Audience: Business users (marketers, lifecycle leads, campaign managers)
By the end of this article, you'll understand:
- What AI Decisioning is and how it fits into your Hightouch workspace
- How it helps marketers scale personalization and continuously improve
- What tools and features you’ll use to build, launch, and optimize campaigns
- How AID balances automation with marketer-defined strategy and control
- How to get started with AID
What is AI Decisioning?
AI Decisioning (AID) is Hightouch’s customer-facing agent that helps marketers launch adaptive, goal-driven campaigns by automatically choosing the best message, channel, and timing for each user.
Instead of building rigid user journeys or manually testing campaign variants, you provide inputs like:
- Target audience
- Desired outcomes (e.g. purchases, sign-ups)
- Creative assets (email, SMS, push)
- Guardrails (frequency caps, timing windows, eligibility rules)
Then, AID learns from behavior and performance data to make real-time decisions for each user—learning and improving with every send.
You’re always in control. AID only operates within the strategy and content you define. It won’t create new messages or send outside your audience rules.
How it works
For every campaign/agent, AID uses reinforcement learning to:
- Ingests your inputs – Audience, goals, creative, guardrails
- Makes a decision – Which message, when, and through which channel
- Delivers the message – Via your connected ESP (e.g., Braze, SFMC, Iterable)
- Measures the outcome – Did the user take the desired action?
- Learns and adapts – Optimizes future sends for better performance
This cycle happens automatically for each user, enabling true 1:1 personalization at scale.
Why use AI Decisioning?
AID helps marketers achieve what used to be complex and time-consuming:
- Run continuous experiments without manual setup
- Personalize at scale without building complex logic
- Measure campaign lift without separate A/B testing infrastructure
- Optimize delivery based on real outcomes—not just opens or clicks
You define the strategy. AID figures out what works—and gets better with every send.
What kinds of campaigns is AID best for?
AI Decisioning is ideal for lifecycle campaigns where user behavior varies and experimentation is hard to manage manually. Examples include:
- Onboarding – Deliver the right nudge to help users activate
- Win-back – Re-engage lapsed or inactive users
- Cross-sell/Upsell – Recommend relevant products based on behavior
- Referral – Drive advocacy from high-LTV segments
- Retention – Reduce churn with personalized engagement
Each campaign is outcome-driven (e.g., purchases, referrals, sessions), and AID continuously learns which message, channel, and timing combinations perform best for each segment.
Key features of AI Decisioning
| Feature | What it does | When to use |
|---|---|---|
| Agents | Define goals, audiences, message options, and guardrails | Every campaign starts here |
| Actions | Add creative content (emails, SMS, etc.) and test variants | After creating an agent |
| Inspector | Preview past and upcoming actions for real users | After creating actions |
| Collections | Dynamically recommend personalized products or content | After creating actions |
| Insights | Analyze performance by user, message, or timing and optimize campaigns based on results | After agent launches |
| Configuration | Define workspace-wide defaults: access, time zones, etc. | Set up once per workspace |
Understand what’s working with AI Insights
AI Insights gives you visibility into how well your campaigns are performing—and why. As AID runs background experiments, Insights surfaces the most impactful trends across:
- Conversion breakdowns by segment (e.g. state, loyalty tier, device type)
- Creative performance by variant, channel, or tag
- Timing insights showing top-performing send windows
- Outcome summaries including lift, statistical significance, and winning group
You’ll be able to answer questions that go far beyond what traditional A/B testing could uncover, such as:
- Which products are driving the highest conversions among new users?
- What subject lines perform best with users aged 25–34?
- Are mobile users more responsive to push notifications or SMS?
This level of transparency helps you validate campaign performance, refine creative and targeting, and scale what’s working.
Get started
To launch a personalized campaign with AI Decisioning (AID), you’ll complete a few setup steps:
Technical setup
- Configure your workspace (one-time)
- Prepare your user and event data
- Connect your messaging destination
Build your campaigns
- Create your agent with audiences, actions, and outcomes
- (Optional) QA delivery with Inspector
- (Optional) Add collections to recommend personalized content
- Analyze performance in Insights
AID setup checklist
Use this table to understand who’s involved in each step of the process and link out to detailed instructions.
| Step | Article | Learn how to… | Who should complete this step |
|---|---|---|---|
| Step 1 | Configuration → | Set global eligibility, time zones, and channel defaults across your workspace | RevOps, Platform Admin, or Technical Owner (One-time setup) |
| Step 2 | Prepare data for AID → | Define your parent audience, structure event models, and prepare tables for AID | Data or Analytics Engineer |
| Step 3 | Set up destination | Connect your messaging platform (SFMC, Braze, Iterable) to deliver AID Actions | RevOps or Technical Admin |
| Step 4 | Agents → | Define campaign goals, audiences, and messages AID can send | Lifecycle Marketer, Campaign Manager |
| Optional | Inspector → | View and QA delivery behavior for individual users | Marketer or QA Analyst (Helpful during setup and testing) |
| Optional | Collections → | Dynamically recommend personalized products or content | Marketer or Merchandiser (Optional if campaign includes recommendations) |
| Step 5 | Insights → | Analyze campaign performance and optimize over time | Marketer, Lifecycle Lead, or Analyst |