TL;DR
Dreamforce 2026 proved AI agents are finally making a difference for life sciences and healthcare, just not in the exam room. The real payoff is happening in the back office, where bots are fixing billing errors and reducing call center hold times. Building an AI foundation that lasts means mapping manual workflows alongside staff, unify claims and clinical data before deploying, and secure every connection from day one.
At Dreamforce 2026, Salesforce played a clip of Parker Harris asking a question I keep coming back to: “Why should you ever log into Salesforce again?”
It’s a compelling idea, and a little scary. Salesforce and its customers have spent more than 20 years cutting clicks and refining page layouts. Now Salesforce is going headless. AIforce, introduced at the event, carries Salesforce data, workflows, and governance into tools like Slack and Claude, so you don’t have to open the CRM to use what’s in it.
For healthcare and life sciences (HLS), Dreamforce also brought proof that agents work in our industry. In past years, the HLS message was mostly about the roadmap. This time, Salesforce and its customers showed agents running in production, and the strongest results came from administrative work like billing and member service, not clinical care.
Those results make a strong case for where to start with agents, and for the data work that has to come first. Below, I’ll cover what Salesforce showed HLS this year, where agents are already paying off, what it takes to get those results, and what to do heading into 2027.
What Salesforce showed at Dreamforce 2026
Life sciences: Life Sciences Cloud and the CRM migration decision
What’s new
In its Life Sciences sessions, Salesforce reported that more than 140 organizations now run on Life Sciences Cloud, more than double the number at last year’s Dreamforce. Some went live in as little as five weeks, using partners’ quick-start packages that put a working baseline in place fast.
What it means
Those numbers arrive at a useful moment. The life sciences CRM that many pharma and medtech companies have long run on Salesforce is moving off the platform by the end of the decade, so most are already planning their next move.
Until recently, that choice came with a big unknown. Life Sciences Cloud was new, and moving to a different platform looked slower and riskier than following the current vendor’s own migration path. Dreamforce removed much of that unknown:
- The platform has momentum. With the customer base roughly doubling in a year, many of your peers have already made the choice.
- Getting started is faster than it looks. Quick-start packages let teams stand up a working baseline in weeks and build from there, instead of waiting years to see value.
Companies can now choose based on what each platform makes possible rather than which migration feels safer. For example, can it run commercial, medical, and patient services together, with agents working across all three?
If you have to re-platform anyway, use the move to rethink what your CRM does for the business.
Healthcare: Pre-built agents and data for payers and providers
What’s new
For payers and providers, the expensive part of any technology project has traditionally been custom work: building the workflows and connecting outside data. Agents inherit that problem. This year, Salesforce shipped more of that work ready to use:
- Payer and provider agents
- Pre-built workflows
- Pre-wired integrations with health data partners HealthEx, Verily, and Viz.ai
What it means
Take a payer that wants an agent to help its contact center answer member questions. Until recently, that meant designing the agent, building the workflow behind it, and connecting outside data sources. Salesforce’s Payer Contact Center now comes with assistive agents built in, so the project starts with the core pieces already in place. The team’s effort goes into connecting the payer’s own data and handling the cases specific to its members.
The better starting point is what Salesforce already ships, with custom work saved for what makes your organization different.
Salesforce demoed where AI agents are paying off in healthcare
Baptist Health South Florida: Recovering revenue from billing errors
The standout session for me was Baptist Health South Florida’s.
Health systems are absorbing lower Medicare and Medicaid payments and longer waits for reimbursement. Baptist Health used Agentforce to catch invoicing errors and recovered millions in payments it otherwise would have lost. When reimbursement is shrinking, collecting everything you’re owed is one of the fastest ways to protect margin, and it doesn’t require renegotiating a single contract.
CVS Caremark: Cutting call volume with voice agents
CVS Caremark brought a payer example. Autonomous voice agents in its contact centers cut call volume by 15%. For a pharmacy benefit manager, each of those calls is a member who got an answer without waiting on hold, and a representative freed up for the calls that need a person.
Why healthcare AI agents should start with administrative work
Neither win is clinical, and that was the pattern across the event: the clearest results came from administrative operations. There are good reasons to start there:
- The rules are already written down. Billing requirements and coverage policies give an agent clear criteria to check its work against.
- The results are easy to measure. Gains show up in numbers finance already tracks. No one has to argue about whether recovered payments count as ROI.
- The work builds the foundation for what’s next. To catch an invoicing error, an agent has to compare billing, claims, and clinical documentation. Connecting those sources builds the same patient record that clinical use cases will need later.
Clinical capabilities will keep expanding, and Salesforce should keep building them. Starting in the back office is how you fund the foundation.
How to get results from AI agents in healthcare
At Dreamforce, my conversations with healthcare leaders came back to two questions: how do you get your data ready for agents, and how do you make sure they support your staff rather than replace them? Some shared that their organizations had moved too quickly, deploying agents before their data was ready or changing hiring plans on the expectation that agents would take over the work, and the results didn’t always follow.
Getting results from agents comes down to three things.
1. Start with the staff who do the work today
Your staff know things your systems don’t. A billing representative knows which payers reject which codes. An intake coordinator knows which referrals usually arrive incomplete and how to fill the gaps.
Before handing a task to an agent, watch how your team does it. Note every point where they:
- Check more than one system
- Make a phone call to fill in missing information
- Rely on what they remember about a payer, provider, or patient
Each of those is a place an agent will get stuck, because the information it needs is scattered across systems or not written down. Document that knowledge and map those processes first, and keep the same people involved to check the agent’s work once it’s live.
2. Unify claims and clinical data with Data 360 and Snowflake
Integrations and APIs move data between systems, but they don’t make it usable. Claims, clinical, and member data come from different sources in different formats, and agents need them cleaned up and combined. Two kinds of platforms handle different parts of that work:
- A lakehouse like Snowflake is where you can clean, join, and model large volumes of claims and clinical history.
- Salesforce’s Data 360 makes that data usable in real time by agents and by people working in Salesforce.
With zero-copy integration, Data 360 can use data in Snowflake without copying it. That matters in healthcare, where every copy of patient data is another copy to secure and audit. It also means agents work from the same records your analysts trust.
3. Secure every agent connection to patient data
Going headless changes how you secure patient data. When agents can act from Slack, Claude, or any other tool, there’s no single login screen to guard, and every connection becomes a new way into the systems that hold that data. Salesforce’s new MCP risk scores address this by checking each connection before it’s authorized and flagging threats like prompt injection and tool poisoning.
How healthcare and life sciences organizations should prepare for 2027
As Salesforce goes headless, the screens its customers have spent two decades refining matter less, and what sits behind them matters more. Agents rely on those processes, rules, and data wherever the work happens, so they need to be connected, trusted, and ready to act on.
That shift is coming either way, and 2027 plans are taking shape now. Four steps can help you get ahead of it:
- Build agents into the projects already on your 2027 roadmap. If a CRM migration or data platform project is planned, design it with agents in mind rather than adding them later.
- Pick one administrative workflow and map it with the people who do it today. Choose one where the gain is measurable, like recovered revenue or call volume, and keep those staff involved to check the agent’s work.
- Connect the data that workflow needs before you deploy. Bring the claims, clinical, and member data together so the agent works from records your team already trusts.
- Secure every connection from day one. Decide which tools can reach patient data before the first agent goes live.
Done this way, agents give clinicians, billing teams, and intake coordinators more time for patients and members.
For the bigger picture from Dreamforce 2026, read Coastal’s Dreamforce 2026 recap. And if you’re working out where agents could deliver the most value in your organization, Coastal’s AI Pathfinder can help you decide where to start.
