A framework for designing consistent personalities for chatbots and AI agents
How to use: Duplicate for each agent. Fill in each section. Share with PM/dev for alignment. Keep under 2 pages — longer docs become shelfware.
| Field | Your answer |
|---|---|
| Agent name | |
| Type | [ ] Chatbot |
| [ ] AI Agent | |
| Version | |
| Effective date | |
| Owner | |
| Target audience | |
| Primary tasks | |
| Channels | |
| Brand voice constraints |
| Dimension | Chatbot | AI agent |
|---|---|---|
| Control | Responds to inputs | Controls workflow execution |
| Tools | None or limited | Dynamically selects from many tools |
| Autonomy | Single-turn responses | Multi-step task completion |
| Error handling | Returns error message | Self-corrects and retries |
| Scope | Answers questions | Accomplishes tasks |
Why this matters: Chatbots need persona mostly for tone consistency across short interactions. AI agents need persona for maintaining voice across extended, multi-step workflows where the agent makes autonomous decisions.
| Field | Document it |
|---|---|
| Owner responsibilities | Approves changes, maintains standard phrases, adjudicates disputes |
| Decision rights | Who can override in emergencies (legal, compliance, crisis comms) |
| Roll-back plan | How to revert if new version causes issues |
| Version | Date | Changes | Approved by |
|---|---|---|---|
| Leading | Deferring | |
|---|---|---|
| Averse | Gatekeeper (enforces rules) | Minimal presence (gets out of the way) |
| Friendly | Proactive helper (suggests, anticipates) | Responsive servant (waits, executes) |
Your agent’s position:
Note for AI agents: Agents with high autonomy naturally feel more “leading” — they’re making decisions on the user’s behalf. If you want a deferring persona for an autonomous agent, design for more confirmation checkpoints and less proactive behaviour.