
Morgan Stanley
Integrating generative AI
Who is affected?
Five groups feel the change, three of them every day
Stakeholder
What changes?
Likely concerns
Influence

Financial advisors
Meeting prep and follow-up are AI-supported; they review every AI output
Loss of control, AI errors, personal liability
High

Support staff
Notes and follow-up drafts partly automated; teamwork with advisors shifts
Losing relevance, unclear future role
Medium

Unit managers
Must drive adoption while keeping service quality high
Quality dip, reputation risk during rollout
High

Compliance & IT
Govern recordings, data and approved AI use
Data privacy, record-keeping rules
High

Clients
Asked for consent before Debrief records a meeting
Feeling recorded, less personal advice
Medium
Force Field Analysis (Lewin)
Driving and restraining forces are almost balanced

Generative AI in daily advisory work
AI Assistant + Debrief
Driving forces
Total strength: 19
Restraining forces
Total strength: 18
Time saved on notes & follow-up (~30 min/meeting)
5
4
Distrust of AI output and fear of errors
Instant access to ~100,000 internal documents
4
4
Support staff fear losing their role
More time for clients and advice quality
4
3
Ingrained routines, no time to learn
Competitive pressure to innovate
3
3
Clients’ privacy concerns, consent refusals
Firm-wide leadership commitment to AI
3
4
Unclear liability and compliance rules
From resistance to measures
Every restraining force gets a concrete answer
We address the cause of each resistance, not just the symptom.
Resistance
Root cause
Change measure

Advisors don’t trust AI notes

Fear of errors and personal liability
Hands-on training with real error cases; rule: advisor checks before anything is sent

Support staff fear losing relevance

Unclear future role
Co-design new roles (AI quality check, complex client service) and development paths

“No time to learn”

High client workload
Protected learning time (e.g. 2 h/week) and a phased rollout instead of a big bang

Clients hesitate to consent

Privacy, feeling recorded
Transparent consent script; opting out has no disadvantage for the client

Liability and compliance unclear

No agreed rules for AI use
Joint AI usage guideline with Compliance: who checks, who signs off, what is stored
Preparing people for change
Five steps to unfreeze the advisory unit before rollout
1 · Unfreeze: prepare people (this part)
2 · Change: pilot & training
3 · Refreeze: anchor new routines
01.

Involve early
Workshops with advisors and support staff: what do they expect, where do they see risks?
02.

Take concerns seriously
Answer hard questions on errors, liability and roles, not just “AI makes you faster”.
03.

AI Champions
Open-minded advisors and support staff test first, then coach their peers.
04.

Clear rules
Human in the loop: the advisor reviews every AI output and stays accountable.
05.

Learning time
Protected time to train and practise, with a phased rollout.

Key message: involvement, clear rules and practical support turn resistance into acceptance. AI supports advisors’ expertise; it does not replace it.

A 12-week pilot
8 advisors + 4 support staff · mixed experience and attitudes
Weeks 1–2

Prepare
Choose participants
Clarify roles + access
Measure baselines
Weeks 3–4

Practise
Four workshops
Fictional client cases
Check practical skills
Weeks 5–8

Apply
Suitable meetings
Consent + human review
Support from champions
Weeks 9–10

Improve
Collect experience
Adjust the workflow
Retest difficult cases
Weeks 11–12

Review
Evaluate the pilot
Check rollout conditions
Expand or improve

Lewin’s Moving phase: learn, apply and improve.
Our proposed local pilot — not Morgan Stanley’s documented rollout.

Training and support
4 practical workshops · 90 minutes each

Use the tools
Find information
Check sources
Draft meeting summaries

Check the output
Spot incorrect figures
Find missing details
Catch invented commitments

Consent + errors
Explain tool use
Follow consent steps
Report problems

Work together
Practise the handover
Allocate follow-up tasks
Approve client messages
2 champions
One advisor + one support employee
Selected from the 12 pilot participants
144 planned learning hours
12 × (6h training + 6h protected practice)
Extra capacity for preparation, IT and champions
Black/Gregersen: Resources (slide 185) and Champions (slide 188). Our proposal.

Clear ownership
before expansion
01.

Advisors review notes and approve client messages.
Support checks completeness and tracks tasks.
Champions: everyday questions.
IT: technical issues.
Compliance: permitted use and consent.
02.

Management provides time and decides on expansion with the required approvals.
03.

The first group keeps using the tools.
Further groups join gradually.

Expand only when skills, quality, consent and support meet the agreed conditions.
Otherwise: improve, retrain and retest.
Our proposed roles and rollout conditions. Communication and measurement follow next.
Psychology of change
Four types of people, four ways to talk
People react to change in different ways. The type of person is one of five factors (Cameron/Green 2012).

Our assumption: most advisors and support staff are realists. They work with facts, risk and clients every day.
So our general style: practical, concrete, step by step. Innovators become AI champions.
Thoughtful
Action-oriented

IS
Thoughtful Realist
“If it isn’t broke don’t fix it”
Give clear facts before the meeting. Show what stays the same and who checks the notes.

ES
Action-oriented Realist
“Let’s just do it”
Let them try Debrief in the workshop. Show the time saved after the first weeks.

IN
Thoughtful Innovator
“Let’s think ahead”
Explain the why: where AI at Morgan Stanley is going. Ask for their ideas.

EN
Action-oriented Innovator
“Let’s change it”
Make them AI champions. They test first and help their colleagues.
Types: Myers-Briggs (MBTI), reduced to four types (Cameron/Green 2012, script p. 69). Which types are common in the unit is our assumption.

Communication plan for the pilot
We talk face to face in the regular team meetings, not in an extra Zoom call

Before

Branch manager
The advisor checks and sends every client email. We talk openly about the new role of support staff.

Week 1

Pilot lead
How client consent, recording and responsibility work. Compliance and IT are in the meeting.

Weekly

Pilot lead
Honest update on what works and what does not. Then a short e-mail and one contact person for questions.

End

Branch manager
The results after 12 weeks and the rules for the next step.
Plan: Black/Gregersen (2008). Channel: media richness (Koch 2004). Times and owners are our assumptions.

Measuring success and making it last

Charting progress (Black/Gregersen 2008): a few clear numbers. For each one: how and how often we measure, the start value and the goal. Start values in weeks 1–2, goals are our assumptions.

Refreeze (Lewin): managers keep supporting the new way of working, we check the results again and reward good use.
01.

Measure
Time for follow-up work, use with client consent, mistakes in notes, feedback from staff and clients.
02.

Decide
In week 12 we compare results and goals. Then: more teams, more training or a better process.
03.

Make it last
One person is responsible, a review every three months, training for new staff, visible support from managers.

Recommendations and reflection

Our doubts: only the press release is public. Everything else about the unit is our assumption. Volunteers in a pilot are often more positive than others. Twelve weeks show first use, not a long-term habit.

Change is a cycle, not a straight line (Cameron/Green 2012). AI tools change all the time, so we keep checking.
01.

Pilot first
Start with the 12-week pilot. Add more teams only when the goals are reached.
02.

Advisor stays in charge
The advisor checks and sends every note. Support staff get a clear new role.
03.

Make it routine
One responsible person, regular training and regular checks.

Thank you very much!