Building agents

Learn how to build an agent for BrazeAI Decisioning Studio™, so you can automate personalized experimentation and optimize outcomes like conversions, retention, or revenue—without manual A/B testing.

About agents

An AI decisioning agent is a custom configuration for the BrazeAI™ decisioning engine that’s tailor-made to meet a specific business goal.

For example, you could build a repeat purchase agent to increase follow-up conversions after an initial sale. You define the audience and message in Braze, while your decisioning agents runs daily experiments and automatically tests different combinations of product offers, message timing, and frequency for each customer. Over time, BrazeAI™ learns what works best and orchestrates personalized sends through Braze to maximize repurchase rates.

To build a good agent, you’ll:

• Choose a success metric for BrazeAI™ to optimize for, such as revenue, conversions, or ARPU.
• Define which dimensions to test, such as offer, subject line, creative, channel, or send time.
• Select the options for each dimension, such as email versus SMS, or daily versus weekly frequency.

Sample agents

Here are some examples of agents that you can build with BrazeAI Decisioning Studio™. Your AI decisioning agents will learn from every customer interaction and apply those insights to the next day’s actions.

Agent use case	Business goal	Using typical methods	Using BrazeAI Decisioning Studio™
Cross-Sell or Upsell	Maximize average revenue per user (ARPU) from internet subscriptions.	Run annual campaigns offering every customer the next-highest tier plan.	Empirically discover the best message, sending time, discount, and plan to offer for each customer, learning which customers are susceptible to leapfrog offers and which customers require discounts or other incentives to upgrade.
Renewal & Retention	Secure contract renewals, maximizing both contract length and net present value (NPV).	A/B test manually, and offer significant discounts to secure renewals.	Use automated experimentation to find the best renewal offer for each customer, and identify customers who are less price sensitive and need less significant discounts to renew.
Repeat Purchase	Maximize purchase and repurchase rates.	All customers receive the same journey after making a website account (such as the same email sequence with the same cadence).	Automate experimentation to find the best menu item to offer each customer, as well as the most effective subject line, sending time, and frequency of communication.
Winback	Increase reactivation by encouraging past subscribers to resubscribe.	Sophisticated A/B testing and segmentation.	Leverage automated experimentation to test thousands of variables at once, discovering the best creative, message, channel and cadence for each individual.
Referral	Maximize new accounts opened through business credit card referrals from existing customers.	Fixed email sequence for all customers, with extensive A/B testing to determine the best sending times, cadence, etc. for the customer population.	Automate experimentation to determine ideal email, creative, sending time, and credit card to offer specific customers.
Lead Nurturing & Conversion	Drive incremental revenue and pay the right amount for each customer.	As privacy policies change at Facebook and other platforms, prior approaches to personalized paid ads become last effective.	Leverage robust first-party data to automatically experiment on customer segments, biding methodology, bid levels, and creative.
Loyalty & Engagement	Maximize purchases by new enrollees in a customer loyalty program.	Customers received a fixed sequence of emails in response to their actions. For example, all new enrollees in the loyalty program receive the same journey.	Experiment automatically with different email offers, sending times, and frequencies to maximize purchase and repurchase for each customer.

Building an agent

Prerequisites

Before you can build an agent, you’ll need to integrate BrazeAI Decisioning Studio™.

Step 1: Contact AI Expert Services

The AI Expert Services team will work closely with you to scope, design, and build your decisioning agent. If you haven’t already, reach out to get started.

You’ll complete the following steps together to build a custom agent that’s right for you.

Step 2: Design your agent

Alongside the AI Expert Services team, you’ll define:

• a target audience,
• the business metric to optimize,
• the actions for BrazeAI™ decisioning agent, and
• any first-party customer data the agent should leverage to drive your business outcomes.

With the design in hand, the team will work with you to identify and complete any additional integration requirements.

Step 3: Set up your delivery platform

Next, the AI Expert Service team will help you set up your marketing automation platform. While the decisioning studio works best with Braze, a variety of other platforms are supported—contact your AI Expert Service team for additional resources.

• braze

Step 4: Launch and monitor

After launching your agent, your AI Expert Services team will continue to monitor and tune it to your agreed-upon design. They’ll also help you make any adjustments, expansions, or modifications to the agent, if needed.