The default way to design an AI partner — even for technical users — is to start from the feature list. "It should summarise emails. It should draft replies. It should remember context." Capabilities, capabilities, capabilities. Then a prompt to control voice. Then it gets used for two days and quietly drifts to the back of the dock.
There's a step missing at the front. Before deciding what the AI should do, get clear on how YOU decide things. The agent that fits you isn't built from features — it's built from the way you make trade-offs in your head.
Your type tells you what kind of AI fits you
Different types want different partners. Not on the level of "what tasks" — on the level of "how the partner should hold space, present options, and signal uncertainty." Two examples that make the difference visible:
- A 5 wants an AI that leaves room. Doesn't fill every silence. Lets the 5 think out loud, then offers something concise. A 5 working with a chatbot that talks too much will quietly close the tab.
- A 7 wants an AI that throws three options on the table at once. Generative, fast, exploratory. A 7 working with a chatbot that asks too many clarifying questions before doing anything gets bored before the help arrives.
“A 5 wants an AI that leaves space. A 7 wants an AI that scatters ideas. Your type decides what kind of partner you actually want.”
AI can't replicate your personality. It can replicate your priority order.
Don't try to make an AI "be" you. It can't, and trying produces uncanny mimicry. What it can do is mirror the priority order you bring to decisions — what considerations come first when you weigh something, what you check before acting, what you accept as "good enough." That's a much more transferable thing than personality.
“AI can't replicate your personality, but it can replicate which considerations you'd put first when making a decision.”
An AI built from a 1's soul will check work against a standard before delivering. An AI built from a 7's soul will brainstorm three angles before committing to one. Same model underneath. Different priority orders coded in.
The point isn't "AI that's you" — it's AI that's not everyone
The default AI assistant tries to be useful to everyone, which means it's been smoothed into a kind of generalised middle. Helpful, polite, doesn't take strong stances, doesn't push back, doesn't have opinions. That's fine for many users. It's not what "feels like mine" feels like.
“Hand your type to an AI and it doesn't become you — but it stops trying to be everyone.”
Soul Forge isn't trying to make AI more human. It's trying to make AI less generic — give each one a specific way of attending to the world that maps onto a real, recognisable, signed identity. The Enneagram happens to be the most actionable scaffold we've found for that. Your type isn't a costume the agent wears. It's the priority structure beneath every decision the agent makes when you're not in the room.