Introducing Adaptive Generation Strategies
Today, we’re excited to announce an enhancement to Jacquard’s Language2 generation engine, which is designed to deliver higher-performing content and further increase the likelihood of outperforming existing content.
Our latest update introduces Adaptive Generation Strategies, a set of improvements that enable Jacquard’s Optimise campaigns to more intelligently tailor their output using AI-driven feedback loops. This ensures we’re generating variants with the greatest potential for strong performance, while still maintaining diversity and creativity. These improvements are live in the platform today. Here’s what’s included in this release:
1. Smarter language, shaped by performance
Language2 now continuously learns from new feedback loops. This means we automatically generate content inspired by both your control content and/or the current winning variant. By analysing the style and techniques in this content, Jacquard creates additional variants using those same proven copywriting patterns.
For continuous trigger-based campaigns, Jacquard also uses ongoing performance signals to focus generation on approaches that are delivering results, while reducing emphasis on those that aren’t. This means that every campaign adapts based on the unique learnings for your touchpoint, journey, and audience.
2. Adaptive variant selection for better results
Jacquard now evaluates your control content alongside the variants it generates using our unique Neural2 performance prediction capability. This helps the system better understand how each piece of content is likely to perform. When the system identifies your control content is already strong, Jacquard now adapts by giving more weight to performance-driven variant types. This increases the likelihood of surfacing winners without compromising variety.
Adaptive Generation Strategies represent an important step in how Jacquard applies learning from live campaigns, allowing generated language to adapt more intelligently over time. No customer action is needed to benefit from these improvements, which will take effect for all customers on Language2.
Word and phrase capitalisation settings
We’ve now added a new feature to the Brand Voice which allows users to define specific words or phrases which must always appear with a specific casing applied. This provides a powerful new way to ensure that your brands, products or other entities are accurately represented in copy, reducing the need for edits and corrections.
At the bottom of the Capitalisation Settings module in the Brand Voice, there’s a new section where you can add specific capitalisation preferences that need to be maintained in your language generation.
.png?sv=2022-11-02&spr=https&st=2026-03-13T09%3A37%3A16Z&se=2026-03-13T09%3A49%3A16Z&sr=c&sp=r&sig=m0ST4aiToXAlWtQhMUS7LvDMopSvJCy8Z%2FFqp4nwiMQ%3D)
Some points to note:
These special capitalisation settings will apply in all contexts, regardless of position and surrounding punctuation.
Having words or phrases in this list has no impact on the frequency of those appearing in the content.
Brand Voice settings will take precedence over Focus Theme capitalisation. For example, if a Focus Theme is defined as Brompton e-motiq but is entered into the new capitalisation setting as Brompton e-Motique, the new capitalisation setting in the applied Brand Voice will take precedence.
Brand Voice Instructions can still be used to influence other aspects of language.
Other improvements
Adaptive generation strategies
We’ve made an improvement to the adaptive generation prompts to better handle control and winning variants that include Personalisation Tags. Now, the new variants will only use tags that were specifically selected in the brief. This is to avoid replicating tags from human controls which may not have been set up in Jacquard.
SMS generation
Our SMS prompt has been updated to encourage more variants with longer content, especially useful for those also sending MMS content.