Morgan Stanley’s user-led AI rollout shows how rethinking deployment can turn a promising technology into a widely adopted workplace tool. Rather than treating generative AI as a one-time software installation, the firm used a phased launch, advisor feedback, role-specific training, and internal champions to expand its AI Assistant to nearly 16,000 financial advisors; OpenAI reports that almost 98% of Morgan Stanley’s advisor teams now use the tool. The example matters because adoption depends not only on technical capability, but also on trust, relevance, workflow fit, and sustained change management.
Phased, User-Led Rollout: Morgan Stanley’s Adoption Strategy
A phased, user-led rollout is a deployment strategy in which an organization introduces a product to selected users, gathers evidence and feedback, improves the experience, and then expands access in controlled stages. In Morgan Stanley’s case, the attribute pairing is “Morgan Stanley—phased, user-led AI rollout”: the company combined executive sponsorship with direct participation from financial advisors who tested and refined an AI assistant for research and client-service work.
Everett Rogers, whose diffusion-of-innovations research remains a foundation of adoption theory, defines adoption as a decision to make full use of an innovation as the best available course of action. His model identifies innovators, early adopters, early majority, late majority, and laggards as distinct adoption groups. Morgan Stanley’s strategy reflected that progression: early users tested the assistant, experienced advisors helped validate it, and broader teams received a more credible and relevant product.
Pilot Deployment: Testing the Product Before Scaling
A pilot deployment is a limited release designed to test usability, risk controls, workflow value, and user confidence before a company commits to organization-wide adoption. Morgan Stanley initially introduced its AI Assistant to a smaller group of financial advisors and then used their feedback to improve search, retrieval, and response quality. This approach was particularly important in wealth management, where inaccurate or poorly contextualized information could affect client communications and regulatory obligations.
The pilot also transformed adoption from an abstract technology initiative into a practical question: could the tool help advisors find approved knowledge faster and prepare for client conversations? Morgan Stanley’s assistant was connected to the firm’s internal research and knowledge resources, allowing employees to ask questions in natural language instead of manually searching a large document library.
Role-Based Enablement: Connecting AI to Daily Work
Role-based enablement means adapting training, permissions, examples, and success measures to the specific responsibilities of each user group. This is a hyponym of user-led rollout because it narrows the deployment around concrete tasks rather than presenting a general-purpose tool with no clear purpose.
For financial advisors, the relevant use cases included locating investment research, summarizing internal material, and preparing for client discussions. These applications were more persuasive than generic demonstrations of chatbot capabilities because they addressed time-consuming activities already embedded in the workday. McKinsey’s research on generative AI estimates that such technology could add between $2.6 trillion and $4.4 trillion in annual economic value across analyzed business functions, but the value depends on integrating AI into specific workflows rather than merely purchasing licenses.
Internal Champions: Turning Employees Into Evidence
An internal champion is a respected employee who experiments with a new capability, demonstrates useful applications, answers peer questions, and communicates practical feedback to the implementation team. Champions are a key hyponym of social adoption: they reduce uncertainty by showing colleagues how the product works in familiar professional contexts.
Morgan Stanley’s advisors played this role by helping the firm understand which prompts, documents, and workflows produced useful results. Their participation strengthened trust because employees could see examples from peers rather than relying only on messages from technology leaders. The result was a feedback loop: advisors supplied real-world needs, the implementation team refined the system, and improved performance encouraged additional usage.
Why Rethinking the Rollout Increased Adoption
The important change was strategic rather than merely technical. A conventional rollout often treats adoption as a communications problem: announce the tool, provide access, publish training, and measure logins. Morgan Stanley’s approach treated adoption as a learning process in which the company had to earn confidence and demonstrate value repeatedly.
Trust and Risk Controls
Trust is the willingness to rely on an AI system while understanding its limitations, sources, and appropriate uses. In financial services, trust must include accuracy, confidentiality, explainability, and compliance. Morgan Stanley limited the assistant to approved internal content and positioned it as a support tool rather than an autonomous decision-maker. This distinction helped employees use AI for retrieval and preparation while keeping professional judgment with the advisor.
The National Institute of Standards and Technology’s AI Risk Management Framework emphasizes governance, measurement, and human oversight. Those principles reinforce why a controlled rollout can outperform an unrestricted launch: employees receive access within clear boundaries, while the organization can observe errors and improve safeguards before expanding use.
