Airship AI Products: Customer Guide
Version Date: Sep. 24, 2026
Executive summary
Airship leverages Artificial Intelligence (AI) to enhance marketing efficiency, increase conversions, and reduce time to value, all while adhering to evolving AI regulations and security standards. As a leading AI-powered cross-channel platform, Airship is committed to transparency and applies AI strategically without compromising data integrity and security. Our AI-powered solutions empower marketing teams to create highly personalized campaigns efficiently by providing intelligent starting points and automating complex tasks like Customer journey optimization and content creation. This document outlines our AI products, compliance framework, data protection practices, and responsible AI principles to demonstrate our commitment to ethical AI use and regulatory compliance. Additionally, we encourage Customers to review our AI Whitepaper.
Definitions
- “AI Functions” or “AI Functionality” means the artificial intelligence systems, technologies, and functionalities employed by or available through the Airship Service, including Agentic AI and Generative AI.
- “Airship Service” or “Service” means the Airship proprietary platform technology, including the Airship SDKs and APIs included in the Airship Library, and programs, features, functions, developer tools, report formats and any updates or upgrades of any of the foregoing made generally available by Airship. For purposes of this Policy, “Airship Service” excludes the Apptimize Service and ASO Services.
- “Customer” means Airship’s Customer using the Service.
- “Customer Data” means electronic data and content submitted by or for Airship’s Customer, or processed by, the Airship Service.
- “End User” shall mean any End User of one or more of Customer Digital Asset(s).
- “Generative AI” means artificial intelligence systems or technologies that can autonomously create or synthesize new content, including but not limited to text, images, and sounds, using machine learning and algorithms trained on input data.
Compliance framework
Airship adheres to industry data privacy standards and is committed to aligning with evolving global AI regulations, including applicable US federal and state laws, and international AI governance frameworks. Our commitment to responsible AI development ensures we maintain compliance with these regulations as they continue to develop.
As a technology provider that incorporates AI capabilities into our platform, Airship operates as a downstream AI Service provider whose Outputs may be used across multiple jurisdictions. Specifically, Airship:
- Is a provider of AI-enabled Services to Customers
- Is not a developer or provider of foundational AI models or general-purpose AI systems
Airship’s AI applications are designed with appropriate risk management controls and primarily support marketing automation and content generation use cases. Our AI systems are engineered to operate within established risk parameters and do not fall into categories typically subject to the most stringent regulatory oversight, often referred to as “high-risk AI systems.”
Airship implements transparency measures and risk management protocols that align with these requirements, including:
- Clear labeling of AI Functionalities in the platform
- Transparency in AI training data
- Human oversight of AI systems
Airship maintains a “human in the loop” approach to AI governance, ensuring human oversight in applying AI recommendations. This governance structure ensures that:
- AI-generated content can be reviewed and edited before sending
- AI delivery models are used only when explicitly instructed
- Predictive audiences are targeted only as requested
Airship utilizes Google’s Vertex AI platform, leveraging models from Gemini for content creation and Imagen for image generation. Airship has incorporated these models with Customer Data protection in mind: data is collected solely to improve Services with explicit consent, and Customers have full control over their data and the option to disable AI features. Customer Data and prompt inputs are not used to train any Generative AI models. Airship emphasizes a “human in the loop” approach, ensuring human oversight in applying AI recommendations. Trust, responsible application development, and ethical considerations guide Airship’s AI strategy, and Airship is committed to transparent practices and data security. Airship also adheres to industry data privacy standards and is committed to aligning with evolving global AI regulations.
AI product inventory
Listed below are the functionalities in the Airship platform that leverage AI with traditional machine learning or LLM models, as well as feature descriptions, the models used, data collected, and data retention. Note that the AI models and functionalities described below are subject to ongoing development and improvement. Airship may update, replace, or modify the underlying models and features at any time to reflect advancements in AI technology or to enhance performance and security.
