Avoiding the Efficiency Trap

The Jevons Paradox teaches us that as a resource becomes cheaper and more efficient to use, its total consumption increases. Last year we saw cheap access to LLM lead to an explosion of use cases, some profound others deeply trivial, all exposing new risks. This year, we’ve seen agentic reasoning become more expensive but our appetite continues to explode. This makes guard rails doubly important: to protect us against the harms of poor (or malicious) use cases, and to not to embed agentic workflows so deeply into our business that token burn becomes crippling. The defining technical leadership challenge is pretty much as it’s always been: find then use the best tools at the best cost. The Four Types of Work framework terms this as Excellence Work.

Excellence Work is a deliberate investment in how a team builds and operates. By improving developer experience, continuous integration pipelines and observability tools, we systematically reduce the friction of delivery. This proactive Kaizen serves as a force multiplier. It lowers the marginal cost of producing the next software feature.

The Jevons Paradox takes hold the moment this efficiency is realised. As the friction of deployment approaches zero, our ability to generate code skyrockets. The agentic era removes historical constraints, lowering the barrier to churning out code to trivially low. Teams can now generate massive applications and deploy hundreds of features in a matter of hours. Excellence drastically increases throughput, but the efficiency trap is that it does not inevitably reduce total expenditure, simply because of the volume

This overproduction carries a severe financial and operational penalty because every deployed feature creates a long-term contract requiring future maintenance and support. For example, large feature sets increase cognitive load with users so require more customer support collateral, also create context debt that degrades the performance of autonomous agents.

Project Work casts a permanent downstream shadow of routine Core maintenance. Poorly focused Excellence Work can do the same. Whilst our capital expenditure on initial delivery drops due to better tooling, our operational expenditure rises to support a vastly larger software estate. Without governance, the system simply becomes faster at producing waste.

The solution is to set and follow Excellence guardrails such as:

• Set an upper limit for Excellence work and prioritise a backlog to work on only what is needed.

• Run subtractive projects to decommission obsolete software and actively reclaim permanent Core capacity.

• Develop a habit to regularly maintain context repos.

• If autonomous agents aren’t building according to expectations, improve the harness first before committing rework.

Excellence Work unlocks the capability to move quickly. Guardrails will ensure that effort is spent solving the right problems and that input resources are used as efficiently as possible. The stewardship of our resources, and those of others in our supply chain, lies at the heart of the Japanese concept of Mottainai, which is foundational to the Four Types of Work framework. Mottainai transforms raw efficiency into genuine organisational durability.

Download the Four Types of Work framework at https://www.sinharay.tech/free-book.

© Mind Rocket Services Ltd 2026. All rights reserved.

Avi Sinharay

CEng, MIET, MEng, MA (Cantab.)

Fractional CTO, Director/VP of Technology

Core Expertise

Technology Leadership, AI Native Dev, Operating Model Design, Engineering Culture

Domains

Health Tech, Media Tech