AI agents can now prepare you for a client meeting in seconds. Here is how to make that preparation accurate, safe and genuinely useful, without losing the relationship that makes it matter.
Part of my series on AI across the deal lifecycle
Introduction: The Thirty Minutes Before Every Meeting
If you work with clients, you know the routine. Thirty minutes before a meeting, you are scrolling through old emails, digging for the last call notes, checking what your colleagues discussed with the same client, scanning the news, and trying to remember the name of the new finance director.
In many firms, the result of that scramble is a one-page briefing called a tear sheet: a quick snapshot of the client, the relationship and what to talk about. Today, AI agents connected to a Customer Relationship Management (CRM) system can build that page in seconds and even capture notes and follow-ups after the meeting.
This matters well beyond investment banking. Wealth advisors, commercial bankers, consultants, lawyers, account managers and sales teams all live or die by how well they know their clients. The opportunity is huge. So are the risks of getting it wrong in front of a client.
The thirty minutes before a meeting should be spent thinking about the client, not searching for them.
In this article, you will learn:
What a tear sheet is, and why meeting preparation matters so much
How AI agents build meeting preparation from CRM and market data
he real value, beyond saving time
Six risks, from information barriers to write-back errors
Eight design principles to make AI meeting preparation safe and trusted
How to roll it out so people actually use it
1. What a Tear Sheet Is, and Why It Matters
A tear sheet is a short summary someone can read in two minutes before walking into a room. A good one usually covers:
Relationship history: how long you have worked together, past engagements and who owns the relationship.
Recent interactions: the last meetings, calls and emails, and what was promised.
Open opportunities and issues: active projects, pending requests and unresolved complaints.
Client snapshot: key financials, recent news and major changes in the business.
People: who will be in the room, their roles and what they care about.
Talking points: what to raise, what to ask and what to avoid.
Done manually, this work is slow and inconsistent. Quality depends on who prepared it, and much of what senior people know about a client lives only in their heads.
A great tear sheet answers one question in two minutes: what does this client need from us today?
2. How AI Agents Build Meeting Preparation
CRM providers now offer agents built for this job. Salesforce, for example, introduced Agentforce for Financial Services in 2025with advisor and banker agents that generate meeting agendas, surface relevant client insights during meetings, and summarize discussion points and action items afterward.
A typical agentic workflow looks like this:
Retrieve from the CRM: contacts, activities, opportunities, notes and service cases for the client.
Enrich with outside data: financial results, news, filings and industry developments.
Highlight what changed: what is new since the last meeting, which is often the most valuable part.
Suggest talking points: questions to ask, opportunities to raise and sensitive topics to handle carefully.
Capture after the meeting: draft notes, action items and follow-up tasks, ready for the relationship owner to review before anything is saved.
Many of these tools work through a simple conversation. Instead of navigating dashboards, you ask: "What do I need to know before tomorrow’s meeting with this client?"
The most valuable line in any briefing is the one that says what changed since we last met.
3. The Real Value, Beyond Saving Time
Time savings are the obvious benefit, but not the most important one:
Consistency: every person walks in with the same quality of preparation, not just the most organized ones.
Institutional memory: what one colleague learned last month is available to the next person who meets the client.
Better follow-through: commitments made in meetings are captured and tracked instead of forgotten.
Faster onboarding: a new team member can understand an account in minutes rather than weeks.
More time for the human part: less time assembling facts means more time thinking about what the client actually needs.
Saving time is the obvious gain. Never forgetting a promise to a client is the real one.
4. Six Risks to Manage
4.1 Information Barriers and Need-to-Know
A CRM often holds information from many teams. In a bank, that may include advisory work, restructuring discussions and lending relationships with the same client. In a law firm or consultancy, it may include work for competing clients. An agent that summarizes "everything we know" can carry confidential information across walls that people are required to respect. The agent must only see what the person asking is entitled to see.
