AI & Consumer Finance
AI advisors, robo-advisors, AI credit and banking, and the agentic tools that touch consumers' money.
NatWest, ASB, Bank of America, Capital One, Commonwealth Bank of Australia and ING published joint principles on transparency, safety, privacy, choice and interoperability for AI agents that shop and pay. The principles are voluntary and written ahead of any regulator, so what they leave to the banks' own discretion is worth reading as closely as what they promise.
New York Life's Wealth Watch survey of 2,278 adults found 82% rated AI financial guidance as at least good, yet only 23% acted on it without checking elsewhere first. People are hedging against tools they cannot evaluate, which is a different thing from trusting them.
Saturn ran 121 money questions five times each across 18 AI models and found wrong or incomplete answers 57% of the time, with errors climbing sharply on complex tax and pension questions. The wrong answers arrived in the same confident tone as the right ones, and the free models most people use fared worst.
NPR reviews new research on chatbot money advice, including an MIT Sloan working paper finding the advice skews sensible on broad questions and gets shakier on complicated ones like job loss. About a quarter of Americans now ask chatbots personal finance questions.
Wealthy clients are running their advisors' recommendations through chatbots and pushing back on fees. The advice relationship is inverting: AI as the second opinion on the human, not the other way around.
A finance professor's framework for why AI money advice fails dangerously: the confident-but-wrong answer is the common failure in finance, because advice is a credence good you can't verify for years — and an engagement-tuned chatbot has an incentive to sound sure exactly when it should hand you to a human.
A 13-question letter to Chair Atkins asks who's liable when a third-party AI agent trades on a retail brokerage — the first congressional probe of the regulatory gap Robinhood's 'open to agents' move created. SEC response due July 31.
A Stanford GSB / MIT Sloan study finds LLMs nudge people toward textbook-sound money habits, but the advice is only as good as the prompt: low-financial-literacy users end up roughly $50K poorer by 60, and women get systematically more conservative advice even from identical prompts.
A study of 18,904 consumers finds fintech use is the strongest predictor of interest in AI financial advice, with interest declining steadily by generation and women less interested than men. Useful demographic texture under the 'a third of consumers now consult AI first' headlines.
A research review of AI advisory moving from portfolio rebalancing toward full personal-CFO territory, and where consumer-protection gaps open as it does.
Finds data-privacy risk and perceived complexity are the biggest drivers of consumer refusal to use robo-advisors, the same trust gap that recurs on every beat.
A practitioner survey: a third of consumers now consult AI before seeing a human advisor, yet 84% still flag data-privacy and accountability worries.
A protocol letting AI agents transact continuously with on-chain permissioning and multi-rail settlement across 30+ partners. The notable twist is an audit trail by design — accountability built into the rails rather than bolted on later.
Agents that can trade and buy on a user's behalf — the consumer-finance and agentic-commerce threads converging in one product.
A feature on how AI advisory tools are reshaping who gives financial advice and what that means for the consumers receiving it.
The full chronological feed lives at On My Radar.
US20260111964A1 · Google Patents ↗
The system constructs structured prompts from a company's historical stock data, financial fundamentals, and AI-compressed news summaries, then directs a large language model to produce analyst-style buy, hold, or sell ratings across 1-, 3-, 6-, 12-, and 18-month horizons. It ingests company identifiers, historical price returns, earnings and revenue figures, raw news articles filtered and summarized by a preprocessing LLM, and news-derived sentiment scores on a −5 to +5 scale. A chain-of-verification step asks the LLM to explain its reasoning and checks date calculations as a hallucination guard; the model can also be fine-tuned on ground-truth forward-return labels using LoRA. JPMorgan is patenting the infrastructure for AI-automated equity research — a system that could influence retail investor decisions through AI-generated ratings without the transparency or accountability requirements that apply to licensed human analysts.
In the news(2)
The full catalog lives at Patent Watch.