Case Study · PPC Agency
Twelve Data Sources. One Place to Look.
A UK ads agency logged into twelve different systems to answer one question. We put the lot in one place, with a layer that protects client names, campaigns and numbers before they reach AI.
Follow the transformation- The business
- Ads agency · ~15 employees · a team of five running client ads
- The work
- AI build
- Delivery
- Built and in active use
01 / The starting point
Twelve logins, and hours of work to answer one question.
Google, Shopify, Meta, TikTok, LinkedIn, Amazon, Apple and five more. Twelve places to log in, each one counting things its own way, each one with its own idea of what a click is. Ask a simple question, like how a client did last month. Someone had to open all twelve, copy the numbers into a sheet, and work it out by hand. Every client. Every month.
- Separate systems the client numbers were spread across
- 12
- Built by hand, one tab at a time
- Every report
- Clients ringing the team for numbers they could look up
- Daily
- Client data was being pasted straight into AI tools
- Yes
02 / The method
Understand it. Then build it.
Why we fix the work first
Twelve logins looked like a tool problem. Drawing the work out showed it was a process problem, so we fixed how the work ran before we built the store.
- 01
We drew it
We followed one client report from start to finish. Where each number lived, who looked it up, how often the same answer was worked out again from scratch. It was dozens of times a week.
- 02
We designed it
One store that all twelve systems feed into, counted the same way. A fixed set of rules maps client and campaign names to codes and transforms numeric values using minimum and maximum levels before sending them to AI.
- 03
We built it
Reports that build themselves. Notes from the team on what changed. A saved copy of every number. And a page where clients look things up without ringing anyone.
- 04
It runs
The agency now has its own reporting platform, built around its client data and workflows, with a protection layer for AI. Both the team and clients use it to get answers directly.
03 / What we built
What we put in its place.
- 01
One place for every number
Twelve systems feed into one store, counted the same way, so there is one answer instead of twelve.
- 02
Protected before AI sees it
AI receives mapped names and transformed numbers. When the response comes back, the system restores the original names and values, so users see familiar client data without any extra steps.
- 03
Any AI can ask it questions
The store speaks one common language, so any AI tool the agency picks up can read it without a special build each time.
- 04
Clients look for themselves
Clients see their own numbers whenever they want, with notes from the team saying what moved and why.
- 05
Reports in seconds
A report that used to take hours of copying now builds itself while you wait.
- 06
Every number is kept
The data is saved as it stood on the day. When a client asks what happened in March, there is proof, not memory.
| The work | Before | After |
|---|---|---|
| Places to look | 12 | 1 |
| Building a client report | Hours of copying | Seconds |
| Client data sent to AI firms | Raw | Names coded, values transformed |
| Getting an AI to read it | A build for each tool | One way in |
04 / The result
One place to look, with client names and numbers protected before they reach AI.
- Places to look for a number, down from twelve
- 12 → 1
- Original client names sent to AI
- 0
- To build a report that used to take hours
- Seconds
- Clients answer their own routine questions
- Self-serve
Twelve scattered systems became one store that any AI can read. Routine reports and everyday client questions now answer themselves. Client and campaign names are mapped to codes and numeric values transformed before AI sees them. The original names and values are restored in the response.
“Twelve platforms felt like a tool problem. Drawing the work out showed it was a process problem. Once you can see where the value is really made, the technology is the easy part.”
Sound familiar?
Different business. A familiar pattern?
- Is someone building your client reports by hand out of half a dozen ad accounts?
- Is your team pasting client data straight into AI tools?
- Do clients ring your account managers for numbers they could look up themselves?
- Could one place for all the numbers become something you sell, instead of something you pay for?
Your version starts with a conversation.
A free 15-minute fit call to explore your workflow and discuss the workshop, including pricing.
Start with a free 15-minute fit call.
We’ll discuss the right approach and pricing on the call.