| On-Device AI for Scenes | |
|---|---|
| Purpose/function | Decisioning |
| Description | Model evaluations let the SDK make contextual decisions based on prompts written by marketer: show/hold back a scene, show specific screen based on text response, select which piece of embedded content to display |
| AI products/models | iOS uses Apple Intelligence by default. iOS and Android also support bring-your-own-model (eg Claude, Gemini) |
| Customer data included in model training | No |
| Data inputs | Information provided in the UI by the customer, including; prompt information, user context (if provided), classification key, description (if provided), categories By default, no data is sent off the device unless customer configures it to do so using the evaluation observer. This data does not come to Airship. |
| Customer facing | Generally Available |
| Dashboard Agent | |
|---|---|
| Purpose/function | Analysis & Chat Assistant |
| Description | Assist users with analyzing their reports and offering recommendations on actions to take, as well as best practices and how to use Airship. |
| AI products/models | Google Gemini 3.8 Flash |
| Model type | Open |
| Customer data included in model training | No |
| Data inputs | Prompt information |
| Customer facing | Available in Early Access/Beta |
| Experiment Results Summary | |
|---|---|
| Purpose/function | Analysis |
| Description | Generates a plain-language summary of A/B test and Intelligent Rollout results; answers follow-up questions via interactive chat grounded in pre-calculated experiment result data. |
| AI products/models | Google Gemini 3.7 Flash |
| Model type | Open |
| Customer data included in model training | No |
| Data inputs | Pre-calculated experiment result data, Prompt information |
| Customer facing | Available in Early Access/Beta |
| Scenes Agent (Native Experience AI Agent) | |
|---|---|
| Purpose/function | Content generation |
| Description | Improvements to v1 agent content generation, create and edit a Scene using chat functionality. |
| AI products/models | Google Gemini 3.8 Flash and Gemini 3.1 Flash Image |
| Customer data included in model training | No |
| Data inputs | Prompt information, uploaded image, Brand Guidelines |
| Customer facing | Generally Available |
| Scene Accessibility Agent | |
|---|---|
| Purpose/function | Content generation |
| Description | Recommend fixes for flagged accessibility issues |
| AI products/models | Google Gemini 3.8 Flash |
| Customer data included in model training | No |
| Data inputs | Prompt; contents of the scene; results of our static analysis of the scene; (future) potentially any other scene-related; customer-entered information (fonts, colors, brand guidelines, etc.) |
| Customer facing | Generally Available |
| AI Email Template Generator | |
|---|---|
| Purpose/function | Content generation |
| Description | Generate an email template from an image or HTML prompt |
| AI products/models | Google Gemini 2.5 Flash |
| Customer data included in model training | No |
| Data inputs | Image, HTML, or prompt provided by the Customer at generate time |
| Customer facing | Available in Early Access/Beta |
| Recommendations AI → Campaign | |
|---|---|
| Purpose/function | Content generation |
| Description | Generate a draft campaign derived from the Audience Pulse recommendations |
| AI products/models | Google Gemini 3.5 Flash, Gemini 3.1 Flash Image |
| Customer data included in model training | No |
| Data inputs | Details from the Audience Pulse recommendations, campaign will prompt for more information information if there are changes requested from the recommendation. |
| Customer facing | Available in Early Access/Beta |
| Journey Map Conversational Agent | |
|---|---|
| Purpose/function | Content generation |
| Description | Conversationally create and manage a journey map and all of its connections. Connect to the journeys AI and AI scenes functionality to create content. |
| AI products/models | Google Gemini 3.5 Flash, Gemini 3.1 Flash Image |
| Customer data included in model training | No |
| Data inputs | Prompt information, Brand Guidelines |
| Customer facing | Available in Early Access/Beta |
| Campaign AI Assistant | |
|---|---|
| Purpose/function | Content organization and generation |
| Description | Upload a campaign brief or conversationally create a campaign with placeholder messages. |
| AI products/models | Google Gemini 3.5 Flash, Gemini 3.1 Flash Image |
| Customer data included in model training | No |
| Data inputs | Prompt information, uploaded campaign brief. |
| Customer facing | Generally Available |
| Recommendation Agent | |
|---|---|
| Purpose/function | Data analysis, recommendations |
| Description | Analyzes the past 12 weeks of RFM data from the Behavioral Targeting features and provides insights into the trends in movement between tiers, and provides recommendations to optimize audience value. |
| AI products/models | Google Gemini 3.5 Flash |
| Customer data included in model training | No |
| Data inputs | RFM measures (events, analysis window, aggregate numbers), Outputs of statistical models, company industry, company website url (if input) |
| Customer facing | Generally Available |
| Shippie | |
|---|---|
| Purpose/function | AI Chat Agent |
| Description | Shippie is designed to answer Customer questions about our product directly within the Documentation site, providing a more efficient and user-friendly support experience. |
| AI products/models | Google Gemini 3.1 Flash Lite |
| Customer data included in model training | No |
| Data inputs | Question/prompts entered |
| Customer facing | Publicly available on documentation site |
| Journeys AI | |
|---|---|
| Purpose/function | Content generation, Image generation |
