A Marketer’s Guide to Adopting Operator Everywhere, Operator Connect, Decisioning Studio Go, and Conversational Agents in Braze: Forge 2026 Product Announcements

One of the biggest takeaways from Braze Forge 2026: there’s a big gap between how fast tech and AI are moving and how much of it marketers can actually put to work. Most teams are running the same processes they always have, and it’s not because they don’t want to adopt AI, but rather because it becomes just another thing on the to-do list to try to figure out where to start. 

The reality is that marketers are moving fast with a lot on their plates, so there’s not a lot of time to step back and ask where AI could make their work easier. It can also be hard to tell where AI makes a real impact. When a team does spot an opportunity, they run into different roadblocks: navigating internal change and driving adoption of new processes and technology, working with different teams across the business, and trying to get leadership buy-in. Yet, marketing technology platforms — like Braze — continue to release increasingly advanced AI capabilities, despite marketers feeling like they’re behind in adopting even the most baseline of features.  

Four Braze product releases announced at Forge 2026 minimize that gap, making AI more accessible for marketers to adopt in their day-to-day work like never before:

• Operator Everywhere
• Operator Connect
• Decisioning Studio Go
• Conversational Agents

TL;DR: AI can now take on more of the hands-on work in Braze, like building Canvases, fixing Liquid, and choosing what each customer sees. Marketers still set the strategy and review everything before it goes live.

And while each of these capabilities lowers the barrier to adoption, they still don’t automatically change your process for you.  If a process is broken today, AI will run the broken version faster. We see this a lot with our clients: a team gets access to new technology before documenting how campaigns actually get built, and only then realizes how much of that knowledge sits with one person. Without a clear brief, the build you get back will be just as unclear.

For each release below, you’ll get what it is, who it’s for, what changes for your team, and what has to be true first so your team can actually adopt it. Disclaimer: Some are still in beta, so you may not have access yet — but each of these new capabilities will be hitting your Braze instance soon.

Where each new release fits in your campaign workflow

Most campaign processes move through the same nine stages. Each release shows up in different ones, so find where your team loses the most time before you pick a starting point.

• Build: create the audience, content, and Canvas logic in Braze.
• Localize: adapt content for different languages and regions.
• QA: check the logic, Liquid, and links. A person signs off.
• Prioritize: decide what to work on, test, or fix first.
• Engage: a campaign gets delivered, or a Canvas decides who gets what, when. At the end of both, a message gets delivered to a customer.
• Convert: turn a customer’s interest into a purchase, an upsell, or a cross-sell.
• Monitor: get insights into timely campaign stats and catch problems early.
• Diagnose and fix: find out why something is off, and repair it.
• Report: show what happened and what it was worth.

● = main use. ○ = supporting use.

Stage	Operator Everywhere	Operator Connect	Decisioning Studio Go	Conversational Agents
Build	●	●		○
Localize		○
QA	○	○
Prioritize		○
Engage			●	●
Convert			○	●
Monitor	○	●	○	○
Diagnose and Fix	●	●
Report	●	●	●	○

Getting started with Operator Everywhere: an assistant with hands on the keys

What it is: An AI assistant built into the Braze dashboard that can do the same work a Braze user would.. It builds Canvases, campaigns, emails, templates, segments, preference centers, customer events and attributes, and landing pages, writes and fixes Liquid, and pulls reports. It works inside the permissions of whoever’s logged in.

Who it’s for: If your Braze knowledge sits with one person (we see this a lot on our client teams), this multiplies their capacity and upskills the rest of your team without a new hire. You don’t need a high level of AI maturity to start using it. The important skill to have is knowing how to prompt it with clear instructions. You need a documented campaign process, a standard brief template, and one person who can review a build in Braze.

How it fits in your workflow: Operator Everywhere does its heaviest lifting in the build, diagnose and fix, and report stages of a campaign. It builds the Canvas from your brief, fixes broken Liquid, and pulls the reports and graphs your team used to wait on. It also supports QA, since it can find every live Canvas that uses a content block before you edit it, and helps monitor when someone needs a quick answer like which email domains are driving bounces. Review and publishing stay with your expert.

How it impacts your team: The biggest change is who can build. A lot of our client teams have one real Braze expert and a few lifecycle marketers waiting in their queue. A lifecycle marketer who’s been waiting on a welcome flow writes a standard brief covering audience, flow, tone, and goal. Operator finds or builds the audience, sets the timing, drafts the messages, and builds a working Canvas. The expert checks the segment logic, tests the flow, and publishes. Builds that used to take weeks can be ready for review in a couple of days.

