When Braze announced that its AI decisioning product works with Klaviyo and Salesforce Marketing Cloud, it was tempting to read it as a reason to skip the migration conversation. For a team that runs Klaviyo well, that reading is partly right. You can test AI decisioning without moving a single flow.

What it doesn’t do is make the migration question go away. It changes the order in which you answer it. If the decisioning layer earns its place, the next job for a lifecycle leader is to work out what it would take to bring the program onto Braze, and whether that is worth doing. Our view is that these are two separate decisions, and that treating them as one is how a low-risk test turns into a high-risk project.

The compatibility itself is documented. Braze’s documentation lists Klaviyo and Salesforce Marketing Cloud alongside Braze as compatible platforms for Decisioning Studio Go, and its orchestration guidance describes both connections for Pro as well. The details of those connections are where the useful decisions sit. 

What Decisioning Studio is, and what changed this month

Decisioning Studio started life as OfferFit, an AI decisioning company Braze agreed to acquire in March 2025 at an announced price of $325 million. The deal closed in June 2025, and Braze introduced the product under its current name at Forge 2025.

The core difference from A/B testing is that the agent isn’t looking for one winner. It uses reinforcement learning to choose, customer by customer, the action most likely to produce the outcome you set, and it keeps adjusting as results come in.

There are now two tiers, and they are different enough to treat as different products. Pro, launched a year ago, is the full-service version: Braze’s data science team helps design the agent. It can optimize for business metrics such as revenue and conversions, and it can draw on customer data from any source.

Go is the self-serve version, announced at Braze’s Forge conference in late September. It is in beta, with general availability expected in October. At launch, it covers email only and optimizes for clicks, choosing among the subject lines, calls to action, and hero images you supply and picking send days and times within your schedule. Braze’s setup flow requires a random control group of at least 5%, which gives you a baseline to measure lift against.

That difference matters when you read the proof points. The headline result in Braze’s materials is Kayo Sports’ 105% increase in cross-sells to its sister service BINGE. It came from a Pro build running ten reinforcement learning models across email, SMS, push, and in-app messages, and Braze reports it. It shows what the full approach can do at scale, not what a first Go agent optimizing email clicks will deliver.

How the Klaviyo and SFMC connections actually work

The detail most summaries skip is that Decisioning Studio doesn’t send anything. According to Braze’s documentation, it triggers API events into a flow in Klaviyo or a journey in Salesforce Marketing Cloud, passing the data needed to populate the message, and your engagement platform does the sending. Anything beyond the supported platforms may need custom engineering, because Decisioning Studio can’t trigger communications on its own.

On Klaviyo, the Go setup runs through a flow. You create a flow using the name and trigger event from the Go portal, add a trigger split and an email node for each template the agent can choose from, and set the email to live. Braze also tells you to turn off the setting that skips recently emailed profiles. It is a small checkbox, but it touches send frequency, so whoever owns your frequency rules should be part of the setup.

On Salesforce Marketing Cloud, setup needs an app package, a query automation that exports subscriber and engagement data once a day, and a Journey Builder journey that receives the API events.

Two things follow. Your current platform stays responsible for consent, suppression, and the send itself, so the compliance and deliverability work you have already done stays where it is. And Braze is candid that Decisioning Studio works best with Braze: its Pro setup documentation rates the native Braze integration as low complexity and the Klaviyo integration as medium. That is Braze’s own assessment, and worth remembering when a pilot starts to grow.

One caution. Go launched in beta only days ago, and the launch announcement describes the workflow around engagement data already in Braze, while the documentation lists all three platforms as compatible. Confirm with Braze which inputs, channels, and tier apply to your Klaviyo or SFMC setup before you plan around it.

What a Klaviyo to Braze migration involves

If a pilot works and the conversation turns to migration, the feature comparison turns out to be the easy part. In our view, the real risk sits in three places that rarely show up in a platform evaluation: consent state, customer identity, and sender reputation. Everything else is rebuild effort, which is substantial but at least visible.