Measuring Adoption Beyond Logins
Adoption measurement is the assessment of whether users repeatedly apply a product to valuable tasks, not simply whether they have opened it. Useful indicators include active-user rates, repeat usage, completion of target workflows, time saved, quality ratings, support requests, and user confidence.
OpenAI’s Morgan Stanley customer account reports that almost 98% of advisor teams use the AI Assistant, a figure that indicates unusually broad penetration. However, the strongest interpretation is not that every interaction creates equal value. The metric should be paired with evidence about how often advisors use the assistant, which tasks it supports, and whether it improves preparation time or knowledge retrieval. Figure 1 could illustrate this funnel: pilot users, trained users, active users, repeat users, and users reporting measurable workflow value.
Feedback Loops and Continuous Improvement
A feedback loop is a repeating cycle in which user behavior and comments inform product changes, and those changes are then tested by users. This is different from a launch-and-leave model, in which the implementation team assumes that initial training is sufficient.
Morgan Stanley’s rollout benefited from this iterative structure because advisors could identify gaps in search results, terminology, and response usefulness. The organization could then improve the experience while maintaining governance. Microsoft’s 2024 Work Trend Index similarly found that 75% of knowledge workers were already using AI at work, while many had adopted tools independently. That finding suggests that formal programs must catch up with employee experimentation by offering secure, approved, and better-supported alternatives.
Lessons for Organizations Planning Their Own Rollout
Morgan Stanley’s example is transferable, but not every organization should copy its technology or timing. The durable lesson is to redesign adoption around user needs, measurable outcomes, and managed expansion.
- Start with a high-value workflow rather than a broad technology announcement.
- Select pilot users who represent different roles, skill levels, and risk considerations.
- Recruit credible internal champions and give them time to experiment.
- Provide role-specific examples, guardrails, and training instead of generic demonstrations.
- Measure repeated task usage and business outcomes alongside license activation.
- Use feedback to improve the product before widening the rollout.
- Make human accountability explicit, especially in regulated or high-impact work.
A practical rollout scorecard should therefore combine adoption and value measures. For example, an organization might track weekly active users, repeat use within target workflows, average time to complete a task, user-reported confidence, error rates, and the percentage of use cases meeting governance requirements. These measures show whether adoption is becoming a habit that improves work rather than a temporary response to an executive initiative.
Conclusion: Adoption Is a Change-Management Outcome
Morgan Stanley’s phased, user-led AI rollout demonstrates why adoption can accelerate when companies rethink implementation. Pilot deployment reduced risk, role-based enablement made the assistant relevant, internal champions built peer confidence, and continuous feedback improved the product. The reported use of the assistant by nearly 98% of advisor teams shows the potential of this model, while broader research from Rogers, McKinsey, Microsoft, and NIST explains why the model works.
The broader implication is that organizations should stop treating adoption as the final step after procurement. It is an ongoing capability built through experimentation, governance, training, measurement, and trust. Leaders planning an AI or software rollout should begin with a narrowly defined workflow, identify representative pilot users, establish clear safeguards, and publish evidence of improvement before scaling. Further reading on diffusion of innovations, responsible AI governance, and workplace technology adoption can help teams turn a rollout into a durable operating practice.
Sources: OpenAI, Morgan Stanley Uses AI to Help Financial Advisors Serve Clients, https://openai.com/index/morgan-stanley/; Morgan Stanley, Morgan Stanley Wealth Management Announces Firmwide Adoption of AI @ Morgan Stanley Assistant, https://www.morganstanley.com/press-releases/ai-at-morgan-stanley; Everett M. Rogers, Diffusion of Innovations, https://www.simonandschuster.com/books/Diffusion-of-Innovations/Everett-M-Rogers/9780743258234; McKinsey Global Institute, The Economic Potential of Generative AI: The Next Productivity Frontier, https://www.mckinsey.com/capabilities/mckinsey-digital/our-insights/the-economic-potential-of-generative-ai-the-next-productivity-frontier; National Institute of Standards and Technology, Artificial Intelligence Risk Management Framework, https://www.nist.gov/itl/ai-risk-management-framework; Microsoft, 2024 Work Trend Index Annual Report, https://www.microsoft.com/en-us/worklab/work-trend-index/2024
