Introducing voice: customize generative AI apps to your style - WRITER

Introducing voice: customize generative AI apps to your style and tone

Writer Team | March 26, 2024

As generative AI tools proliferate, enterprises have paid close attention to output breadth and sophistication. But for generative AI to be usable in the enterprise, it’s not enough to simply generate outputs. You also need to meet compliance, brand, and accuracy standards. Without the appropriate guardrails in place, generative AI will be difficult to scale to your highest-impact use cases and workflows.

That’s why we’re excited to add new voice capabilities to our suite of AI guardrails, making it easier than ever for teams to build generative AI apps that generate outputs in their company’s unique voice and writing style.

Introducing voice capabilities

To enable our voice capability, simply upload sample copy, and WRITER will analyze the nuances of your writing to build a voice profile. The profile can be applied to every output by default, ensuring consistency across your entire workforce. You can also create profiles for specific products, channels, audiences, or asset types.

Starting today, voice will be available in our AI apps and rewrite feature. AI apps can also be created, enabling you to generate on-brand outputs for your most unique or highest-value use cases.

Writer voice

AI’s default, neutral business voice

Marketing longform

Professional style that’s easy to understand and engaging to read

Internal operations

Very approachable style that uses a casual vernacular

CIO exec voice

Straightforward and professional style that’s slightly technical

Financial journal

Verbose and sophisticated, bordering on academic

To activate this feature, just upload some sample text, and Writer will get to work analyzing the subtleties of your writing to construct a voice profile. This profile can then be set as the default for all your content, guaranteeing a uniform voice across your entire team. Or, if you prefer, you can create distinct profiles for different products, channels, audiences, or asset types.

Powered by specialized, WRITER-built LLMs

Most other generative AI tools create branded writing with a generic LLM — relying on users to prompt it with the correct instructions. This is problematic. Users may struggle to articulate the subtleties of their writing style, generic LLMs may fail to replicate a style correctly, and organizations can’t leave it to chance that all users across all functions will prompt their LLMs the same. To address this, we built two specialized models to add to the WRITER Palmyra family of LLMs:

This approach creates nuanced and accurate voice profiles, capturing stylistic tendencies that a writer might struggle to articulate in a prompt. When it comes to creating outputs, voice generation is able to analyze a profile to create content that accurately matches the writer’s style. Combined, these two models empower your brand leaders to create and scale a voice profile across your organization, resulting in consistent, on-brand content everywhere people work.

An example voice profile generated from sample text

On-brand, enterprise use cases

Voice can be used across a variety of AI apps to quickly deliver content that’s on-brand or stylized for specific use cases.

The increasing importance of AI guardrails

As enterprises adopt generative AI, their need to manage outputs is increasingly urgent. Voice is just the latest example of how we’re prioritizing the governance features of our full-stack platform — making WRITER the preferred generative AI platform for the world’s most reputable brands, such as Vanguard, and more.