Consent comes first. Braze tracks email subscription as opted in, subscribed or unsubscribed, and by default treats a profile with a valid email address as subscribed until someone changes it. Subscription is also tracked by channel and by subscription group. A list imported without its unsubscribe history, spam complaints and channel-level consent can mail people who already said no, so the mapping from Klaviyo’s consent and suppression data needs to be designed and signed off before any profile loads. SMS consent and sender registration run on their own track. In our Sailthru migration guide, we recommend a final suppression export from the old platform before the account is closed, so anyone who opts out mid-migration is carried across.

Identity comes second. Braze keys customers on an external ID, and the choice you make during migration becomes the identity everything else reads. Braze’s Shopify integration, for example, uses the Shopify customer ID, and its documentation warns that a plain email address carries a security risk. It matters beyond messaging: Braze’s data preparation guidance for Decisioning Studio is built around the external ID as the unit of customer identity, and says recommendation quality depends on the quality of identity resolution. A messy identity decision at migration is likely to show up later as a noisy signal for the agent.

The build is the part everyone budgets for. Klaviyo’s developer documentation refers to Django-style template logic, while Braze personalization is written in Liquid, so templates need porting, and automations need rebuilding, followed by QA against the live flows.

Sender reputation is the third risk. A new platform means a new sending setup: domains, authentication, and IP warming. Braze’s documentation calls IP warming the most important step in email setup, and its own guidance puts typical warming at two to six weeks depending on audience size and engagement, with automated warming starting from your most engaged customers. That makes the timing of a cutover a business decision as much as a technical one, particularly for a brand with a heavy fourth quarter. On our Weight Watchers engagement, an SFMC to Braze move, the team handled email template migration and paid close attention to domains and opt-in forms so that deliverability held through the transition.

Then comes the cutover. In our Sailthru guide, we suggest keeping the old platform active for at least 30 days after the new one is fully live, and the same logic applies here: the old platform is your fallback and your last source of suppression data.

When migration becomes the real question

Braze’s own documentation is direct that Decisioning Studio works best with Braze. Plugging into Klaviyo or SFMC makes the first step smaller, but it doesn’t promise the same experience at every step. So the useful question isn’t whether a pilot is possible. It is what you would need to see before a pilot turns into a migration plan.

We would look for three things. First, whether the agent beat its random control on a metric the business cares about. A Go agent optimizes for clicks, which is a proxy, and a lift in clicks that doesn’t carry through to revenue is a reason to keep testing, not to change platforms. Second, whether the next step is something your current setup can’t carry. Pro can optimize for revenue and draw on other data sources, and Braze documents it as working with Klaviyo and SFMC too, so the real comparison is the cost of that integration against the cost of a migration. Third, whether your team can run a rebuild, a consent mapping, and an IP warm-up alongside a live program.

Two things shouldn’t carry the argument alone. AI is the first: Klaviyo has AI of its own, and its product updates describe delivering messages at each recipient’s optimal time using reinforcement learning, so the difference lies in how decisions are made and what they optimize for. Channel count is the second. Klaviyo lists email, SMS, RCS, WhatsApp, and mobile push, and Braze lists mobile app and web messaging, SMS, RCS, WhatsApp, LINE, and KakaoTalk alongside email. If a channel drives the case for migration, name the specific experience you need and test it against what your current platform offers today.

We have said before that moving platforms on the strength of a single factor isn’t realistic; in that case, a pricing change, and that due diligence has to come first. A promising AI result is also a single factor. It is a good reason to look closely, and not on its own a reason to move.

Wrapping up

Decisioning Studio makes it cheaper to find out whether AI decisioning works for your program. It doesn’t make it cheaper to change platforms, and keeping those two costs apart is what lets you decide on evidence. A pilot can be stopped at any point. A migration, once the old platform is switched off, is much harder to unwind.

If you’re weighing a Decisioning Studio pilot on your current Klaviyo or SFMC setup, or sizing up what a Braze migration would involve, our team at Mavlers is happy to talk it through.