EPS Digital case study / content systems

ESTableTennis case study

19–21 hours a week for two articles and two videos. Now a full week on five platforms.

I run the content for ESTableTennis, a table tennis brand I own. I rebuilt how it gets made: AI does the repeat work, I make the calls. No hires.

Before AI 19–21 hrs/wk

Produced 2 articles and 2 videos.

owner-recalled, not time-tracked
Now roughly 13–21 hrs/wk

Produces a full week across five platforms: 7 videos on 5 platforms, carousels, 21 pins, 7 threads and articles.

measured from session logs, includes review and render time

01 / Problem

The bottleneck wasn't ideas

Research, articles, videos, social copy and pins were treated as separate jobs, even when they came from the same insight. Every platform meant starting again.

Repeated setup

Each format reloaded the same context, research and planning from scratch.

Output without a job

More posts only help when each one has a purpose: useful, reach or trust.

Distribution gaps

A finished article or video could still miss its related posts, its channel version or its website embed.

Judgement still matters

Equipment and performance advice needs evidence, a point of view and a person who signs it off.

02 / Cost

19–21 hours bought very little

That time covered planning, two articles, two videos and site upkeep. It left no room for a presence beyond a few posts.

Where the weekly hours went

  • planning2 hrs
  • two blog articles4 hrs
  • at least two video edits8–10 hrs
  • website and seo upkeep5 hrs
owner-recalled, not time-tracked

What that time produced

In April and May 2026 the site published 10 articles: 6 in April, 4 in May. Before the system, I put out fewer than 10 social posts.

articles verified in wordpress social count owner-reported

03 / Opportunity

Make the topic the unit, not the post

A researched topic can feed an article, a short video, a carousel, pins and a thread. Do the thinking once, then adapt it to each channel. Same idea, different native format.

1 topicresearched once
articlesanswer-first, linked
7 videosone master, 5 platforms
carousels, 21 pinsweekly batches
7 threadsconversational

04 / AI

What the AI does

It handles repeatable work: research, outlining, drafting, adapting one idea to each channel, narrating, rendering and scheduling. It never chooses what deserves to exist.

One topic, one weekly cycleA reviewed source asset, native channel outputs, and a better next cycle.
  1. 01 / Human-led strategy

    Choose what deserves attention

    Audience need, search opportunity and business relevance.

    • Topic clusters
    • Useful, reach or trust role
    • Batch planning
  2. 02 / AI production

    Build once, adapt natively

    A strong source asset, then branches into channel formats.

    • Answer-first article
    • Short video and social posts
    • Pinterest pins and carousels
  3. 03 / Human-owned release

    Review, distribute, improve

    Quality gates stay visible. Misses strengthen the process.

    • Check facts, voice, links, format
    • Schedule supported channels
    • Find gaps, update the checks

05 / Human role

What I decide, and where I sign off

I stopped doing the repeat work, not the thinking. Three approval gates sit between an idea and a live post.

Gate A / Before production

Does this topic deserve the work?

I own topic priority, audience intent, the content job and the business path.

Gate B / Before approval

Is this accurate and useful?

I check evidence, advice, voice, links, visuals and channel fit before anything is approved.

Gate C / Before distribution

Can this channel carry it?

I schedule only formats that are stable and supported. Manual steps stay manual.

What stays manual

Strategy and topic choice. Accuracy of equipment and mental-performance advice. Quora answers. Any channel my current plan can't schedule.

What I fixed when it slipped

Some finished videos went live without an article embed. I added an embed-gap check to the weekly routine so it can't repeat quietly.

06 / Implementation

What's under the hood

A set of connected tools, each doing one job, with a notes vault holding the shared memory. You don't need to know any of them to use the result.

01 / memory

Shared knowledge

An obsidian-style markdown vault holds the brand voice, plans, past failures and rules. Every session reads it first.

ai brain
02 / plan and write

Drafting and checks

One model plans, another builds and audits, so no model grades its own work.

claude codecodex
03 / make assets

Articles, voice, images, video

Articles go into wordpress and get scored in rank math. Narration uses a cloned voice. Scene images come from flux. Video is composed from code.

wordpressrank mathvibevoicehugging facehyperframes
04 / render, schedule

Carousels, pins, posting

A headless browser renders carousel slides and pins. Everything is queued across five platforms in one batch.

headless chromemetricool

The pipelines improve themselves, with my approval

Each weekly pipeline (videos, carousels, pins, threads) logs the failure modes it hits, then proposes changes to its own checklist. Nothing changes until I approve it.

approval-gated improvements

Five platforms, one master video

Each daily video goes to TikTok, Instagram Reels, YouTube Shorts, Facebook Reels and Pinterest video pins. One render, adapted per platform.

Make.com scenarios exist in my account but are switched off. They are not part of this system.

07 / Outcome

What changed

Less time per week, more formats, more platforms. The publishing record backs the volume.

Blog articles 10 → 100+

10 articles in April and May 2026, then 100+ in the following couple of months. Roughly 10x the volume.

baseline verified in wordpress 100+ owner-reported
Social posts <10 → 120+

Fewer than 10 social posts before. 120+ platform posts after consistent use.

owner-reported
48

posts published July 14–18, 2026.

project record
53 → 101

posts in the catalog by July 18, 2026.

project record
No hires

One operator ran all of it.

owner-reported

The 10 articles published before the system

verified in wordpress

April / 6

  1. Overcoming Performance Anxiety in Table Tennis: 7 Proven Techniques That Stop the Panic Before It Destroys Your Game
  2. The Best Mindset Strategies to Develop Resilience in Table Tennis in 2026
  3. How to Perform Under Pressure as a Table Tennis Player (And Win More Tight Matches in 2026)
  4. The Best Table Tennis Racket?: Here’s How to Find the Right One and Actually Win More Games in 2026
  5. Table Tennis Psychology: 7 Proven Mental Shifts That Stop You Losing Points in Your Head
  6. Best Table Tennis Table in 2026: Top Picks, Buying Tips, and What I Wish I Knew Sooner

May / 4

  1. Why Do I Play Worse in Matches Than in Practice? 7 Proven Causes That Destroy Your Match Performance
  2. How to Grip a Ping Pong Paddle Correctly: 7 Proven Techniques That Transform Your Control and Eliminate Costly Errors
  3. Best Outdoor Ping Pong Table: Top Picks for Every Backyard and Budget in 2026
  4. How to Deal with a Cheating Opponent in Table Tennis: 5 Proven Strategies That Stop Your Mental Game from Collapsing

How I measured this

Where each number comes from

  • 19–21 hrs/wk before: my recollection, task by task. I didn't track it.
  • roughly 13–21 hrs/wk now: supervised session time from my logs, July to September 2026. It is wall-clock time including review and render waits, not hands-on effort. Videos roughly 8–14 hrs, pins roughly 3, carousels roughly 1–2, threads roughly 1–2.
  • 10 articles, April and May: counted in the WordPress posts list, dates above. The 100+ and the social counts are my own tally.
  • 48 posts and 53 → 101: from the project record, as of July 18, 2026.
  • I make no traffic, ranking, lead or revenue claims here.

Your workflow

Find the repeat work worth fixing

On a free AI Fit Call we look at where AI could take repeat work off your plate, where a person should keep the decision, and what a sensible first step looks like.

Book a free AI Fit Call

One conversation. A clear view of where to start.