City skyline

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.

Advisor meeting with clients

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 workshop

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.

Advisor reviewing notes

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.

Team meeting

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.

Office towers

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.

Colleague coaching a colleague

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.

City skyline

Thank you very much!