: financial reasoning at advisor grade. The architecture reflects that, in ways, a general-purpose AI with a finance plugin cannot replicate.

The system is a multi-agent ensemble — a Core Router classifies every query and routes it to specialized agents. A Spending Agent handles budgeting and cash flow. An Investment Agent handles portfolio construction and rebalancing. A Planning Agent handles retirement, estate, and tax scenarios. Each runs in parallel or in sequence, depending on the query, and is coordinated by an orchestration layer that keeps context clean rather than dumping everything into a single prompt. The full technical breakdown is here if you want to get into the weeds.

The math problem is where this matters most. Large language models are fluent but arithmetically fragile — they can confidently miscalculate compound interest or portfolio allocation in ways that produce reasonable-sounding but wrong answers. Rockport pairs its LLM reasoning layer with deterministic computational engines that handle the underlying math with absolute precision. When it tells you that retiring at 60 has a 74% success rate under your current assumptions, that number came from a Monte Carlo simulation, not a language model estimation.

There’s also a compliance layer that most people don’t think about until they need it. Every output goes through 138 automated checks before it reaches you — validating numerical accuracy, suitability, disclosure compliance, and privacy standards. That’s not something a general-purpose chatbot bolts on later. It’s a system designed around the reality that financial advice in a regulated domain carries consequences that a hallucinated restaurant recommendation does not.

What this actually means for you

If you use ChatGPT for everything and want a rough picture of your spending without opening another app, the new finance feature is a reasonable convenience. OpenAI is good at building accessible onramps.

If you want an AI that knows whether your portfolio allocation matches your retirement timeline, can explain why your net worth moved this week in terms of specific market events affecting your specific holdings, remembers that you reclassified Uber as transportation three months ago and applies that consistently, and has already outperformed every major frontier model on the actual CFP® exam — that’s a different product category entirely.

Rockport wasn’t built to be a feature. It was built to be your financial advisor. Right now, you can try it for $1 for an entire year — which, given the comparison above, feels like the easier call.