Our team has been using it to generate graphs from Braze data, run reports, audit Canvases, fix broken Liquid, and use it for Query Builder when team members don’t know SQL. 

Cool things it can do: 

• Builds a working draft Canvas from a standard brief, then tells you what was unclear and what it assumed.
• Writes Liquid from a plain-language description, fallback included, and explains and fixes Liquid that’s broken.
• Works as a query builder for team members who don’t know SQL.
• Generates graphs from Braze data and runs reports.
• Audits your Canvases, like finding every live Canvas that uses a content block before you edit it.
• Answers questions live in the room, like which email domains are driving bounces over the last 30 days.

Helpful Operator Everywhere prompts our team uses — the more detailed, the better:

• Build a Canvas: “Build a welcome Canvas for users who created an account in the last 7 days and haven’t purchased. Email 1 welcomes them and highlights [value prop]. Wait 3 days. If they still haven’t purchased, send a follow-up with [incentive]. Exit anyone who purchases. Save as a draft. Don’t launch it.”
• Plan before building: “I want a win-back journey for customers who haven’t purchased in 90 days. Before you build anything, tell me the audience logic, timing, and exit conditions you’d use, and what you need from me.”
• Build from a brief: “Here’s our campaign brief [attach]. Build the Canvas as described. List anything that was unclear and what you assumed.”
• Write Liquid: “Write Liquid for a subject line that mentions the last product category the customer viewed. Fall back to ‘Something new for you.’ Tell me which attribute you used.”
• Fix Liquid: “This Liquid is throwing an error. Explain what’s wrong in plain language, then fix it: [paste Liquid].”
• Pull data without SQL: “Find users who performed [custom event] in the last [14] days where [property] equals [value]. How many are there?”
• Check Canvas performance: “How many users were sent step [2] of [Canvas name], and how many advanced to the next step?”
• Audit before you edit: “Which live Canvases or campaigns use the content block [name]? I need to update it and want to know what it touches.”

Before you start: Operator works off your existing process, so any gaps will show up quickly. Write down your campaign process before Operator builds on top of it. Create a standard brief template . since the quality of the build depends on the quality of the brief. Name who validates and publishes, since whoever publishes owns the output. Decide who can use Operator before rollout, because more people building means governance matters more.

The bottom line:  Your Braze expert spends less time building and more time reviewing, and your marketers can get campaigns live without waiting on them. That frees up time for the work that actually drives results: the audience, the message, and the offer.

Getting started with Operator Connect: pairing your favorite LLM with Braze so the brief powers the build

What it is: The same Operator capabilities, inside the AI tools your team already uses, like Claude, Cursor, or ChatGPT. It runs on the Braze MCP, a secure connection between where your team plans and where campaigns live. Your existing permissions carry over, and best of all: you no longer have to live in a sea of a million different tabs and platforms to get the answers you need.

Who it’s for: If your team is already using AI tools (LLMs, specifically) outside of Braze, with or without rules in place, this gives you a governed way to connect them into Braze.  Your team keeps using the tools they already like, with guardrails in place. But first,you need to decidewhich AI tools you will approve to connect to Braze, and who gets access. The biggest challenge here is learning how to clearly name the audience, timing, channel, and goal of a campaign to the LLM and make sure it understands how that connects to what lives in Braze

How it fits in your workflow: Operator Connect shows up most in build, monitor, diagnose and fix, and report stages of your campaigns. The brief carries your plan from the LLM into Braze as a draft Canvas, scheduled summaries and drop alerts keep an eye on live results, and you can ask where people fall off or have a Liquid error fixed without leaving the chat. It also supports QA and prioritization, since it can tell you which existing segments and attributes you could use and what to test first.

How it impacts your team: Our team has personally seen this cut customers’ time to market by up to 90%. What we often see is a lifecycle strategist spends a week in Claude working out a win-back plan. Audience, angles, and timing are settled. Then the handoffs start: copy it into a brief, have one person build the email, another person builds that Canvas, send it to the Braze expert, wait for a rebuild from scratch. The more steps, the more room for information to get lost in translation.

With Operator Connect, the strategist attaches the brief and asks for a draft Canvas. If a Liquid tag throws an error, they describe it and get it fixed in the same chat, all without leaving Claude. The draft then goes to the Braze expert, who checks the audience logic, reads the Liquid, and tests the flow before anything goes live. Operator can execute the build, and your team can QA and test.

