Braze Forge 2026 Wrap-Up: The Announcements That Matter and What to Do About Them

We’re back from Las Vegas, a little short on sleep and long on notes. Forge 2026 ran September 28 to 30 at the Fontainebleau under the theme “Made for This,” and the message from the main stage was clear. Last year Braze introduced its AI building blocks: BrazeAI Operator, Agent Console and Decisioning Studio. This year was about putting those pieces to work in the daily grind of running a program.

Four product launches, one big partnership and a night of awards. Here’s what was announced, where each piece stands today, and what we’d do about it if we were sitting in your seat.

1. BrazeAI Decisioning Studio Go: 1:1 decisioning without the build

This was the headline. Decisioning Studio Pro, launched a year ago, is powerful but custom. It’s typically configured with help from Braze’s data science team around a specific business goal. Go is the self-service version.

You give an agent an audience, a set of approved message variants and guardrails on how often and when it can reach people. The agent then experiments on its own to find the variant, time of day, day of week and frequency most likely to engage each person. Braze says a single Go agent can test more than one billion combinations at once, and that an agent can be deployed in minutes. Variants can include subject lines, creative and calls to action.

Braze’s own materials describe the first Go agents as email click optimizers initially, with more channels to follow. It’s in beta now, with general availability expected mid October.

Our take: Go removes the setup barrier, but it doesn’t remove the need for good inputs. An agent can only pick among the variants you give it, and it learns from the events you track. Before you turn one on, make sure you have three or four variants that are genuinely different (not the same email with a new button color), clean click and conversion tracking, and frequency rules your team agrees on. Teams that skip that step will get an agent that optimizes noise.

2. Agentic Standards: QA that keeps up with AI-speed production

Here’s the problem Braze is solving. AI can now build a campaign in minutes, then it sits for two days waiting for someone to check links, segments and unsubscribe footers by hand. CTO Jon Hyman noted that many teams still manage their QA requirements in spreadsheets.

Agentic Standards lets you set your rules once in Braze. It then reviews every campaign and Canvas against them, covering setup, copy, links, personalization logic, brand guidelines and compliance. Each check comes back as pass, warning or fail. Working with Operator, it can also make the fix, with human sign-off wherever you want it, and every check and correction is logged for audit.

Sportsbet, which operates in a heavily regulated space, said in the launch announcement that it has already caught missing unsubscribe links and eligibility segments in testing. Agentic Standards is in beta now, with general availability expected sometime in October.

Our take: This is the launch we expect regulated brands in financial services, healthcare and gaming to jump on first. The catch is that the tool is only as good as the rules you write. Most teams we work with have their standards scattered across a spreadsheet, a brand PDF and one senior marketer’s memory. Getting those into a single, written rule set is the real project, and it’s worth doing before you switch this on.

3. BrazeAI Operator Connect: Braze from inside Claude, ChatGPT and Copilot

Operator is the AI assistant inside the Braze dashboard. Operator Connect takes it outside. Using the Model Context Protocol (MCP), it brings Braze into tools like Claude, ChatGPT, Microsoft Copilot and Snowflake Cortex Code, plus other MCP-compatible tools.

In practice, a marketer drafting a campaign brief in Claude can ask Operator to look at recent email performance in their Braze workspace, plan the next campaign and build the structure inside Braze. The marketer then jumps into the dashboard for the final review. Developers can also connect their own agents and internal platforms to Braze through it.

One detail we like: every action is authenticated as the individual user and governed by that user’s existing Braze permissions. An agent can’t do anything the person couldn’t do themselves. Operator Connect is in beta now.

Our take: That permission model is smart, and it also means your Braze user roles just became a lot more important. If half your team has admin rights because it was easier at implementation, an AI agent working on their behalf has admin rights too. A quick roles and permissions review is a good first step before you roll this out widely.

4. Conversational Agents: two-way messaging that sells

Braze also introduced Conversational Agents, which hold real-time, context-aware conversations with customers across WhatsApp, SMS, RCS and web. They answer questions, clarify what a customer is after and recommend products, using brand guidelines, guardrails and knowledge sources like product catalogs and FAQs. They’re available in beta now.

Our take: The quality of these conversations will come down to the quality of the knowledge you feed them. If your product catalog has gaps or your FAQs haven’t been touched since 2023, fix that first. We’d also start narrow, with one channel and one use case, before going always-on.

5. A strategic collaboration agreement with AWS

Outside the product launches, Braze signed a strategic collaboration agreement with Amazon Web Services. The most practical piece is that Amazon Bedrock is now integrated into BrazeAI Agent Console, giving teams more choice in the models behind their Braze agents. The deal also covers industry-specific solutions, global infrastructure, and continued expansion in AWS Marketplace, where eligible customers can buy Braze using their existing AWS commitments.

Our take: If your company already has an AWS spend commitment, talk to procurement. Buying through AWS Marketplace can make a Braze expansion or renewal a lot easier to get approved.

6. The Torchies: proof that it works

The 2026 Torchie Awards on Monday night gave the announcements some real-world backing. A few results stood out to us:

• AB InBev won Campaign of the Year for a Zé Delivery program in Brazil that tied Brahma purchases to customers’ favorite football clubs. More than one million fans took part, and it lifted average spend per Brahma order by 30%.
• Wealthsimple was named Collaborative Team of the Year after marketing and engineering shipped its Monthly Millionaire experience in two months. More than a million people took part in the first month.
• Sarah Lands Ramrup of Nestlé Purina was named Marketing Leader of the Year. A program targeting a 15% conversion rate hit 24% in its first quarter, and Decisioning Studio cross-sell work produced roughly a 28% lift against holdout.
• Zahra Khan of Pizza Hut was named Rising Marketing Star, with the highest-scoring submission across the judged categories.

That Purina number is worth pausing on. AI decisioning has been a big promise for a few years now. A 28% lift against a holdout is the kind of result that gets budget approved.

Where everything stands

What we’d do in the next 30 days

Every one of these launches does more work for you, and every one of them depends on groundwork. Decisioning needs clean events and real variants. Agentic Standards needs written rules. Operator Connect needs tidy permissions. Conversational Agents need good knowledge sources. None of that is glamorous, and all of it decides whether these tools pay off.

If you want a starting point, we’d pick one journey where you’re still running manual A/B tests and get it ready for Decisioning Studio Go when it goes GA. At the same time, start writing down your QA rules. You’ll need them for Agentic Standards either way.

Thanks to everyone who stopped by to see us in the Engagement Expo. If we didn’t get time together in Vegas, we’d still love to talk. You can book a free Braze platform health session with our team, or take our 5-minute Marketing AI Readiness Assessment to see where your program stands across strategy, data, platform, content, people and governance. Or just reach out and tell us which of these launches you want to dig into first.