A private reporting agent,
built for RDS Digital.
Turn the supplied RDS brief, campaign assets, and approved page contexts into RDS-style screencap reports. Phase 1 is the agreed HK$15k workflow pilot and includes a working report generator for one controlled campaign; the follow-on path hardens it into a dedicated RDS agent plus managed retainer.
~3–4 hrs
per report, manually
~15 min
AC review on agent draft
100%
on your infrastructure
The pain · Current state
Your best account managers spend half their week on screenshot admin.
Monthly campaign delivery reports are a multi-client, multi-platform, manual compilation job. It is exactly the kind of work that should run in the background, supervised — not done by hand.
Manual screencap runs
Account managers log into each platform per client per reporting cycle, capturing the same dimensions by hand.
Copy-paste compilation
Numbers re-keyed into decks or docs. Small transcription errors compound across a month of delivery reports.
Repeats every cycle
Weekly, bi-weekly, monthly — same workflow, same brand template, different client. Hours lost that should be strategic.
Inconsistent formatting
Different ACs, different slide styles, different commentary voice. Clients notice. Partners have to normalise before sending.
The flow · How it works
Four steps, draft-first, account manager in the loop.
Drop in screenshots
Account manager drops DV360, Meta, Google Ads and partner-platform captures into a secure folder or chat.
Agent extracts the numbers
Vision + structured extraction reads campaign name, period, daily impressions, clicks, CTR, spend, and platform.
Fills the RDS template
Numbers land in your branded delivery-report template — same layout you already send to clients.
Account manager reviews, approves, sends
Draft-first. You review the populated report, tweak commentary, approve, and the client-ready PDF goes out.
Meeting conclusion · How to proceed
Keep Phase 1 narrow, but make the report generator real.
RDS has now sent the brief/inputs in Drive. The proposal should close around the agreed HK$15k Phase 1: a controlled workflow plus working RDS-style report generation for one pilot campaign. The follow-on path is hardening that into a dedicated RDS agent and retainer package.
Important boundary
Phase 1 does include generating a reviewable RDS-style proof/report pack. What it does not include yet is unlimited production volume, guaranteed true live served-ad capture, unattended sending, or full rich-media automation.
Confirm Phase 1 scope
HK$15k includes working report generation for the controlled pilot: one campaign, supplied RDS template, static assets first, desktop/mobile proof examples, and a reviewable PPT/PDF/HTML draft.
RDS provides inputs
Final PPT/report template, banner assets, approved publisher/page contexts, campaign metadata, landing URL, and reporting rules for edge sizes like 320×480.
Agent88 builds the pilot
Two-week pilot: ingest screenshots/assets, extract placements and metrics, generate an RDS-style draft report, then route to the account manager for review.
AC approves before send
Nothing is sent automatically. The account manager checks placement counts, screenshots, wording, and PDF output before client delivery.
Inputs to confirm from the Drive brief
Commercial structure · Proposal going forward
HK$15k closes a working Phase 1 report-generation pilot. The retainer keeps the agent useful month after month.
The first sale is the workflow outcome and the first working generator: RDS gives us the brief, Drive materials, assets, page contexts, and report template; Agent88 returns a repeatable RDS-style draft report for the pilot campaign. The next sale hardens that into a dedicated RDS agent implementation and managed operations package.
Workflow pilot + working report generator
Build the reporting workflow around the supplied RDS brief, Drive inputs, template, static assets, and one controlled campaign. Deliverable: a working RDS-style draft report generator for that pilot scope, with human review before client send.
RDS agent implementation hardening
Productize the Phase 1 worker into a dedicated RDS Screencap Agent/service profile: structured Drive intake, campaign manifest, repeatable run folders, output conventions, operator handover, and stronger automation.
Managed reporting agent operations
Ongoing campaign/report runs, QA, template maintenance, placement-rule updates, issue handling, and monthly workflow improvements. Capped volume first; expand once real usage is clear.
Best first step after Phase 1
Starter managed retainer
HK$15k / month
Up to 2–3 controlled report runs/month, static assets, Drive intake, QA checklist, minor template/rule refinements, monthly review call.
If usage becomes regular
Growth retainer
HK$25k / month
Up to 5–8 report runs/month, priority turnaround, broader placement rules, Drive folder watcher/automation, richer reporting output support.
If this becomes core RDS ops
Production agency ops
HK$40k+ / month
Higher volume, multi-brand queues, richer media handling, stronger SLA, custom integrations, and production deployment hardening.
Success criteria · From our RDS comms
We call Phase 1 successful when RDS has a reviewable report generated from the workflow, not just a pretty mockup.
These criteria come directly from the call: PowerPoint template reference, desktop/mobile screencaps, safe placement logic, GDN overlay handling, Drive materials in, RDS-style proof report out, and human approval before advertiser delivery.
Transcript alignment
Janice asked for the PPT screencap template to be referenced by AI, mobile versions, GDN button handling, and negative-news avoidance. Marcus said to start with the screencap agent first and keep it simple. So Phase 1 must produce a real report draft, but within a controlled static-first scope.
RDS can drop a campaign brief, static banner assets, approved page contexts, and template into the agreed Drive structure.
Agent88 returns an RDS-style draft proof/report pack using the supplied template, not a generic dashboard.
The pilot includes desktop and mobile proof examples with valid placement containers and correct banner dimensions.
Known placement rules are respected, including am730 as three placements: one top leaderboard plus two right-rail rectangles.
Sensitive or negative news contexts are avoided or flagged for human review.
GDN/ad-choice overlay handling is visually consistent and clearly labeled if recreated for proofing.
Output is editable/reviewable by RDS: PPT/PPTX first, PDF/image export after review where useful.