| Description | Answer guided prompts to generate draft Journeys in the Airship Platform. Prompts include the purpose of the Journey, the information you want to convey in your messages, message types, trigger, tone, language, and images. After completing the form, the Journey Map displays all linked components. Your selected message types will be included in a Sequence and/or Scene. |
| AI products/models | Google Gemini 3.5 Flash, Gemini 3.1 Flash Image |
| Customer data included in model training | No |
| Data inputs | For content generation: prompt inputs, app name, vertical name, For image generation: prompt input |
| Customer facing | Generally Available |
| Brand Guidelines | |
|---|---|
| Purpose/function | Content generation |
| Description | Brand guidelines are used as an input to Journeys AI to ground generated content with Customer Brand and Style Guidelines. |
| AI products/models | Google Gemini 2.5 Flash |
| Customer data included in model training | No |
| Data inputs | Brand Guideline documents, sample writing samples provided |
| Customer facing | Generally Available |
| Generative Content | |
|---|---|
| Purpose/function | Content generation |
| Description | Use AI to create message content faster and according to guidelines you set. Enter seed text, then generate text to see the versions side by side. Customer can adjust the message tone, length, sentiment, and language. Customer can also include your brand name and emoji. |
| AI products/models | Google Gemini 2.5 Flash |
| Customer data included in model training | No |
| Data inputs | Prompt inputs |
| Customer facing | Generally Available |
| Send time optimization | |
|---|---|
| Purpose/function | Determine best send time for a user |
| Description | Optimal Send Time is an algorithm that determines the best hour for optimal engagement activity — when each individual member of the Customer’s audience is most likely to receive and act on your message. |
| AI products/models | Airship built machine learning model |
| Customer data included in model training | Any Customers with optimal send time enabled, uses only the customer's own data for the model. |
| Data inputs | Application lifecycle and send events, timezone information |
| Customer facing | Generally Available |
| Predictive Churn | |
|---|---|
| Purpose/function | Predict when someone is likely to become inactive |
| Description | Predictive churn analyzes your audience for users that exhibit behaviors indicating they are likely to become inactive, and tags the users as High, Medium, or Low risk. You can use this information for analysis & targeting users in the various churn buckets. |
| AI products/models | Airship built machine learning model |
| Customer data included in model training | Any Customers with predictive churn enabled, uses de-identified, aggregated Airship Customer Data to create a global model for all Customers |
| Data inputs | Application lifecycle and send events |
| Customer facing | Generally Available |
| Intelligent Rollouts | |
|---|---|
| Purpose/function | Automated variant optimization and rollout. |
| Description | Intelligent Rollouts uses a multi-armed bandit algorithm to automatically identify the best-performing message variant within an A/B experiment and progressively shift traffic toward it as statistical confidence builds. Instead of manually declaring a winner at the end of a test, the system continuously learns and reallocates sends in real time, so your audience receives the best-performing content faster. |
| AI products/models | Airship built machine learning model |
| Customer data included in model training | No, uses only the Customer’s own data for the model |
| Data inputs | Message variant direct engagement (click to open, clicks, message read) from within the Customer’s own experiment |
| Customer facing | Generally Available |
Data protection & privacy practices
Data collection and consent
Airship collects and uses Customer Data in the Airship platform solely to optimize marketing campaigns or otherwise provide requested Services to the Customer. We adhere strictly to consent frameworks, obtaining explicit Customer consent before collecting or processing any Customer Data.
Customer data protection guarantees
A core principle in Airship’s application of AI is Customer Data protection. Customer Data and prompt inputs are not used to train any Generative AI models. The data entered into prompts is used for the sole purpose of generating Output content for use in marketing campaigns or for reporting purposes.
Opt-out mechanisms
Customers have full control over their data usage with options to not use or otherwise disable the AI functionalities. Airship offers the Customer capabilities to determine the types of data to be used or processed with the generative AI functionalities in the Airship platform.
Intellectual property rights
Between Customer and Airship, Airship does not claim any ownership rights in any new intellectual property created within the Generated Output. To the extent that Generated Output is retained in the Airship platform, it is treated as Customer Data. Data retention for Generated Outputs adheres to Airship’s Data Retention Schedule.
Responsible AI principles
Human oversight mechanisms
While Airship’s AI functionalities provide valuable ideas and recommendations, we believe the best results come from a combination of AI and human oversight. Our AI products require a human to make the final decision on applying recommendations and capabilities. This balanced approach allows our Customers to benefit from AI’s transformative potential while maintaining control over their brand voice, Customer relationships, and strategic direction.
Transparency practices
Airship maintains open communication with Customers and end-users about using AI in our platform. We clearly explain and communicate when, where, and how AI is used in the Airship platform and how data is stored and protected throughout the process. This transparency promotes trust and empowers users to make informed decisions about their data and interactions with our AI-powered solutions.
