Start free

Case study · Luxury fashion · travel creators · Influencer marketing agency, UK

A luxury handbag launch roadshow: from a one-line brief to a priced creator proposal, without opening a platform

An agency had a week to staff a luxury handbag brand's launch roadshow for a new travel handbag, in Manchester and Birmingham and then London, with travel creators. The whole job ran inside Claude with Zaver connected over MCP and Gmail connected for sending and reading mail. Day one: 77 travel creators found and checked, 62 emailed with personalised rate requests. Day two: quotes read from the inbox into a comparison sheet with a markup tab. Day three: a client-facing proposal with live follower counts, reasons per creator and backups, then audited against the quotes. Over the following days 18 creators replied with rates, and 7 were confirmed for the campaign.

Claude (desktop app)Zaver Radar over MCPGmail connectorSpreadsheet and doc artifacts

62personalised rate-request emails sent the same day from 77 creators found and checked, each approved by the marketer first
18creators replied with their rates over the following days; 19 replies had arrived by the next morning
7creators were sent final go-ahead confirmations after the client approved the proposal
3 daysfrom the first line of the brief to the client's go-ahead on a proposal with backups and an audit

The job

The brief was short and the deadline shorter: a luxury handbag brand was launching a travel handbag with a roadshow in Manchester and Birmingham the following week, followed by a launch event in London. The agency needed local travel creators in each city, their public contacts, their rates for an event visit plus Reels, Stories and TikToks, and a client-ready proposal inside a fixed budget. The usual way is a discovery platform for the search, a CSV export, a mail merge tool, an inbox, a spreadsheet and a document, with a person carrying data between them for several days.

How it ran, step by step

Every scene below is a redacted rendering of the real exchange: the ask as typed, the tools that ran, and what came back. Handles, names, emails and prices are masked.

  1. 1

    Day 1 · morning

    The brief, in plain words

    Find influencers from Manchester and Birmingham who talk about travel

    Claude called Zaver's database search first. It came back thin and mostly brands, so it ran a discovery search per city instead and returned about 25 travel creators with follower counts and a one-line fit note each, with travel agents and brand accounts left out. It also flagged a large account that Radar marked as possibly a brand and reminded the marketer that city comes from the bio, so a few profiles were worth a manual look.

    search_creator_databasediscover_creatorsget_job_status

    Chat scene: the brief and the first discovery results for Manchester and Birmingham, handles masked
  2. 2

    Day 1 · morning

    Public contacts, and three creators dropped

    Find their email id as well

    Contact enrichment ran on all 23 profiles. Claude scheduled its own check-back and reported when it finished: 18 had a public email. Reading the full bios that came back with the contacts, it moved three creators to a "probably not a fit" list, including an illustrator born in Manchester who now lives in Los Angeles and a magazine editor whose phone code put her outside the UK.

    enrich_creator_contactsget_enriched_contacts

    Chat scene: enrichment complete, 18 of 23 with public emails, masked
  3. 3

    Day 1 · midday

    One draft approved, 15 personalised emails sent

    Draft a first sample email and if ok then I'll give a go ahead for all

    Claude drafted the rate request with the four deliverables (event visit, Reel, Stories, TikTok), asked the marketer for the details it did not have, and took two rounds of edits: name the event a roadshow, and ask for the TikTok link and Instagram handle. On "ok send to all now" it sent 15 separate emails from the connected Gmail account, each with the creator's first name in the greeting, and listed the five creators who had no email and needed an Instagram DM instead.

    Gmail send_message

    Chat scene: the approved email draft and the confirmation that 15 emails went out
  4. 4

    Day 1 · afternoon

    Two more searches: the wider UK, then London

    Rerun the search and email id discovery for other travel based UK creators

    A UK-wide run found 35 new creators across London, Scotland, Wales, Bristol, Leeds and Northern Ireland; 33 had emails after enrichment, and 32 were emailed with a variant that asked whether they could travel. The London-only discovery job then stalled at 15 percent twice. Claude said so, started a different search, and when that stalled too it fell back to a keyword search of Zaver's database, which produced 19 London creators with emails. The marketer asked for only the travel-focused ones, so 15 were emailed and the four food accounts were left out. Total for the day: 62 creators contacted.

    discover_creatorssearch_creator_databaseenrich_creator_contactsGmail send_message

  5. 5

    Day 1 · evening

    Replies read from the inbox, rates in a table

    Can you also read emails from Gmail?

    Nine creators had already replied. Claude read the threads and sorted them: three interested with full rates (visit fee, Reel, Stories, TikTok, plus a media kit and TikTok links), one asking for the date, two declines, three auto-replies. It also corrected itself on which address the emails had gone from. All three with rates wanted the brief, so it offered to draft one reply with the dates and times.

    Gmail search_threadsGmail get_thread

    Chat scene: nine replies read from Gmail, three with rates in a table, names masked
  6. 6

    Day 2

    A comparison sheet with a markup tab

    Check email and build a sheet with Instagram URL, TikTok URL, cost for event, per Reel, per Story, per TikTok, travel cost. Add an x% agency markup on each quoted cost.