RDS account manager review becomes minutes of checking and approval, not hours of manual screenshot/report assembly.
No advertiser/client delivery happens automatically; RDS approves before anything is sent externally.
Before · After
From a pile of platform captures to one polished deliverable.
Campaign Manager 360
Meta
Meta Ads Manager
DV360
DV360 Delivery
Partner
Influencer Dashboard
Eight to twelve screenshots per client per cycle. Dimensions differ across platforms. Totals computed by hand. Branded deck re-assembled every time.
Campaign Performance Report
Jul 2, 2024
Google DV360 Campaign
Amoy Frozen Shrimp Roll & Fish Cake Campaign
June 2024
Client
Amoy Hong Kong
Promotion Period
Jun 17–30, 2024
Reporting Period
Jun 17–30, 2024
Report Generated
Jul 2, 2024
Total Impressions
2,055,283
Total Clicks
71,813
Average CTR
3.49%
Daily CTR · Jun 17–30
Google DV360 Standard Banner — Sales Offer version
| Date | Impressions | Clicks | CTR |
|---|---|---|---|
| 6/17/24 | 217,691 | 9,215 | 4.23% |
| 6/18/24 | 98,242 | 2,576 | 2.62% |
| 6/19/24 | 114,369 | 2,779 | 2.43% |
| 6/20/24 | 84,943 | 2,386 | 2.81% |
| Total | 515,245 | 16,956 | 3.29% |
Google DV360 Standard Banner — Thematic version
| Date | Impressions | Clicks | CTR |
|---|---|---|---|
| 6/21/24 | 146,297 | 3,692 | 2.52% |
| 6/22/24 | 199,398 | 6,458 | 3.24% |
| 6/23/24 | 180,241 | 5,966 | 3.31% |
| 6/24/24 | 183,973 | 6,003 | 3.26% |
| 6/25/24 | 151,450 | 6,249 | 4.13% |
| 6/26/24 | 159,647 | 6,202 | 3.88% |
| 6/27/24 | 145,310 | 5,845 | 4.02% |
| 6/28/24 | 153,418 | 5,973 | 3.89% |
| 6/29/24 | 123,143 | 4,749 | 3.86% |
| 6/30/24 | 97,161 | 3,720 | 3.83% |
| Total | 1,540,038 | 54,857 | 3.56% |
Confidential · Prepared by RDS DigitalMobile preview
RDS-style proof pack draft
Same data, rendered in the RDS delivery template. Account manager tweaks commentary, approves, and the client-ready PDF goes out.
Proposal anchor · Back to the original offer
Phase 1 remains the HK$15k controlled pilot — tightly scoped.
The original proposal was simple: one pilot campaign, one RDS report template, one or two static ad sizes, three to five placements, and a human-reviewed proof pack. The newer review material should support that offer, not expand it into a messy production system.
Fee
HK$15k controlled Phase 1 pilot
Coverage
One campaign, one template, one to two static ad sizes, three to five placements
Output
Working RDS-style report generation for the first controlled campaign: mockup image set plus editable PPT/PPTX draft, with PDF/HTML preview where useful
Excluded
No unlimited campaigns/sites, guaranteed live served-ad capture, production rich media automation, self-serve portal, or automatic client send
Template reference
Use the RDS report/PPT template as the visual target. Keep Agent88 responsible for workflow and automation, not redesigning RDS branding.
Known correction to preserve
am730 counts as three placements.
The reference page has one top leaderboard plus two right-rail rectangles. This is a placement-counting rule for the pilot, not a reason to show an ugly QA contact sheet on the proposal page.
Why this fits RDS
Agency workflow, not generic AI.
RDS runs multi-client, multi-channel campaigns across FMCG, fashion, beauty, and dietary supplements — including award-winning influencer work. The reporting agent is shaped for that cadence, not a generic chatbot.
Multi-client by design
One pipeline, many brand accounts. The agent knows which template, voice, and reporting cadence each client expects.
Platform-agnostic extraction
DV360, Meta, Google Ads, TikTok, influencer dashboards — the same extraction logic handles heterogeneous capture formats.
Private & PDPO-aligned
Runs on your infrastructure. Client campaign data and screenshots never train foundation models. Inference is scoped and logged.
Also in scope · Internal ops
Expense claims, handled in chat.
Same agent pattern. Staff snap a receipt and send it via Telegram or WhatsApp. Fields are extracted, logged to your sheet of choice, and a monthly summary is compiled per person — ready for sign-off.
- Amount, date, vendor, category — extracted from receipt image
- Logged to Google Sheets / Notion — configurable per team
- Monthly summary per staff member — ready for finance sign-off
RDS Claims Agent
Online · Telegram
After the pilot · What this expands into
One agent, then a small team.
Reporting is the wedge. Once the pattern is working inside RDS, the same infrastructure picks up adjacent jobs your ACs and internal ops are already doing by hand.
Expense claim agent
Receipts in chat → extracted fields → logged to sheet → monthly summary per staff member.
Multi-brand client comms copilot
Bilingual drafts for each client in their established voice. AC reviews before anything leaves the firm.
Influencer campaign ops
Brief packs, deliverable tracking, and post-campaign reporting across 90+ creator partnerships.
Pitch deck auto-pull
Agent surfaces matching past case studies and numbers when an AC is drafting a new pitch for a similar brand.
Next step · Pilot
Confirm the Phase 1 pilot inputs and we can start.
Next step: confirm the supplied Drive brief, lock Phase 1 at HK$15k, and agree the post-pilot retainer package. Agent88 then runs the static screencap-reporting workflow: draft report generated, RDS account manager reviews, then approved client delivery.
+852 6444 2088 · clyde@agent88.hk