TL;DR
Two-thirds of financial AI pilots stall due to fragmented data and siloed workflows, not model capability. Dreamforce 2026 revealed the fix: use headless FSC, Agentforce, and zero-copy data to automate low-risk internal workflows (like meeting prep), establish strict compliance guardrails, and prove ROI before rolling out client-facing agents.
At Dreamforce 2026, Salesforce shared a statistic that should wake up every financial executive: half of Americans now turn to AI for financial advice, up from 10% a year ago. Among Gen Z, that number reaches 77%, just as an $84 trillion intergenerational wealth transfer begins shifting to them and millennials.
Salesforce framed its answer to this shift using Cumulus Bank, its keynote demo institution. In the scenario, a customer asks Anthropic’s Claude to help her consolidate her accounts. With the customer’s permission, the Cumulus agent joins the chat through a secure Model Context Protocol (MCP) connection. It conducts a needs analysis, presents a personalized offer, and books a banker meeting, all without the customer ever leaving Claude.
Behind the scenes, the agent simply executed Cumulus’s existing Salesforce needs-analysis and referral workflows. The futuristic client experience was built entirely on business logic the bank already owned.
Yet most institutions cannot run this play today. Salesforce reported that two out of three financial firms with AI stalled in pilot blame scattered data and siloed internal processes, not the AI model.
The path forward requires working from the inside out: untangle one operational workflow at a time and connect the data where it already lives, rather than funding a multi-year overhaul.
Dreamforce 2026 Highlights for Financial Services
Headless Financial Services Cloud and AIforce
Financial Services Cloud is now fully headless, decoupling customer data and workflows from standard Salesforce screens. AIforce carries those capabilities into any interface — from Claude and Slack to custom advisor desktops — while enforcing existing enterprise permissions.
At Rocket Pro, Rocket Mortgage’s B2B lending division, a Slack connection built in a two-day workshop went live in August; weeks later, account executives created live leads at an event simply by photographing business cards. This lets institutions deploy agents directly into existing employee and client workspaces, eliminating the friction of custom front-end builds.
Agentforce Operations
Agentforce Operations addresses the process side of stalled pilots. Teams upload a whiteboard sketch or SOP document, and Agentforce Operations maps the workflow, flags bottlenecks, and rebuilds the process with agents, automated tasks, and compliance checks. Risk exceptions and final approvals remain with human staff, while unified monitoring flags any KPI that slips. This allows firms to digitize complex operations using documentation they already own.
Data 360 and Informatica MDM
Data 360 and Informatica tackle data fragmentation. Data 360 reads data directly where it sits in Snowflake or Databricks without duplicating records into the CRM. Informatica MDM maintains a single source of truth for client identities, while MCP connections allow agents to query live market and transactional figures. When client records live across conflicting systems, zero-copy access grounds agents in real-time data without forcing a multi-year migration.
The Agentic Advisor Experience
Salesforce’s agentic advisor tools are available in customer orgs today. Run My Day prioritizes daily tasks across compliance, service, and retention, while Meeting Concierge automates meeting prep, transcript summaries, and follow-ups (launching on Microsoft Teams in October, Slack to follow). As advisors manage higher account loads for younger clients, these tools automate routine preparation without requiring custom development.
How to Move a Financial Services AI Pilot into Production
Ryan Radding, who leads Salesforce’s North American financial services business, structured the keynote around three questions:
- How do you show up in a client’s AI conversation trustworthily?
- How do you untangle processes within budget?
- What is the return?
Answering those questions in reverse yields the fastest results: untangle one internal process, earn compliance’s trust on it, prove the return, and then extend the capability to clients.
Untangle One Workflow Your Own People Run
Starting internally ensures that if an agent makes a mistake, the impact stays inside the firm. Client meeting prep is an ideal starting point for wealth firms, while banks and insurers can apply the same logic to onboarding or claims triage. Bank of America’s Journey of a Meeting workflow handles prep, summaries saved to Salesforce, and follow-ups for 35,000 bankers and advisors, saving up to four hours per meeting across a million annual meetings.
Using Meeting Concierge, the underlying agent framework already exists. Advisory operations leaders and system owners simply need to define three operational elements:
- Data Sources: Identify which backend platforms supply meeting data and which fields require manual re-keying.
- Data Conflicts: Establish which system serves as the authoritative source when platforms disagree on a client record.
- Division of Labor: Assign administrative steps (assembling packets, drafting follow-ups) to the agent, while keeping recommendations with the advisor.
Resolving these data sources and conflict rules dictates what to connect through Data 360 and Informatica, and that clean data foundation carries directly over to your next workflow.
Bring Compliance an Agent with a Defined Scope
Eran Agrios, Salesforce’s head of financial services product, noted on stage that financial advice “cannot be probably right.” Greg Jacobi, who leads Anthropic’s banking practice, added that a bank’s existing risk framework is what allows it to move quickly, provided audit trails are enforceable. Before an agent enters production, present compliance with explicit answers across four areas:
- Access Limits: Define which client records the agent can read and which remain off-limits.
- Action Boundaries: Specify which tasks the agent completes autonomously versus those requiring advisor sign-off.
- Accountability: Assign an individual employee to own the outcome of every agent action.
- Auditability: Ensure compliance can track every query, decision, and system permission used by the agent.
Approving this scope for a lower-risk internal workflow creates an established governance template for higher-stakes operations like underwriting or claims payouts.
Measure the Return Before You Scale
At Bank of America, every AI deployment begins with a documented ROI case and is tracked continuously post-launch. Apply that same discipline at pilot scale: measure baseline research hours and completed meetings per advisor for 30 days prior to launch, then track reallocated time afterward. Because agentic platforms use consumption pricing (such as Flex Credits and Anthropic token usage inside Claudeforce), project costs from historical meeting volume and audit them against usage reports after month one.
Bringing AI Agents to Your Clients
That seamless Cumulus Bank scenario is what’s possible when you fix your internal mechanics first. The work required to connect live data, establish explicit compliance boundaries, and automate meeting prep builds the exact operational engine needed to show up safely in a client’s AI conversation.
When autonomous agents take on the heavy lifting of background research and administrative paperwork, advisors gain back what technology was supposed to deliver all along: the time to guide clients through the complex, emotional, high-stakes decisions that still require human judgment.
For the broader perspective on Salesforce’s platform announcements, read Coastal’s Dreamforce 2026 Recap. If you are evaluating which operational workflow to move out of pilot first, Coastal’s AI Pathfinder is a zero-cost strategic session that audits your Salesforce architecture, evaluates candidate use cases, and delivers an actionable roadmap for your leadership, compliance, and finance teams.
