We Blew £30K on AI Workflows Trying to Scale Creative Ops

D2C Diaries is proudly sponsored by Wayflyer

Running an ecom brand and struggling with cash flow? Whether you're scaling ad spend or restocking bestsellers ahead of peak season, Wayflyer provides fast, non-dilutive funding tailored to e-commerce businesses. No equity. No months of waiting. Just the capital and insights you need to grow, on your terms.

Check them out: partners.wayflyer.com/soar


In this episode, Olly sits down with Soar COO Sam to unpack the hard lessons from their early experiments with AI.

After spending over £30K on tools, workflows, and automation frameworks that quickly became obsolete, the team realised they were solving the wrong problem.

This episode dives into the strategic shift that followed, from prompt engineering to context engineering, from static workflows to scalable infrastructure.

You’ll learn:
 – What context engineering actually looks like inside a creative team
 – The difference between business-level vs process-level context
 – How to build agentic systems that don’t collapse under real-world complexity
 – How Soar’s internal AI Committee drives bottom-up adoption across departments

If you're building with AI inside a fast-moving team, this is the blueprint we wish we had six months ago.

Watch Andrej Karpathy’s LLM Breakdown Video: https://youtu.be/EWvNQjAaOHw
Comment on our Youtube "AI" to receive Sam's run-through on how to 10x creative strategy by using AI-powered context engineering

00:00 Introducing Sam and his role with AI 
07:10 Where our AI systems broke (and why)
11:10 Business-level vs process-level context
14:15 Codifying human judgment into systems
15:30 Flashy tools vs infrastructure that scales
17:45 Our internal AI committee
28:50 Slope vs intercept thinking in AI builds
32:30 Data warehousing
39:35 Creating UGC briefs with MCP
52:20 Setting brands up for the next 6 months
1:00:00 The S-tier tools we're actually using