Gentic Influencer

Workflow

1. Brand setup (required first)

Check if a brand profile exists with get_brand. If empty or missing, walk the user through save_brand before anything else. A good profile needs:

• description: brand identity, values, voice, target audience
• matchmakingInstructions: what makes a creator a good fit (content style, audience demographics, deal-breakers)
• activeCampaignContext: current campaign or product focus
• initialOutreachCallToAction: the CTA paragraph for outreach emails — tells creators what to do next (e.g. product seeding offer, "reply if interested", "book a call")
• emailSignOff: sign-off line for outreach emails (defaults to "Team {brand_name}")

If the user has a website, use fetch_page to pull context and draft the profile for them.

2. Upload existing creators (optional)

If the brand already has creator relationships, use upload_excluded_creators to exclude them from future searches. Two input modes:

• usernames array — pass usernames directly (the agent reads a CSV or paste and extracts them)
• googleSheetsUrl — brand shares a public Google Sheets link; the tool fetches it server-side and extracts usernames from the first column

The tool auto-normalizes input: lowercases, strips @, and extracts usernames from Instagram/TikTok profile URLs. Deduplicates on insert. Capped at 5,000 per call.

3. Search creators

Use search_creators to find creators. Key behavior:

• Always search both platforms (Instagram and TikTok) unless the user specifies one. Run both calls in parallel.
• Write descriptive searchText — "fitness micro-influencers who post workout routines and meal prep" beats "fitness influencers"
• Use filters: minFollowers, maxFollowers, country, minEngagementRate to narrow results
• campaignContext auto-reads from the brand profile — only override if the user gives campaign-specific instructions
• Results are stored automatically. Each search gets a unique run_id.

4. Score creators

After searching, run score_creators to AI-score the matches against the brand profile.

• Scoring is async. Tell the user results are processing and will be ready shortly.
• Score both platforms if both were searched.
• The brandContext auto-assembles from the brand profile — don't override unless asked.
• Scoring writes final_match_score (0-100) plus sub-scores back to the database.

5. View results

Use get_creator_results to review scored creators inline. When presenting results:

• Format as a ranked table with columns: rank, username, platform, followers, engagement rate, match score, one-line analysis
• Bold any creator with score >= 80
• Below the table, highlight the top 3 with a short paragraph each: why they're a fit, any risk factors, and their personalized pitch angle
• If total exceeds count, mention how many more results are available
• If results have null scores, they haven't been scored yet — tell the user to run scoring first

6. Draft outreach emails

After scoring, use draft_personalized_creator_outreach_email to generate personalized emails for top creators.

• Requires a brand profile with initialOutreachCallToAction set — prompt the user to add one via save_brand if missing.
• Only drafts for creators that have been scored and don't already have an email.
• minScore defaults to 70 — only high-quality matches get emails.
• Capped at 10 per call. Emails are stored in the database and visible in get_creator_results and export_creator_results.
• Sign-off is appended automatically from emailSignOff (or defaults to "Team {brand_name}").

7. Export

When the user wants to share or save results, use export_creator_results to generate a CSV download link. Mention the link expires in 1 hour.

Important notes

• All tools are org-scoped. Users only see their own data.
• Mention costs when the user is doing a large search (50+ results).
• If a user asks to "find influencers" or "find creators" with no prior context, start from step 1.