Are You Paying for AI You're Not Using? A Braze AI Optimization Audit for Marketing Leaders

Are You Paying for AI You're Not Using? A Braze AI Optimization Audit for Marketing Leaders 

By Sue Murray

Introduction

If you're running Braze, you're already paying for a fully AI-enabled customer engagement platform. Intelligent send-time prediction, automated channel selection, generative copy assistance, churn and purchase prediction; it's all in the license most Braze enterprise customers already hold.

Here's the uncomfortable part: you might not be using all the AI functionality. Across the implementations we've reviewed, it's common to find a large majority of licensed AI capability sitting dormant. It was likely enabled during setup but never operationalized into how campaigns and journeys get built and run. That's not a technology gap. It's a budget gap. You negotiated for this capability, you're paying for it every renewal cycle, and it's not helping your engagement metrics or your team's productivity.

This isn't a recommendation to purchase additional technology. It's an opportunity to evaluate whether you're maximizing the value of the one already powering your customer engagement. It's an AI Utilization Audit you can run this week, on the platform you already own.

What's Included in Your Braze License Currently?

Before you can find the gap, you need to know exactly what you're not using. Here's the core AI capability set built into Braze:

Intelligent Timing : Instead of sending a campaign to your entire list at 10:00 AM because that's when someone scheduled it, Intelligent Timing predicts each individual user's optimal engagement window and sends it to them accordingly. One campaign, thousands of personalized send times. 

Intelligent Selection: Functions like automated, continuous A/B testing at the individual level. Instead of running a single test and picking a winner for everyone, it dynamically routes each user to whichever content variant is statistically performing best for users like them.

Intelligent Channel: Automatically selects the best-performing channel  (e.g., push, email, SMS, in-app ) for each user based on their historical response behavior, rather than forcing every user down the same channel path a marketer picked manually.

AI Item Recommendations: A recommendation engine for product or content suggestions, most used in retail and media use cases, that personalizes what gets shown rather than relying on static merchandising rules.

AI Copy Assistant: Generative content support built directly into campaign creation, intended to speed up first-draft copywriting inside the platform.

Predictive Churn:  Flags users who are statistically likely to disengage before they do, giving the Marketing Team a window to intervene rather than reacting after the fact.

Predictive Purchase: Identifies users with a high likelihood to convert, useful for targeting spend efficiently or for suppression logic so you're not wasting sends on users who were never going to buy. 

Every one of these AI functions in Braze exists to replace a manual decision (e.g., a send time, a channel choice, a content pick, a segment definition)  with a data-driven one. That's the entire value proposition. If your team is still making those decisions manually while you’re paying for the AI/automated version, you're paying twice: once for the license, once for the labor.

Why adoption of AI features stall?

The common assumption is that marketing teams simply don't know these features exist. In practice, that's rarely the real blocker. The actual reasons adoption stalls are more specific and more fixable:

• The AI feature got flipped on during implementation but never made it into a real workflow. Enabling a toggle during setup isn't the same as building it into how campaigns get planned and launched. Plenty of instances have Intelligent Channel technically "on" but the functionally remains unused.

  
  

• Nobody owns validating whether it's working. Without someone running Intelligent Selection against a manual control group periodically, teams have no evidence it's outperforming what they'd do by hand, so it quietly gets ignored in favor of familiar habits.

• Data readiness gaps. Predictive Churn and Predictive Purchase both need sufficient behavioral event history to produce statistically reliable predictions. If your event tracking is thin or inconsistent, the models underperform and teams conclude the feature "doesn't work" rather than diagnosing the lack of data problem underneath it.

• A trust gap with the black box. Marketers who've spent years manually picking send times and channels are often uneasy handing that decision to an algorithm they can't fully see inside. Without a governance framework that explains how the model works and what guardrails exist, teams default back to manual control.

• No review policy for generative content. AI Copy Assistant frequently sits disabled by policy, not by choice, because nobody defined who reviews AI-generated copy before it ships, so it's easier to leave it off than to figure out the process.

None of these are technology problems. They're operational and governance problems sitting on top of technology that already works.

You Can Run an Audit To See What AI Your Team Is Using 

Below is a 9-point audit that you can do, and it will help you score your level of Braze AI usage.  Once you know your level of usage, then you can take actions to start incorporating them, or increasing their usage, across your campaign operations.  To start, pull up your Braze dashboard and go through these 9 points with an open mindset. This is the same type of review Celerity run as part of their Platform Optimization engagement, but you can easily do the first pass yourself. 

Point 1. Is Intelligent Timing enabled account-wide, or only on a handful of campaigns someone remembered to turn it on for?