    Ten more replies had arrived overnight. Claude read them and built a workbook of 30 creators: 16 who replied by email (source "Database") and 14 from a partner's list the marketer pasted in (source "RM"). Each row has both profile links, the four rates, travel, package price, notes and status. The markup went on a second tab as formulas driven by one yellow cell, so the quotes stay untouched and the percentage can change later. It also flagged what was not like for like: per-frame versus per-set Story prices, one quote in US dollars, one rate card locked in a PDF.

    Gmail search_threadsGmail get_threadspreadsheet artifact

    Spreadsheet scene: quoted rates with a markup tab driven by one editable cell, names masked
  7. 7

    Day 3

    The proposal, with live counts, reasons and backups

    Create a client-facing proposal with creators included and costs proposed. Give links and follower counts. Give alternate options with client-side cost. Give reasoning on the strength of each creator.

    Claude planned three creators per city inside the budget, with the exchange rate, markup and contingency each in one editable cell on an internal tab, then wrote a client-facing document that shows only the planned creators and their price. It pulled live Instagram follower counts for all nine from Zaver, added a backup section of ten creators with the same deliverables priced, and a "Why this creator" column per row. The live counts exposed weak spots: one planned creator had 3,000 Instagram followers, another's Reel would run on a 27,000-follower second account, and three were entertainment rather than travel creators. It proposed swaps that stayed within budget.

    get_creator_profiledoc artifact

    Document scene: the Manchester line-up with reasons per creator, and the audit of the exported PDF
  8. 8

    Day 3

    The audit

    Evaluate this PDF for match and any mismatch

    The marketer exported the proposal to PDF after editing it by hand, and asked for a check. Claude compared it with the quotes, the live doc and the workbook. The arithmetic was right, but the total had gone over budget with the budget rows removed, one creator's price did not match any quote, the creator count in the overview was stale, one TikTok handle had a letter missing, and the exchange-rate note had been deleted. It then re-checked every creator against the full brief of one Reel, three Stories and one TikTok, and listed exactly what each package still needed.

    Read PDFdoc readworkbook read

Who did what

The marketer

  • Wrote the brief in one line
  • Approved the email draft and asked for two edits
  • Said "send to all" three times, once per batch
  • Pasted a partner's creator list
  • Set the budget, markup and the deliverable package
  • Edited the proposal by hand and asked for an audit

Claude

  • Chose between database search and discovery per city
  • Scheduled its own check-backs on long jobs
  • Drafted and personalised every email
  • Read and sorted replies
  • Built the workbook, the markup tab and the plan
  • Wrote the proposal and checked the PDF against the data

Zaver, over MCP

  • Creator discovery by city and niche
  • Database keyword search
  • Public contact enrichment
  • Live profile and follower counts
  • Fit notes and brand-account flags
Outcome

Sixty-two creators were contacted with personalised rate requests on the day the brief was written. Rates were in a comparison sheet the next morning. A client-facing proposal with live follower counts, reasons and ten priced backups was ready on day three, and a second plan for the London launch event was built from the same pool the same day. Over the following days 18 creators replied with their rates (19 replies of all kinds had arrived by the next morning), and once the client approved the line-up, 7 creators were sent final go-ahead confirmations.

Run the same flow

Connect Zaver to Claude or ChatGPT at /mcp, connect Gmail in the assistant's connector settings, and start with the first ask. The playbooks cover each step on its own.

Find influencer email addressesVet a creator list before outreachWhat to pay an influencerConnect Zaver to Claude

Questions

Is this a real campaign?

Yes. It is reconstructed from the actual working session in Claude, with the brand, the creators' names and handles, email addresses and the agency's budget and markup removed. The scene images on this page are redacted renderings of those messages, not screenshots of the original thread.

What did Zaver do, and what did Claude do?

Zaver supplied the data: discovery by city and niche, database keyword search, public contact enrichment and live profile counts, all over MCP. Claude did the reasoning and the writing: choosing which search to run, drafting and sending emails through the Gmail connector, reading replies, building the workbook and the proposal, and auditing the PDF. The marketer approved every send and set every commercial number.

Did anything get sent without approval?

No. Each batch went out only after the marketer approved the draft and typed "send to all". Claude asked for missing details before the first send and listed the recipients before each batch.

Where did the creators' emails come from?

Public business contacts only: the address in the creator's bio or linked page. Creators without one were listed as "no email, DM" and were not guessed.

What would it cost in Zaver credits?

Roughly: each discovery search is 1 credit, each contact enrichment 1 credit per profile, and live profile lookups for the proposal 1 credit each. The whole run fits inside the Growth plan's monthly pool many times over.

Can I run the same flow?

Connect Zaver to Claude at /mcp, connect Gmail in Claude's connector settings, and start with the first ask on this page. The playbooks linked below cover each step on its own.

Reconstructed from the working session and published with the customer's permission. Brand, creator identities, contact details and commercial figures are redacted. Scene images are renderings of the redacted messages, not screenshots of the original thread.