Cool things it can do:

• Takes a campaign brief from your LLM and builds it as a draft Canvas, with no retyping.
• Reads your results and gives you a table of every step’s entered, proceeded, and exited counts.
• Pulls your top 5 and bottom 5 Canvases by conversion rate and tells you what stands out.
• Flags segments that look like duplicates, and custom attributes that look unused or inconsistently named.
• Fixes Liquid error in the same chat, and can wait for your OK before it applies the fix.
• Suggested Content Optimizer on its own once it read the brief, to make the messaging more targeted.

Helpful Operator Connect prompts our team uses — be specific, ask for drafts, and start small:

• Read-Only: “Pull this week’s reporting results for Canvas [name]. Give me a table of every step’s entered, proceeded, and exited counts. Don’t change `anything.”
• Plan with your Braze data: “I want to win back customers who haven’t opened the app in [60] days. Which existing segments and attributes could I use, and what’s missing?”
• Draft and build: “Here’s our campaign brief [attach]. Build this as a draft Canvas in the EMEA workspace. List anything that was unclear and what you assumed. Don’t launch it.”
• Fix and monitor: “This Liquid is throwing an error in [Canvas name, step]. Explain what’s wrong, then propose a fix. Don’t apply it until I say so.”
• Recurring reports: “Create a report that sends every Monday at 9 AM, it needs to be a  performance summary for [Canvas name]: entries, conversions, and any step with an unusual drop-off.”

Before you start: Connecting more tools to Braze can make a good process more efficient, and poor processes more open to risk.. A strong brief is what carries your plan from your LLM into Braze, so start with a standard template.  If your process isn’t documented, the brief won’t have much to work with. Decide which AI tools you approve and who gets access. Plan for human review before anything goes live, since Connect builds drafts and someone with Braze expertise should check them. Start with read-only prompts, then move to one simple draft like a welcome flow.

The bottom line: Your team is already planning campaigns in AI tools, and now that work can flow straight into Braze as a draft, with a person reviewing before anything goes live. Of the four releases, it’s gotten the least attention, but it’s likely to have the biggest impact on how fast your team ships.

Getting started with Decisioning Studio Go: 1:1 personalization without the data science team

What it is: Decisioning Studio Go (Go) hands the choice of variant, creative, send day and time, and frequency to an AI agent, picking from up to 100,000+ combinations per person and learning as results come in. It runs on native Braze data, optimizes for clicks and opens, and holds out a random control group that gets your usual experience, so you can measure the uplift yourself.

Who it’s for: If you’re a lean team that personalizes the old way, with two subject lines, an A/B test, two weeks of waiting, and a winner sent to everyone, this is for you. It fits evergreen journeys like lapsed shopper win-back, credit card activation, or upsell. And it’s fully self-serve, so you don’t need to get in line for data science or engineering to use it. You aren’t building the AI models or writing code, but you do need to be comfortable letting an agent choose what each customer sees, and reading uplift against a control group.

How it fits in your workflow: Go lives in engage and report stages of your campaigns. The agent decides what each customer gets and when, and the control group shows you the uplift. It also supports the monitoring stage, since it keeps learning as results come in. It doesn’t replace your build or QA stages. A human still writes the content options and approves every one of them.

How it impacts your team: You get more targeted campaigns and more testing without more headcount. You stop waiting on test results, stop depending on engineering to build decision logic, and stop hand-building personalization rules. , Your team’s work shifts to creating content variants that are meaningfully different, setting guardrails on offers and frequency, and reading results against the control group. For a team of two or three, that’s a better use of the week than running one test at a time. 

Cool things it can do:

• Chooses the variant, creative, send day and time, and frequency for each person, from up to 100,000+ combinations.
• Gives one customer a quick reminder on a Sunday morning, and another a benefit-focused message on a weekday at a slower cadence, with nobody building that logic by hand.
• Keeps learning from results, so performance improves over time.
• Splits your audience automatically into a decisioning group and a random control group, so you can see real uplift.
• Runs on native Braze data, with no data science team and no services engagement.

Before you start: Go removes the engineering gate, but it still needs good inputs. Start with one evergreen journey that has a real decision in it. Give the agent content that’s meaningfully different,because five near-identical subject lines won’t give it much to learn from. Set offer and frequency limits before launch, since the agent will pick any option you give it. Define what a win looks like up front, because Go optimizes for engagement, not revenue. If revenue or LTV is the goal, look at Decisioning Studio Pro. Then assign someone to compare the decisioning group against the control group and decide whether to expand, adjust, or stop.