4.2 Garbage In, Confident Out
CRM data is often stale or incomplete: old contacts, missing notes, duplicate records. A human reading raw records notices the gaps. An agent turns them into a polished summary that looks complete. Poor data quality used to be an inconvenience. With AI, it becomes a front-line risk.
4.3 Wrong Facts in Front of a Client
Calling a client by the wrong title, referencing a deal that never happened, or quoting an outdated number does real damage to credibility. Every fact in a tear sheet should trace back to a source record or document, with a date.
4.4 Write-Back Errors
The most powerful feature, letting the agent update the CRM from meeting notes or a quick prompt, is also the riskiest. An error written into the CRM becomes the "truth" that every future tear sheet repeats. Sensitive details can also end up recorded where they should not be, such as Material Non-Public Information (MNPI) or personal information about a contact.
4.5 Privacy
Tear sheets are about people. Personal details beyond what the business relationship needs should stay out, and data protection rules on how personal information is collected, used and retained still apply when an agent does the collecting.
4.6 Knowing the Summary, Not the Client
As I discussed in "The Junior Analyst Problem," there is a risk that people become reliant on the tool and stop building their own understanding. A relationship manager who only knows what the tear sheet says will struggle the moment the conversation goes off script. A tear sheet can tell you about a client. Only a relationship tells you what they need.
A polished summary of bad data is still bad data, only more convincing.
5. Eight Design Principles for Trusted Meeting Preparation
Inherit the user’s entitlements. The agent sees only what the person asking is allowed to see, and information barriers are enforced as access rules, not reminders.
Cite every fact. Each item links to its source record or document, with the date it was last updated.
Show what is missing or stale. Flag old contact data, gaps in recent activity and conflicting records, instead of hiding them.
Separate facts from suggestions. Talking points and recommendations are clearly labeled as AI suggestions.
Approve every write-back. No CRM update is saved without review by the relationship owner, and every change is logged.
Filter sensitive content. Keep MNPI, unnecessary personal data and privileged material out of summaries and notes.
Measure accuracy. Regularly sample tear sheets for errors, and make it one click to report a wrong fact.
Fix the data. Treat CRM data quality as a prerequisite for AI, with clear ownership for keeping records current.
Trust is built one cited fact, one flagged gap and one approved update at a time.
6. Rolling It Out So People Actually Use It
The best tool fails if nobody trusts it. In my experience leading AI enablement and training programs, adoption comes from a few simple practices:
Start with one team and one use case, such as meeting preparation for a single coverage group, and expand from there.
Teach verification, not just prompting. Show people how to check a tear sheet against its sources in under a minute.
Find champions inside each team who can share real examples of time saved and mistakes caught.
Make feedback effortless, so every wrong fact becomes a fix rather than a reason to stop using the tool.
Measure both usage and quality, not just how many tear sheets were generated.
Adoption does not come from the launch email. It comes from people who watched the tool catch a mistake.
7. Beyond Banking: The Same Pattern Everywhere
Wealth management: preparing for client reviews, including life events and portfolio changes.
Commercial banking: reviewing a business client’s full relationship before a credit discussion.
Consulting and law: briefing partners on a client’s history across many engagements.
Sales and account management: preparing for renewals, upsells and escalations.
Healthcare and public services: summarizing a case history before a meeting, where privacy rules are especially strict.
Every profession that meets clients can walk in better prepared, as long as privacy walks in too.
Conclusion: Better Prepared, Still Personal
AI tear sheets can turn the thirty-minute scramble into two minutes of focused preparation and give every client the benefit of everything the firm knows. That is a genuine edge.
But the edge only holds if the facts are right, the walls are respected and the people using the tool still know their clients. Let AI prepare the page. Let people build the relationship.
Preparation can be automated. Caring about what the client needs cannot.
How does your team prepare for client meetings today, and what would you trust an AI agent to do first? I would love to hear your view in the comments.
Explore more: Start Here · Topics · Archive · About