Why it matters: Intelligent Timing must be toggled on a per-campaign or per-Canvas-step basis in Braze as it's not a single global setting. That means "we turned it on" during implementation often meant "we turned it on for the three campaigns we were building that week," and every campaign built since then defaulted back to a manually chosen send time.

Where to check: Pull up a list of active campaigns and Canvases and check the delivery settings on each for whether Intelligent Timing is toggled. Compare that count against your total active campaign count.

Good answer/Higher Score: "It's the default setting in our campaign templates, and we spot-check quarterly."

Bad answer/Lower Score: "I think it's on for our welcome series" (and nobody's checked anything built since).

Point 2. Have you compared Intelligent Selection's performance against a manual control group in the last 90 days or are you assuming it's working? 

Why it matters: Intelligent Selection dynamically routes users to the best-performing variant, which means, by design, it hides its own control group from you unless you deliberately hold one out. Without a holdout, you have no way to prove it's beating what a human would have picked manually so you're trusting the algorithm on faith.

Where to check: Look for a deliberate holdout segment (often 5-10% of the audience) excluded from Intelligent Selection and receiving a fixed variant, with reporting that compares the two groups' performance. 

Good answer/Higher Score: "We run a 10% holdout and review lift quarterly & it's outperforming manual picks by X%." 

Bad answer/Lower Score: "We turned it on and results seem fine" (with no comparison baseline at all). 

Point 3. Does your team have visibility into which Canvases use Intelligent Channel versus hardcoded, manually chosen channel logic?

Why it matters: It's common for a Canvas to be built with hardcoded channel steps (send push, then if no open, send email), logic a marketer designed by hand sitting right alongside newer Canvases using Intelligent Channel's dynamic selection. Over time, nobody can tell you, which is which without opening every Canvas individually, so the org has no idea how much of its messaging is optimized versus manually guessed.

Where to check: An audit of Canvas step configurations, ideally documented in a running inventory (Canvas name, channel logic type, last reviewed date) rather than tribal knowledge.

Good answer/Higher Score: "We maintain a Canvas inventory doc (60% use Intelligent Channel), and we have a backlog to convert the rest."

Bad answer/Lower Score: "Not sure, as it depends on who built it and when."

Point 4. Are AI Item Recommendations being used to personalize customer experiences, or are customers still receiving static product and content recommendations?

Why it matters: Braze AI Item Recommendations are designed to automatically personalize the products, content, or offers each customer sees based on their individual behaviors, preferences, purchase history, and affinity patterns. They can eliminate missed opportunities to increase engagement, conversions, average order value, and customer lifetime value.

Where to check: Review your active email campaigns, Canvases, Content Cards, and in-app messages to determine where product or content recommendations are being displayed. Identify how many experiences are powered by Braze AI Item Recommendations versus static lists or manually selected content. Then compare the performance of AI-driven recommendations against static recommendations by measuring click-through rates, conversion rates, average order value, and revenue generated.

Good answer/Higher Score: "We've deployed AI Item Recommendations across our primary product recommendation journeys, regularly measure their business impact, and continuously optimize the recommendation strategy based on performance."

Bad answer/Lower Score: “Our merchandising team updates the recommended products every few weeks," or "I think we looked at AI recommendations during implementation, but I'm not sure whether we're using them today."

Point 5. Is there a governance policy for AI Copy Assistant output such as who reviews it, and what the approval path is? Or is it disabled by default because no one ever defined the rules?

Why it matters: This is as much a legal/brand risk question as a technical one. Generative copy tools can produce off-brand tone, factual inaccuracies, or claims that create compliance exposure (especially in regulated industries like finance or healthcare) if nothing reviews the output before it ships. Many teams disable the feature entirely rather than deal with defining a review process which means you're not capturing any of the productivity benefit either.

Where to check: Ask whether there's a documented approval workflow (who reviews AI-drafted copy, what the sign-off step looks like) versus the feature simply sitting toggled off.

Good answer/Higher Score: "AI drafts get flagged for brand/legal review before scheduling, documented in our SOP." 

Bad answer/Lower Score: "We turned it off because we weren't sure how to handle it."

Point 6. Has anyone assessed whether you have sufficient event and behavioral data volume for Predictive Churn to be statistically reliable, or is it running on thin data?

Why it matters: Predictive Churn needs a meaningful volume of historical behavioral events per user to produce a reliable prediction. Sparse event tracking (e.g., only tracking email opens and nothing else) starves the model of its power. A model running on thin data doesn't fail loudly; it just quietly produces low-confidence or misleading churn scores that teams act on without realizing the underlying data was never sufficient.

