Blueshift: Customer AI

Overview:

Blueshift’s patented Customer AI blends predictive, generative, and agentic intelligence to deliver real-time decisioning. It powers the entire platform, enabling smarter decisions across the customer journey.

AI Assistants help marketers generate content and automate repetitive tasks.
AI Agents manage and optimize campaigns in real time.
AI Predictors forecast behavior, channel engagement, and conversion likelihood to enhance targeting and personalization.

Blueshift's Recommendation Studio, built with patent pending Artificial Intelligence technology, is the most flexible and transparent system for building recommendations that can be used by non-technical marketers. Blueshift automatically computes various forms of recommendations: based on events (re-targeting), collaborative filtering, affinity based, trending items, catalog change based etc. These recommendation types can be sorted, filtered and combined in different ways to suit your business needs. Recommendations can be configured from a visual studio that offers previews for different users, and once configured, can be attached to any email, web or mobile template with 1 click.

Blueshift's Responses to Questions on Customer AI

AI Predictors

1. Does your solution offer predictive scoring for customer actions? What scores are available, i.e. buying propensity, user engagement, and affinities? Are scores out-of-the-box and/or customizable?

Yes, Blueshift provides numerous standard predictive models out-of-the-box, including conversion, engagement, churn, and retention models. Blueshift also provides the flexibility for marketers to create their own custom predictive models that are tailored to meet their specific business requirements.

In addition, Blueshift offers channel engagement predictive models to better understand a user's propensity to engage in a specific channel. Out-of-the-box, Blueshift provide channel engagement predictive models for the following channels: email, SMS, push, and in-app.

Additional information can be found in Blueshift's documentation at https://help.blueshift.com under the topic: Predictive Studio

2. Can predictive scores be used for dynamic segmentation and to automatically trigger communications based on a user's score?

Yes, predictive scores are immediately available for use throughout the Blueshift platform, including within the customer profile, segmentation, recommendations, campaigns, and templates. Marketers can use predictive scores to create highly targeted segments that focus on high or low propensity customers so they can remove low-performing audiences by defining which score ranges to target in their campaigns. Similarly, predictive scores can be used in campaign journey building to move customers through different paths and receive different messages and offers based on their propensity to perform a certain action. Predictive scores can also be exported or syndicated to any 3rd party application, paid media channel, or back to your internal systems for deeper analysis.

Predictive models, as well as the scoring against the models, are deployed and updated automatically as new data becomes available without involvement of data scientists or IT team. Predictive scores are refreshed on the user profile on a daily basis, whereas the predictive models are re-run and optimized on a weekly basis.

3. Does your solution provide an interface for business users to generate, deploy, and refresh custom predictive models without involvement of a data scientist or IT?

Yes, Blueshift provides an intuitive interface for marketers to build predictive models without involvement of data scientists or IT team.

Our platform does the heavy lifting and automates data analysis so that no additional work is required by marketers or their data scientist teams to clean, de-dupe, and normalize raw data or do other processes to get the training data ready or remove noise from the models. By looking at all of the behaviors across every user, Blueshift’s predictive scores can understand and predict users' intents and propensities to take certain actions. Once computed, predictive scores are immediately available throughout Blueshift’s platform.

Furthermore, we use a white box, transparent approach, rather than a black box approach, in building predictive models, which allows marketers to get more insight into the model inputs and interpret the model performance. Through rich visualization, marketers have full visibility into which user attributes and behaviors impacted the model the most and can track how the performance of the predictive score changes over time. Marketers can then take this insight to refine models to their desired outcomes.

Additional information can be found in Blueshift's documentation at https://help.blueshift.com under the topic: Predictive Studio

4. Does your solution have the ability to integrate with and bring in in-house models into the system?

Yes, Blueshift provides our customers the ability to bring their own in-house models and scores as custom attributes on the user profile. Using our easy-to-use, marketer friendly segmentation builder, marketers can access their in-house model scores to create segments that target the appropriate users based on these scores. Similarly, the in-house model scores can be used to orchestrate lifecycle campaigns via our intuitive, visual journey builder to move users through different paths and receive different messages and offers based on the in-house scores.

5. Does your solution offer transparent models that provide detail into the model feature inputs and the performance outputs?

Yes, Blueshift’s approach to predictive intelligence is to be fully transparent and customizable. Through rich visualization, marketers have full visibility into which user attributes and behaviors impacted the model the most and can track how the performance of the predictive score changes over time. Marketers can also customize models to their unique business requirements as needed.