Bias prevention measures
We support high standards of ethical AI practices, ensuring that our AI algorithms are fair and unbiased. We have added guardrails with our model provider settings to ensure the content safety rating is set to a level to avoid harmful results. Using both model capabilities and Airship configurations, we aim to prohibit harmful content and avoid or treat sensitive topics like politics, religion, or controversial issues with caution.
Security protocols
Protecting Customer Data is our top priority. Airship implements robust security measures to protect personal information within our AI systems. We employ encryption, access controls, and other industry-standard security practices to ensure that data is protected at all stages of collection, storage, and processing.
Customer data use in AI functions
Airship is committed to transparency regarding how Customer Data is used within our AI-powered features. Airship does not use Personal Data or End User data contained in the Customer Data or prompt inputs to train or fine-tune any Generative AI models. This section explains the two distinct ways Customer Data may be used to enhance AI functionality, and the safeguards in place for each.
Aggregate data for broader AI improvement
When Customers enable certain AI features, Airship may use de-identified and aggregated data to improve AI Functions across the broader Airship platform. This aggregate data includes:
- Vertical/industry classification (e.g., retail, travel, media)
- Content classification (e.g., promotional, transactional, informational)
- Performance metrics (e.g., engagement rates, response patterns by content type)
What is NOT included in aggregate data:
- Customer logos
- Brand names or other Customer-identifying information
- Specific campaign content or messaging
- Any data that could identify the Customer as the source
This aggregate approach allows Airship to develop insights such as “users in this vertical who receive this type of content respond in this way”—without ever attributing that insight to a specific Customer. Individual Customers cannot be identified from this aggregated data.
Human oversight mechanisms
When Customers enable certain AI features, Airship may use de-identified and aggregated data to improve AI Functions across the broader Airship platform. This aggregate data includes:
- Vertical/industry classification (e.g., retail, travel, media)
- Content classification (e.g., promotional, transactional, informational)
- Performance metrics (e.g., engagement rates, response patterns by content type)
What is NOT included in aggregate data:
- Customer logos
- Brand names or other Customer-identifying information
- Specific campaign content or messaging
- Any data that could identify the Customer as the source
This aggregate approach allows Airship to develop insights such as “users in this vertical who receive this type of content respond in this way” — without ever attributing that insight to a specific Customer. Individual Customers cannot be identified from this aggregated data.
Customer-specific AI enhancement
Airship also uses certain Customer-provided information to improve AI Functions solely within that Customer’s own account. This Customer-specific data may include:
- Brand Guidelines (mission statements, brand voice, tone preferences, personalities)
- Website URLs (when optionally provided by the Customer for context)
- Historical campaign performance data specific to that Customer
- Uploaded images and content used in AI-assisted content generation
This data is used exclusively to:
- Tailor AI-generated content to the Customer’s brand voice and style
- Provide personalized recommendations based on the Customer’s own campaign history
- Enhance the relevance of AI suggestions within the Customer’s account
Customer-identifying materials such as logos, brand names, and proprietary content are never used to train aggregate models or improve AI Functions for other Customers. These materials remain siloed within the Customer’s account and are used only to enhance that Customer’s personalized AI experience.
Customer controls
Customers retain full control over how their data is used in AI Functions through Service settings. Customers maintain full control which data inputs are provided to AI Functions (e.g., Brand Guidelines, website URLs).
Opting out of AI
Airship’s AI features are designed to enhance the functionality of existing products, not operate as standalone tools. All AI features within the platform are clearly marked so that users know when they are interacting with AI, and engaging with these features is entirely optional. Customers choose when and how to interact with AI functionality, and choosing not to use AI features will not impact access to or use of the Airship platform. For example, with Airship Brand Guidelines, Customers can set all brand guidelines manually or optionally provide source material and let Airship derive guidelines using Generative AI. The same applies across all AI-enabled products: Customers can complete any task manually without leveraging AI. Airship does not automatically process Customer content through AI Functionality. It is always up to the Customer to determine the types of data to be used or processed with any AI tools or features, and Airship maintains a “human in the loop” approach, ensuring human oversight in applying any AI recommendations or Outputs.
For Customers who wish to disable AI features entirely, Airship offers a full opt-out option. Opting out of AI requires a contractual addendum to the Customer’s agreement, under which Airship will disable all AI features within the Customer’s platform environment. Because AI capabilities are integrated across Airship’s product suite, the opt-out applies to all AI features collectively and cannot be selectively applied to individual products. Customers interested in opting out should contact their Airship account representative to initiate the process.
Conclusion
By choosing Airship as your AI-powered cross-channel Customer experience platform, you gain a partner dedicated to helping you navigate the opportunities and challenges of AI in marketing while maintaining the highest standards of regulatory compliance, data protection, and ethical AI use.