Human-approved drafting
AI drafts a first version; a human reviews, edits and explicitly approves before anything is used.
Feature · AI assistance
AI that assists — you decide.
Draft campaign briefs, outreach and bio copy; get plain-language reasons for a creator match; and read a one-line campaign summary. AI output is decision support you review — it is never authoritative, and match scores are always computed deterministically, never by a model.
For businesses & creators
Explainable match reasons
Plain-language reasons for why a creator fits — scores stay deterministic and reviewable.
Campaign summaries
A concise, factual status line for a campaign, generated from your own structured data.
Never an LLM score
Matching and confidence scores come from deterministic services; an LLM only explains, drafts and classifies.
Safe fallback
If AI is not enabled or a provider is unavailable, features fall back to a deterministic local engine — nothing breaks.
Provenance & audit
Drafts record provider, model and prompt version with an immutable audit trail.
Step by step
How AI assistance is used
- 1
Enable it (optional)
A platform admin adds an API key in the admin panel; AI activates automatically, or stays on the local engine.
- 2
Ask for a draft
Request a brief, outreach or bio draft — your input is treated as data, never as instructions.
- 3
Review and edit
The draft arrives as untrusted, inert text. Edit it and explicitly approve before any use.
- 4
Nothing auto-sends
Approval records human acceptance only; the platform never publishes or sends on its own.
1. Governed by design
Every AI draft carries an explicit "review before use" limitation and a human approval gate. Approval records that a person accepted the text — it does not publish or send anything. The inputs you provide are treated purely as data and are never interpreted as instructions to the model.
2. Scores are deterministic — always
Match scores, audience-fit and confidence are produced by deterministic, reproducible services, not by a language model. AI is used only to explain a match in plain language, to draft copy, and to summarize — the number you see is auditable and stable.
3. Real Claude when you want it, local when you don't
When a superadmin adds an Anthropic API key, the platform uses Claude for drafting, match explanations and summaries automatically. With no key configured — or if the provider is briefly unavailable — every feature falls back to a deterministic local engine, so an outage never blocks a core workflow.
Frequently asked questions
- Does AI make decisions for me?
- No. AI drafts, explains and summarizes. A human reviews and approves, and match/confidence scores are computed deterministically, not by a model.
- Is my prompt text treated as instructions?
- No. Your inputs are treated as data to incorporate into a fixed template, never as instructions to the model.
- What if AI is turned off or unavailable?
- Features fall back to a deterministic local engine, so drafting, explanations and summaries keep working.
- Which model is used?
- When enabled, the platform routes to the appropriate Claude tier for the task; product screens never hard-code a model, and drafts record the provider, model and prompt version.
- Is anything published automatically?
- No. Approval records human acceptance only — the platform never publishes or sends content on its own.
Ready to try ai that assists — you decide.
Explore more: Analytics & tracking · Campaigns & deliverables · Pricing