The bottom line: Your team can run 1:1 personalization itself, with no ticket to data science. The built-in control group also gives you real numbers to bring to leadership when you’re ready to expand.

Conversational Agents: two-way messaging that knows who’s talking

What it is: An AI concierge that talks with customers in natural language and runs on your Braze data. Braze is positioning it for WhatsApp, SMS, and RCS, with a broader rollout later this year.

Who it’s for: If customers already text back or open a chat on your site, and your team can’t staff that conversation at 11 PM, this is for you. It fits moments where a customer has a question or a choice to make and the right answer depends on who they are: product recommendations, returns, onboarding, FAQs, and support.

How it fits in your workflow: Conversational Agents sit in the engage and convert stages of your campaigns. A customer reaches out, and the agent answers in the moment with recommendations and offers based on who they are. You’ll still do some build up front. Catalogs, workflows, and rules need to be in place before it goes live. It also supports the monitoring and reporting stages. A webhook step can send conversations to an auditing tool, so a person can review what the agent said and offered. Webhooks do more than that, too. The agent can open a ticket or update your CRM. If a customer reaches out after business hours, they can submit a ticket that triggers a follow-up once your team is back. 

How it impacts your team: Your work shifts from scripting chatbot flows to structuring data and writing clear instructions. We built one in beta, acting as a shoe-shopping assistant on a demo storefront. We set up two Catalogs — Shoes and Discounts — as the agent’s knowledge. We wrote one workflow with a description that told the agent when to use it. We pulled a tier attribute (Silver, Gold, or Diamond) off the customer’s profile to determine if they had available discounts, and added a webhook to Pipedream so we could audit the conversations. 

Cool things it can do:

• Recommends products in natural language, using the catalogs you point it at.
• Reads a real customer attribute, like a Silver, Gold, or Diamond tier, and answers with the offer that fits.
• Follows rules you write into the workflow, like only bringing up a discount when the shopper asks.
• Scopes each workflow to certain apps, subscription groups, and audiences, so a Diamond-tier customer can get a different experience than a first-time buyer.
• Makes webhook calls (POST, GET, PATCH, or PUT) to other systems or an auditing tool.
• Works beyond shopping. The same setup can handle support, returns, onboarding, and FAQs, in an app or on the web.

We’ve tested this in beta: A shopper asked for hiking shoes with a lot of cushion and got three recommendations. They said one pair looked great but a little pricey, and asked about sales. The agent checked their tier, saw Diamond, found the matching offer in the Discounts catalog, and brought the price down 25%.  It only offered the discount because the shopper asked, which was a rule we built into the workflow.

We wrote almost no chatbot script. The agent relied on the catalogs, the tier attribute, and the workflow description, so keeping that data clean is now part of campaign prep. One catch from the build: Braze auto-refreshes Catalog content every six hours, so a price or offer change can take that long to reach the agent without a manual refresh. 

Before you start: The agent runs on the data and rules you give it. Keep your Catalogs current, since it recommends what’s in them. Make sure profile attributes are accurate, because a wrong tier means a wrong offer. Write clear workflow descriptions, since they decide whether the agent picks the right job. Clean up consent and subscription groups before an agent starts talking. Plan for the questions it shouldn’t handle, and assign someone to review what it said and offered on a regular basis.  

We also set up a separate workflow for returns and exchanges, so the same chat can switch jobs based on what the customer asks for.

Pro tip: Start with one focused workflow, like product recommendations or returns, before expanding.

The bottom line: The work is mostly data and instructions you already manage in Braze, so you can start with one workflow instead of a big cross-team project.  A customer who starts a chat is already showing intent, and a fast, relevant answer gives you more chances to drive a purchase, cross-sell, or upsell within the rules you set.

TL;DR — These new capabilities can meaningfully change the way your work gets done. This guide is meant to help you get there. 

The gap between what AI can do and what marketing teams can actually use won’t close on its own. Each of these releases takes a different piece of work off your team’s plate, from building and reporting to personalization and real-time conversations. But none of them will fix a process that isn’t working today. If your campaign steps only live in one person’s head, document them before you turn anything on.

From there, start small with one welcome flow, one evergreen journey, or one chat workflow. Have the AI build drafts, keep a person reviewing, and use the results to decide where to expand. A small, measured win is also the easiest way to get leadership on board.

Not sure where to start? We’ll review your current campaign process and map out which release fits first. Reach out to schedule time to connect.

---

Meet the authors