Where to check: Review your custom event tracking plan and event volume per user over time. Braze's own documentation outlines minimum data recommendations for predictive features which is worth reviewing against your actual event catalog.

Good answer/Higher Score: "We track 15+ custom behavioral events per user and validated model confidence during setup."

Bad answer/Lower Score: "We track opens and purchases, that's about it" (and nobody's checked if that's enough).

Point 7. Are you using Predictive Purchase to prioritize customers most likely to convert, or are all customers still receiving the same campaigns regardless of their likelihood to purchase?

Why it matters: Braze Predictive Purchase uses AI to identify customers who are most likely to make a purchase based on their behaviors and historical engagement patterns. Rather than treating every customer the same, marketing teams can use these insights to prioritize high-value audiences, tailor messaging by purchase propensity, suppress customers who are unlikely to convert, and allocate marketing resources more effectively. The result is more relevant customer experiences, improved conversion rates, and better return on marketing investment. 

Where to check: Review your audience segments and active campaigns to determine whether Predictive Purchase scores are being used for audience selection, personalization, suppression logic, or journey branching. Evaluate whether campaigns targeting high-propensity audiences are outperforming traditional segmentation approaches and whether predictive audiences are reviewed and optimized on a regular basis.

Good answer/Higher Score: "We actively use Predictive Purchase scores to prioritize campaign audiences, personalize customer journeys, optimize media spend, and regularly measure conversion lift against standard audience segments."

Bad answer/Lower Score: "We built the predictive model during implementation, but I don't think any of our current campaigns actually use it," or "We still send the same promotional campaigns to everyone."

Point 8. Do you have a recurring, scheduled review of AI feature performance, or was adoption a one-time decision made at launch and never revisited? 

Why it matters: AI features aren't "set once, done forever". As audience behavior shifts, data volume grows, and Braze itself regularly updates these models. A feature that was correctly evaluated and enabled at launch two years ago may be underperforming today simply because nobody's looked at it since. Without a recurring review cadence, feature adoption becomes a historical decision rather than an active one.

Where to check: Determine if there's a standing meeting, dashboard, or quarterly business review item specifically covering AI feature performance and not just campaign performance broadly.

Good answer/Higher Score: "It's a standing agenda item in our quarterly platform review."

Bad answer/Lower Score: "We set it up during implementation and haven't really revisited it."

Point 9. Can you name, right now, which AI features are included in your license but have never been turned on at all? 

Why it matters: This is the most basic litmus test in the whole checklist, and it's often the most revealing. If nobody on the team can immediately answer this without looking it up, it's a strong signal that AI utilization has never been treated as something to actively manage. You can't optimize what you can't inventory.

Where to check: Cross-reference your Braze contract/license terms against your actual enabled features in the dashboard in a simple side-by-side list.

Good answer/Higher Score: A team member can list them off, or points to a maintained feature-inventory doc.

Bad answer/Lower Score: A long pause, followed by "let me get back to you on that."

Score Your Answers

0-2 checked: You're leaving significant licensed value on the table. This is a common result,  and it means real budget is going unused every renewal cycle. 

3-5 checked: Partial adoption. Some features are live, but without consistent validation or governance, ROI is inconsistent and probably invisible to leadership.

6-7 checked: You're in the minority of Braze customers operationalizing what they're paying for. Worth documenting and sharing internally as this is a genuine competitive advantage most teams don't have.

Turning Insight into Action

Completing this audit is only the first step. The real opportunity lies in transforming AI capabilities into measurable business results. The challenge isn't enabling more features; it's operationalizing the ones they already own. That means embedding AI into everyday marketing workflows, validating its impact, ensuring the underlying data is ready, and establishing the governance and processes that enable marketing teams to adopt AI with confidence.

Organizations that realize the greatest value from Braze don't simply activate AI; they continuously optimize it. That's where Celerity's Platform Optimization Services make the difference. We partner with clients to conduct a comprehensive review of their Braze environment, evaluating:

• AI feature adoption and utilization 

• Campaign performance and deliverability 

• Customer journey & Canvas performance 

• Marketing operations and workflow efficiency 

• Customer data quality & AI readiness 

• Governance, compliance, and AI oversight 

• Audience and personalization strategies 

• Business outcomes and ROI 

The result is a prioritized roadmap that identifies where the greatest opportunities exist to improve customer engagement, increase marketing productivity, and maximize the value of your Braze investment.

The Bottom Line

Your organization has already invested in one of the industry's most powerful customer engagement platforms. The question isn't whether Braze has the capabilities. The question is whether your organization is fully leveraging them.

Don't measure the success of your Braze investment by the day it went live. Measure it by how much more value you're realizing from the platform one year later.