With summer wrapping up and the Winter ’27 release around the corner, it is a good time to look at what Summer ’26 provided. The features are live, the early quirks have surfaced, and you can see what teams are doing with them.

The updates span Agentforce, Marketing Cloud, Data 360, Service, Sales, Commerce, security and development tools. Most of the updates have the same theme: Salesforce is moving AI from something that answers questions to something that does the work.

1. Real-Time Offer Management

Marketers have been promising the right message to the right person at the right time for years now. The idea was never the hard part. Making it happen across systems was the challenge.

Real-Time Offer Management inside of Salesforce Marketing Cloud puts two pieces together. Global Promotions Management handles the rules: eligibility, rewards, limits, promotion logic. Offer Management handles the delivery: how those promotions show up in email, web, mobile and messaging.

It gets more interesting when you pair that with your customer data and Salesforce Personalization. Instead of only deciding which campaign or audience someone belongs in, Salesforce can help decide which eligible offer makes the most sense for that person right now. Agentforce can build the offers and content.

If you are already running Data 360, Marketing Cloud, Personalization, Commerce or Loyalty Management, then this is where your promotions start to feel like one program instead of several.

Worth noting, the strategy is still needed. Areas like eligibility rules, approvals, audience logic and the experience a customer takes still needs someone thinking hard about them.

2. Agentforce Vibes 2.0

Vibes is the most technical update on this list and probably the most interesting.

Instead of helping a developer write one piece of code at a time, Vibes is built to take a broader request, help plan the work, pull in Salesforce-specific context and then generate or update the code. Summer ’26 adds Plan Mode, live Lightning Web Component previews, Salesforce Skills and Rules, and support for Model Context Protocol (MCP).

The context part is what matters. Plenty of tools can generate code. Far fewer understand the org they are generating it for. Salesforce is betting that platform awareness is the difference between code you can use and code you must rewrite. Giving developers time back from documentation hunting, metadata digging and boilerplate development.

Related Content: From AI Hype to AI Systems: What I’m Seeing in Salesforce’s State of Marketing Report

AI-generated code still needs testing, security checks and a human who signs off before it goes to production.

3. Vibe Coding for Custom Lightning Types

This one is niche, but it is critical in how your Agentforce experiences feels to the user.

Custom Lightning Types let developers control how an agent displays structured information. Without them, a customer asking about an order gets a block of text with the order number, shipping status and delivery date buried somewhere inside. With them, they get a clean card.

Summer ’26 lets developers describe the experience they want in plain language and have Agentforce Vibes generate the underlying configuration and components.

Using Agentforce to supply the right answer is only half the process. Making it easy to read and act on is the other half.

4. Security Mesh

The more connected your Salesforce environment gets, the more your security team has to keep track of.

Security Mesh, part of Salesforce Security Center, pulls security signals from Salesforce and outside platforms like Okta and CrowdStrike into one view, then turns scattered alerts into risk scores.

A single login at an odd hour does not look like much on its own. That same login sitting next to a signal from your identity or endpoint tools can look very different. Security Mesh is built to connect those events instead of leaving them in separate tabs for someone to notice.

That matters more as agents get access to more data and permission to take more actions. Agentic governance and visibility must grow along with the technology capabilities.

Related Content: Salesforce is Tightening MCE Security

Where to Start

Start with a problem you already have. Where would real-time data improve the customer experience? What repetitive work is eating your team’s time? Are your data, security and governance practices ready for experiences that are more connected and more autonomous?

The real takeaway from Summer ’26 is that AI has stopped being a feature inside Salesforce and started being a layer across it. It shows up in how marketers make decisions, how developers build, how sellers and service teams work, how organizations connect data and how security teams watch all of it.

This is a good moment to stop evaluating AI tools one at a time and start thinking about how your people, processes, data and technology need to fit together as these capabilities keep maturing.

Winter ’27 is not far behind, and we will have a read on that one soon. In the meantime, if you want a second opinion on which of these matter for your setup, we are happy to talk it through.