Everyone is talking about what AI will replace. Let's talk about what it won't.

The Remote Labor Index, a benchmark developed by the Center for AI Safety and Scale AI, tested the best AI agents on the market against real-world, economically valuable projects. The kind of work your teams do every day. The result? The highest-performing agent achieved an automation rate of just 2.5%.

Not 25%. Not 50%. Two and a half percent.

That number should change how you think about your entire AI strategy. Because the real opportunity isn't in the 2.5% that AI can automate. It's in the 97.5% it can amplify.

Automation Is the Wrong Question

Most boardroom AI discussions start and end with "what can we automate?" That focuses your budget and strategy on a sliver of the picture.

The far bigger question: how do we make the other 97.5% of our workforce dramatically more effective?

Think about what your most valuable people actually do. They make creative judgements. They navigate ambiguity. They solve problems without a template. What slows them down isn't a lack of talent. It's everything around the talent: searching for assets across disconnected systems, re-doing work because upstream specs changed, manually checking outputs against brand guidelines, coordinating handoffs between teams and tools.

That operational drag is where AI delivers its real value. Not replacing the person. Removing the friction.

The Numbers Everyone Gets Wrong

Here is where most AI investments go sideways. Organisations focus on GPU pricing and infrastructure benchmarks. They measure what AI costs per hour.

But the real cost is iterations. Rework. Failures. A render at $2.80 per GPU hour looks more expensive than one at $2.20. But if the cheaper option requires 6.8 iterations on average while the more expensive one needs only 3.9 (because it understands your production requirements), the actual cost per finished deliverable drops by 28%.

The metric that matters isn't cost per GPU hour. It's cost per finished output.

The Data Gap Nobody Wants to Talk About

I see this first-hand in media and entertainment. The industry is moving fast on AI adoption, but the data to support purchasing decisions, infrastructure planning, and go-to-market positioning is almost entirely absent. Studios don't know what hardware to buy. Hardware vendors don't know which workflows to target. And everyone is overpromising what AI can actually do today.

What we're seeing in practice, on real productions with real deadlines, is that AI-augmented workflows deliver 5 to 10% cost and time savings. Not the 50% or 80% figures that slide decks promise. But 5 to 10% compounding across every project, every quarter, with the numbers improving as the system learns. That's not a headline number. It's a real one.

How Catalyst Makes This Work

Standard workflow tools understand infrastructure: containers, scheduling, resource allocation. What they don't understand is whether the output meets your brand standards, passes regulatory review, or aligns with creative intent.

That gap is where organisations lose money. And it's exactly why we built Catalyst.

Catalyst is a technology accelerator built on Microsoft Azure that understands the context and semantics of your content. It works across your existing environment, whether that's Azure, AWS, Google Cloud, or on-premises, and closes the gap between what your infrastructure can process and what your production actually needs. Because it understands your assets, your specifications, and your creative intent, it catches problems earlier, reduces unnecessary iterations, and learns from every execution.

That's how you move from 6.8 iterations to 3.9. That's how 5 to 10% savings compound. And that's how you amplify the 97.5% of work AI cannot automate.

The Question Worth Asking

The organisations that win won't be the ones chasing full automation. They'll be the ones that figured out how to amplify everything else.

We'd welcome the chance to show you where those returns are hiding in your operation.

Book a Catalyst Discovery Workshop with Support Partners. In less than 90 minutes, we'll map one of your high-impact workflows, show you where the iteration waste sits, and design a proof-of-concept with measurable outcomes on the platform you already own.

Contact us to schedule your workshop with "97.5% Workshop" in the subject line and we will book a discovery session with you.