One Concept, One Example, One Question

The concept

When work is under pressure, my first instinct is to ask for more resource, but what would that help actually be able to take on?

Capacity isn't just a headcount or a number of hours, its usefulness depends on how well it matches the work. A person or tool may be able to carry out the repeatable parts yet lack the context or authority to own the result. If everything comes back for review, the original workload has only changed form. That's what I mean by saying capacity has a shape.

This is especially visible when AI is proposed as extra capacity. A model can compare records and draft a note quickly, but it doesn't know which discrepancy carries business risk and can't accept a trade-off on the team's behalf. Another person faces a similar limit when the relevant context or authority remains with someone else.

Both can help, but the task may need reshaping first. The useful diagnosis starts by separating a shortage of execution capacity from work held up by missing context or decision authority. Sometimes the real bottleneck is review, because someone still has to check what comes back. Naming the bottleneck makes it easier to decide what can move and what must stay with the person who holds the judgement.

The example

Imagine a team trying to finish a data migration before a fixed deadline.

At first the problem looks like volume. The mapping backlog is growing, but only a few people know both systems well enough to recognise a dangerous mismatch. The team asks whether another analyst can help, and on paper the answer looks simple: more hands should mean more progress.

The work separates when they look more closely though. A prepared field mapping checked against agreed rules can be handed over, but other cases turn on business meaning, so the experienced person has to interpret them. Decisions about whether a messy historic value is acceptable stay with whoever has the authority to accept the risk.

If the analyst receives a share of the backlog without those boundaries, the senior person still has to explain the ambiguous cases and recheck the results. The nominal handover creates a second queue for the same person.

A better split is narrower. The team defines a set of low-risk mappings with clear rules and examples. The new analyst works through that set and records exceptions in a fixed format, while the experienced person handles cases that depend on policy or business meaning and samples the completed work rather than reviewing every row.

AI can work inside the same boundary by producing a first comparison note and flagging inconsistent mappings for review. It doesn't approve a mapping or decide whether a historic value is acceptable. The original bottleneck now receives fewer, better-defined cases instead of every row.

The question

Ask yourself:

What would extra capacity need to own to release this bottleneck?

If the answer is vague, adding a person or a tool is likely to create another handoff for the same people to manage.

Elsewhere this week…

• I published AI Autonomy Needs Boring Operations, on the operating controls that make AI autonomy dependable: clear authority, evidence, recovery, and ownership.

• The SecondBrain Framework now has a versioned behaviour-contract inventory and validator, so its role boundaries, content lifecycles, privacy, automation, storage, and test assumptions can be checked for drift before release. The update also clarifies DTM hand-off and support for vaults outside iCloud.

Support

Everything I publish here is free to read, and always will be.

I maintain this newsletter and my website in my own time. If something here has been useful to you, and you'd like to help cover the costs behind that work, you can support me through Ko-fi.

Generative AI is a powerful tool. I openly use it to conduct research, challenge my thinking, generate drafts, and review my work. AI never owns what I think, or my personal judgement. Everything I publish is a reflection of my own thoughts and opinions.