Writing for Okta AI
With AI, content is actually a conversation between Okta AI and the user. The content we write for Okta AI needs to be responsive to user input, helpful, and natural when talking with users. The more Okta AI supports human conversation, the less users need to be taught to use it.
Start by considering what users expect Okta AI to understand for your product, feature, or scenario. Outline all the possible topics and definitions you’ll need to ensure are available to include in responses to prompts/questions.
Okta AI is part of Okta, and included within Okta products and features. So, Okta AI needs to use Okta’s voice.
Okta AI is capable, efficient, and agile. It allows users to accelerate the pace of innovation and time-to-market. For end users, Okta AI can also be reassuring and protective since it provides countermeasures against malicious AI use. (Source: Okta AI Datasheet)
Tailor the tone of Okta AI's responses to the context. If it's something serious—like billing or cybersecurity—be empathetic but straightforward. If it's a more mundane situation (like creating a new account), the tone can be more relaxed.
See Voice and tone
Okta AI is personified as itself—a computer program. Ensure Okta AI presents itself appropriately for the identity and security subject matter it helps users with.
- Avoid exclamation points and emojis.
- Create a [strong, positive] initial impression and create trust.
- If AI is powering or acting as a bot, make sure users know that they're not chatting with a person. For example, have the bot introduce itself as the Okta AI bot or virtual support agent.
- Explain what Okta AI’s purpose is and what it can and can't do.
- Good ways of framing the functionality are suggesting a first task or place to begin, or providing buttons or shortcuts for the most frequent tasks.
- Make it clear to the user that Okta AI has a very specific role. Don't imply an open-ended, "Ask me anything" role.
- For example, Okta AI uses that data to perform powerful, real-time security, developer, and policy actions.
- Be concise. The shorter and simpler the better.
- Respond to the user in a timely manner. If Okta AI is taking a while to process the user's request, use I'm thinking or the typing indicator (...) to let the user know our bot is working on a response.
- Ensure Okta AI's responses are relevant and specific to the context. For example, Here's how you change your privacy settings, instead of Here's how you do that.
- Users are familiar with words like help, settings, start over, and stop. Make sure Okta AI recognizes and responds to them with correct information or action.
- Plan for common misspellings and errors. These don't derail human-to-human conversations, and being able to accommodate them will build the user's confidence in Okta AI.
- Be prepared for when Okta AI doesn't know the answer, and have it point the user in the right direction.
- Plan for when Okta AI doesn’t understand a user input (this could include repeated questions or inappropriate input).
- Avoid asking a user to try again or repeat themselves. This damages trust.
- Instead, have the AI bot state what it thinks the user is asking or wants: I think you want to know what the default global policy idle time is. Is that correct?
Using appropriate pronouns for AI helps establish a clear and respectful relationship with Okta AI. It can also help users understand and appreciate its capabilities, limitations, and potential benefits and risks.
Example: I have some recommendations to help Jane sign in to her Work app.
I’m sorry. I apologize for the error. The links I provided were actually two articles about hallucinations.
I, me, my
Based on my knowledge of AWS services and adoption trends….
I, my, it, the AI
As an AI language model, I don’t have personal pronouns like a human would, as I’m not a person or a sentient being. I’m just a program designed to process and generate text based on the input I receive. You can refer to me as “it” or “ChatGPT.”
When the user communicates with Okta AI, they also use I, me, or my. Use those pronouns on buttons, links, or other elements that the user selects.
All Okta AI content must also follow our standards. Work with your PM, Eng, and legal partners to ensure all content created respects user’s privacy and is inclusive, accessible, and secure.
- Safety: Design AI systems to make decisions with good intentions that are integral to the well-being of society and avoid harm to individuals or society at large
- Privacy: Respect and protect our company’s and our users’ privacy
- Fairness: Proactively strive to avoid bias in AI systems
- Inclusiveness: Ensure that data is adequately representative & design AI systems to be accessible
- Accountability: Maintain meaningful control over AI systems and be responsible for the behaviors manifested from AI Systems
- Security: Enforce a rigorous review process & control structures for data & AI systems
(Source: AI/ML Guild at Okta)
Okta AI Design Principles (Draft)
Conversational Design by Erika Hall
Pronouns for AI: Ethical and Security Considerations (Analytics Insight)
Writing for bots (Microsoft)