A Freelance Ads Consultant Measured What AI Saves Him: Lionel Fenestraz and 14 Hours a Week

Running paid advertising for a client involves far more than setting up campaigns. There are briefs to write, search terms to sort, product descriptions to optimize, audits to run and a report to deliver every month. For a consultant working alone, those hours decide how many clients one person can serve.

Many people claim AI saves time. Lionel Fenestraz measured it, task by task, in his own work.

This article is for marketers, and for anyone with a professional skill who is thinking about working for clients on their own. We look at how he uses AI, what he measured, and then, in a separate section, at what his experience still has to do.

Who he is

Lionel Fenestraz is a freelance pay-per-click and conversion-rate consultant based in Alicante, Spain. He has more than seven years of experience managing Google Ads and Meta Ads for vacation rental, B2B and e-commerce clients, and works in Spanish, English and French. His website tells clients that they work directly with him, from the first call to the monthly report.

How he uses AI

In a guide published on his website in June 2026, he describes three levels of using Claude:

• Chat for one-off tasks. Drafting a difficult email, translating a landing page or generating copy ideas, always with a detailed setup prompt.
• One Project per client. He keeps a Claude Project for each active client, plus one for his own work, with instructions and files such as client documents, his templates and examples of his writing, so every conversation starts with the client's context.
• Claude Code for recurring work. He uses Claude Code to automate recurring workflows; he describes the API as the way to automate tasks such as processing Google Ads exports.

The five tasks he relies on most are:

1. A first brief for a new client, drafted from the client's website and his own template
2. Google Ads search term analysis: sorting queries by intent, finding negative keywords and generating ad copy ideas
3. Meta descriptions and title tags for large product catalogs, checked by hand on the top products
4. The first draft of monthly reports, to which he adds the strategic observations himself
5. Technical SEO audits that prioritize what to fix, alongside a conventional site crawl

He also keeps a prompt library: every prompt that works is saved with an example input, an example output and notes on when not to use it.

What he measured

He tracked his time from July 2025 to January 2026 while using Claude Projects and Claude Code. By his account:

Task	Before	After
New client brief	90 minutes	30 minutes
Monthly report per account	4 hours	1.5 hours
Search term analysis	2.5 hours a month	45 minutes a month
Initial technical SEO audit	6 hours	2.5 hours
Technical email replies	30 minutes a day	10 minutes a day

In total he estimates a saving of about 14 hours a week. He used that time to take on three new clients in six months without hiring, to write his own research and to give existing clients deeper analysis. He estimates that the 20 to 30 hours he spent setting everything up paid for itself in two to three weeks.

Where he says AI fails

He is unusually specific about the limits:

• Recent numbers. Asked for a current industry statistic, Claude may invent a plausible figure, so he never asks for statistics without supplying the source.
• Large calculations. Across a spreadsheet of many campaigns, it can get a few cells wrong, so he does exact calculations in code rather than in chat.
• Missing business context. It does not know what happened to a client last quarter unless he tells it.
• Generic voice. Without examples of the voice he wants, the writing comes out generic.

His rule is to use AI for drafts, ideas and repetitive operations, and to read every important email, monthly report and campaign brief end to end before sending it. He also advises anonymizing personal data before sharing it with a model and telling clients in writing how AI is used.

What experience gave him that AI could not

This section is our analysis, not a claim made by Fenestraz.

• Knowing which tasks to hand over. He chose tasks with structured input, predictable output and regular repetition. That judgment comes from years of doing the work by hand.
• Knowing what a good report says. The AI drafts the narrative; the strategic observations, which depend on a client's history, come from him.
• Knowing how AI goes wrong. His list of failures, and his notes on when each prompt should not be used, are the kind of caution only experience produces.
• Turning time into growth. He measured the savings and decided what they were for: more clients and deeper work, not just shorter days.

What you can copy

These are our suggestions, drawn from his story.

1. Give each client a workspace. A project with the client's documents and your templates means you stop repeating background in every request.
2. Save the prompts that work. Write down the prompt, an example input and output, and when not to use it.
3. Measure before and after. A task-by-task table is more convincing, to you and to clients, than a general claim.
4. Keep calculations and decisions out of chat. Check numbers in code, and read everything before it goes to a client.
5. Protect client data. Remove personal information before sharing data with a model, and tell clients how you use AI.

Notes on the sources

All facts in this article come from Fenestraz's own website: a guide published on June 15, 2026, and his About page. The time savings are his own measurements of his own work and have not been independently verified. His website's legal notice lists the site as owned by him personally.

Sources

• Claude for digital marketing consultants: the 2026 guide, Lionel Fenestraz, June 15, 2026
• About, lionelz.com