Our AI-powered predictive models are trained on historical data to learn behavioral rules, which separate converting users vs non-converting users. Our platform then automatically derives user behaviors from user clickstream events and campaign engagements. Specifically, we derive hundreds of behaviors such as recency, frequency, time spent, catalog affinities (category affinity, brand affinity, product attribute affinity) for each event. These behaviors, along with user attributes, are referred to as features/variables and are fed into the AI model. The model then learns an optimal combination of features, which leads to conversion and a scoring function is learned.

AI Assistants

6. Does your solution utilize deep customer understanding and customer data to automatically generate personalized, brand-aligned content across channels using AI, demonstrably reducing campaign development time while improving relevance and marketer efficiency?

Yes, Blueshift's Assistants are trained on effective marketing techniques as well as an individual brands past marketing efforts and the complete set of data attributes that are stored within the Blueshift platform. By utilizing this rich data set, the AI Assistant is able to not only create effective marketing content and assets, but use the context of the brand and data to use the Liquid personalization language to create highly personalized and engaging content.

7. What specific marketing assets (e.g., email subjects, body copy, CTAs) can your AI Assistants generate and what channels are supported by your AI Assistants?

Blueshift's AI Assistants help marketers in creating images as well as text that can be used for subjects, pre-headers and body content in email, SMS push and web campaigns. In addition, Blueshift's AI Assistant can generate both HTML content as well as Liquid personalization code for marketers using text or voice commands helping to streamline the process of template creation.

AI Agents

8. What specific marketing tasks can your AI Agents autonomously execute, and how are they configured to achieve predefined business goals?

Blueshift offers a range of AI agents as part of our Customer AI suite, designed to enhance marketing automation and personalization. Our flagship AI agent is the Blueshift Campaign Optimizer, which boosts email performance by generating AI-powered subject lines and preheaders. It allows for dynamic personalization using customer, transaction, and catalog data, and facilitates real-time A/B testing to optimize engagement.

Additionally, we are developing more agents to tackle various marketing tasks, such as:

Audience Creation Agents: Build dynamic segments from real-time behavior.
Reporting Agents: Surface campaign insights and trends.
Multichannel Agents: Test and optimize across email, push, and in-app channels.

These agents are designed to streamline campaign management, optimize engagement, and deliver highly personalized customer experiences at scale.

9. How do your AI Agents use real-time customer data to make decisions and self-optimize campaigns without direct human intervention for each action?

Blueshift's AI Agents are designed to be intelligent autonomous agents that can seamlessly create, execute and monitor tasks on behalf of a marketer. For instance, Blueshift's Campaign Optimizer will generate subject lines and preheaders for use in A/B tests and once launching the test will continuously monitor and re-allocate between test and control variants based on real time metrics until statistical significance has been reached and a clear winner can be determined based upon the metric that the marketer has chosen.

10. Describe the guardrails and oversight mechanisms available to marketers when deploying AI Agents for automated campaign execution.

Blueshift's AI Agents contain numerous guardrails when performing tasks for marketers. AI Agents are configured to screen out sensitive and offensive language and will check for industry, brand and campaign relevancy when creating content. When creating subject lines and pre-headers, the agent will also check for consistency with the template copy and check for liquid syntax validity and empty variables. Finally, the marketer can provide additional instructions to the agent to take into consideration when completing its tasks such as a specific tone to use or a specific audience demographic to target.

11. How do your AI Agents learn from past performance and evolving customer behavior to continuously improve their decision-making and task execution?

As Blueshift's AI Agents perform their tasks, they evaluate customer engagement with the content that they have created to influence future content creation. Additionally, the Agent can display the various variants that it creates and the marketer can express their preferences by liking and disliking variants to help the agent determine what resonates, automatically improving future content suggestions.

Recommendations

12. Does your solution have a visual interface in which marketers can create and manage recommendation types without technical knowledge? Is this a native capability or through a partnership/ additional integration?

Yes, Blueshift's recommendation studio provides marketers with an intuitive, easy-to-use drop-down interface to build recommendation schemas that pull directly from your catalog in a self-service manner, without the need for IT teams or data scientists. Marketers can create content blocks with product recommendations, recommended offers, or other brand content like blogs or videos that adapt and dynamically personalize to every individual user in real-time based on their current context, recent activity, and their product, category, or brand interests.

