Over 16,000 financial advisors at Morgan Stanley now use AI @ Morgan Stanley Assistant in their daily workflow, with 98% of advisor teams relying on the GPT-4-powered internal chatbot to search a research corpus that grew from 7,000 to 100,000 documents. The firm then shipped AI @ Morgan Stanley Debrief on the same blueprint — a Whisper + GPT-4 meeting-summary tool that turns Zoom recordings into CRM-integrated client notes and follow-up emails, compressing follow-ups that previously took days into hours.
"This technology makes you as smart as the smartest person in the organization. Each client is different, and AI helps us cater to each client's unique needs," said Jeff McMillan, Head of Firmwide AI at Morgan Stanley.
The rollout needed confidence that AI could meet strict accuracy and compliance standards in regulated advice. McMillan's team — with David Wu and Kaitlin Elliott — set three goals: faster information retrieval, summarization automation, and client-tailored insights, each gated behind custom eval datasets where advisors graded AI outputs against expert baselines before deployment.
The eval framework evolved across three phases: summarization evals where advisors graded GPT-4's condensation of research content for accuracy and coherence; translation evals for multilingual clients plus joint work with OpenAI engineers to fine-tune retrieval as the document library expanded; and a daily regression suite in production, catching weaknesses before they reached advisors. "We went from being able to answer 7,000 questions to a place where we can now effectively answer any question from a corpus of 100,000 documents," Wu said. McMillan added: "Now, advisors can engage clients on topics they haven't discussed before because the friction between knowledge and communication has gone to zero."
The production stack combines GPT-4 for generation, Whisper for Debrief's speech-to-text, and a retrieval-augmented generation pipeline tuned jointly with OpenAI on Morgan Stanley's proprietary research corpus. OpenAI's zero data retention policy addressed one of the firm's earliest compliance concerns — "One of the first questions we get is, is our information going to be used by OpenAI to train the public ChatGPT? The OpenAI team's willingness to ensure zero data retention has been really impactful," Wu noted. For Debrief, separate eval datasets represented different meeting types, and every advisor reviews and adjusts AI output before sending, preserving human oversight.
Document access across the advisor base jumped from 20% to 80%, freeing advisors for higher-value client work. "The feedback from advisors has been overwhelmingly positive. They're more engaged with clients, and follow-ups that used to take days now happen within hours," Elliott said.
The team is extending the eval-driven blueprint beyond wealth management — Assistant functionality is already scaling to the institutional securities group, and Debrief is being adapted for investment bankers meeting CFOs. "We're building platforms that will support many other use cases," Wu said. McMillan framed the bigger picture: "It's a fundamental change which both improves the quality of our content and creates new products and services that only people who are close to the problem can imagine."