Case Study · AI with Claude
AI-Powered Campaign Report Automation Case Study
An AI reporting app, with Claude as the engine, turns Adverity exports into a finished, branded campaign report.
The result: a campaign report that took three to five hours is ready in a few minutes, consistent regardless of who produces it, with Claude writing the optimizations and recommendations.
- Before
- 3 to 5 hours
- per report: manual transcription into a PPT
- After
- A few minutes
- per report: generated with Claude, then reviewed
The Customer
A media agency trading team
The agency's trading team produces mid-campaign and post-campaign reports for every client and campaign it runs.
At a glance
- Customer
- Media agency trading team
- Input
- Adverity exports, uploaded directly or from Google Drive
- Output
- Branded mid-campaign or post-campaign PPT or PDF
- AI engine
- Claude, through Anthropic's API
- Built on
- Supabase, AWS and Inngest
- AI pattern
- Prompt Chaining
The Problem
Every report was built by hand
Building a mid-campaign or post-campaign report at the agency takes a user three to five hours of manual work. Figures are read out of platform exports and retyped into a PPT, charts are pasted in as images, and the written analysis is composed from scratch every time. The work is slow, repetitive and prone to error, and it grows in direct proportion to the number of campaigns.
Current process
During and after each campaign, a user pulls one Adverity export per channel, opens each workbook in Excel and reads off the totals and breakdowns. Those figures are retyped into a PPT template, one slide at a time. Charts are built in Excel and pasted in as images rather than native objects. The user then writes the optimizations and recommendations from memory of how the campaign ran. A senior reviews the finished file over email or chat before the PPT goes to the client.
Pain points
Each report consumes three to five hours, repeated across every client and campaign. Manual transcription from export to slide introduces errors with no way to detect them, and charts pasted as images cannot be corrected without rebuilding them. Metric selection is inconsistent across channels, and the quality of the written analysis depends on the author rather than the data. Review happens informally, leaving no record of what was approved or by whom.
The AI Solution
AI report automation with Claude as the engine
The solution is a web application, with Claude as the engine, that turns Adverity exports into a finished, branded mid-campaign or post-campaign PPT. A user uploads the exports, and the app parses them, computes the performance verdicts, builds native PPT charts and uses Claude to write optimizations and recommendations grounded in each campaign's own figures. A report that took three to five hours is ready in a few minutes, and the output is consistent regardless of who produces it. Users now review a finished report instead of building one by hand.
How the AI reporting app works
It is an internal reporting tool for the agency's trading team. Users upload one or more Adverity exports, directly or from Google Drive, and choose which campaigns from the upload go into the report. The app handles multiple channels and selects the correct metric vocabulary for each. Users preview the PPT in the app and edit it through a chat with Claude, then download it as a PPT or PDF, in the agency's branding or on the client's own template. Reports can be routed to a senior for review before they go out, and admins can see what each report costs to generate, along with a usage and audit view of who generated what. The application runs on Supabase, AWS and Inngest and follows a Prompt Chaining pattern: exports are parsed and analysed in code, Claude writes the insights in a single call through Anthropic's API, and the PPT is built in code, with retries, data checks and human review around it.
- Parse and analyseIn code
- Write the insightsClaude, in a single call
- Build the PPTIn code
Retries, data checks and human review around it
See the AI in Action
1. New report.
Accepts Adverity exports uploaded directly or imported from Google Drive.
2. Reports.
Reduces a three to five hour manual task to a few minutes.
3. AI spend usage.
Shows admins what each report costs to generate, with a usage and audit view of who generated what.
4. Activity and edits with Claude.
Previews the PPT in the app and lets users edit it through a chat with Claude.
5. Benchmarks.
Computes performance verdicts against plan targets.
More key features of the AI reporting app
Lets users pick which campaigns from an upload go into the report
Selects the correct metric vocabulary for each channel automatically
Generates native PPT charts rather than pasted images
Uses Claude to write optimizations and recommendations grounded in each campaign's own figures
Downloads as PPT or PDF, in the agency's branding or on the client's own template
Routes reports to a senior for review before they go to the client
Get In Touch