Additional information can be found in Blueshift's documentation at https://help.blueshift.com under the topic: Recommendation Studio

13. Describe what types of recommendations your solution supports (i.e. collaborative filtering, trending items, similar items, new items, expiring items, etc).

Within Blueshift’s recommendation studio, marketers can define recommendation schemes for everything from abandoned content, related items, top trending items, top sellers, most viewed, “users also bought,” next-best-product, expiring items, recent additions, and more. For recommending relevant, contextual predictive content/items/products, we use matrix factorization and collaborative filtering algorithms.

Marketers also have advanced controls to define what kinds of content to include or exclude within each content scheme, mix and match recommendation types, and set how to backfill recommendations to ensure that users receive the most relevant and highest converting offers and messages.

14. Does your solution ingest product, content, and offer catalog data and use dimensions of that data (i.e. product name, price, etc.) within dynamic recommendations?

Yes, Blueshift easily ingests your content and product catalogs in a very flexible way to leverage all the metadata fields from your catalogs. We provide the ability to refresh the catalog based on a scheduled cadence by setting up automated catalog import jobs or for scenarios where real-time update is needed, we also provide a Catalog API that can be used to dynamically update the content and product attributes (i.e, price, availability, etc.) immediately. Many of our customers have large and extensive catalogs, where millions of products are constantly changing over time, and our platform has been architected to scale and accomodate any catalog size to meet your current and future needs as your content and product library continue to grow and evolve over time.

Using our built-in recommendation engine, Blueshift can then easily recommend the most relevant next-best items from the catalogs based on the catalog metadata and user data and behaviors and serve that content dynamically across channels.

Additional information can be found in Blueshift's documentation at https://help.blueshift.com under the topic: Catalogs

15. Does your solution allow us to import recommendations from our in-house team? If yes, describe how.

Yes. Blueshift supports customers importing recommendations from their in-house team by providing the ability to easily upload their recommendations directly into Blueshift as recommendation feeds. Our customers that have a data science team and have their own recommendations typically will combine their external recommendations with our AI-powered recommendations that are built using our recommendation studio and then test, experiment, and optimize the best performing combination of recommendations that result in the highest engagement and conversions in their marketing messages.

Additional information can be found in Blueshift's documentation at https://help.blueshift.com under the topic: Import recommendations overview

16. Does your solution support combining multiple types of recommendations within a template? If yes, describe how.

Yes, using Blueshift’s recommendation studio, marketers can create multiple content blocks within a single recommendation scheme. Multiple recommendation types can then be incorporated into a single message template to ensure that users receive the most relevant and highest converting offers and messages.

17. Does your solution generate unique recommendations for every user based on real-time data? Does it offer the ability to preview recommendations for different users and add filters?

Yes, once recommendations are used within a message template and the campaign is running, recommendations presented to each user update in real-time to each user based on their current context, affinities, recent activity and behaviors, and the latest catalog data. This ensures that the content remains current and relevant and avoids fatiguing users. Blueshift allows marketers to set up message templates that incorporate recommendations once and be confident that the content delivered is always relevant even as users' interests and intentions evolve over time.

Marketers can also preview what recommendations look like for any given user or users within a segment within the message editor.

18. Can your solution generate ready-to-deliver content using templates that select different items (text, images, offers, etc.) for different individuals based on fixed rules, predictive models, or both?

Yes, Blueshift offers powerful, dynamic content capabilities that are highly customizable and scalable. Marketers are able to generate dynamic content and product recommendations with items from your content and product catalogs that are made specific to each user based on their unique user behaviors, preferences, and product affinities.

One of our key differentiators is that with our predictive and machine learning algorithms, marketers are able to provide the type of product and content recommendations that Netflix and Amazon offer without the need for a team of data scientists or a manual process of building complex rules. Within an intuitive dropdown interface, marketers can easily define what kinds of content to include or exclude within each content scheme, mix and match recommendation types, and set how to backfill recommendations to ensure customers receive the most relevant offers and messages content blocks. Once established, content blocks can be easily leveraged across any channel including email, mobile, and web to present relevant, contextual content and product recommendations to each individual.

Marketers can also use the Liquid template language to define business rules and personalize the email template with texts, images, offers, etc. that are specific for each user based on any user attribute or user predictive scores.