Agentifying operations users could do directly removes space for skill expression

What it is

Some operations are the point of participation: hand-grading a photo, setting type, steering a character, writing the sentence yourself. Handing these—tasks users can already do and care about doing—to an agent saves operation time at the cost of skill in use and expression. For someone who wants to grade their own photo, "one-tap enhance" is not a saving; it is a removal.

Why it happens

An operation is a means for some users and an end for others. Agentification compresses "process-as-output" tasks—creating, tuning, competing—into "accepting results," closing the channel through which users express intent via parameter choices and technique. Output homogenizes, and the learning path is cut: without doing the work by hand, skill stops growing, and the tool loses value for its most invested users.

Studying it

Compare agent-only, tool-only, and agent-with-expandable-manual forms on creative tasks, measuring output diversity, satisfaction, and later unaided performance. Process data show whether expert users route around the agent; interviews separate "I want it finished" from "I want to make it" as drivers of mode choice.

Where it stops holding

The claim applies to operational tasks users care about; for pure procedural work (forms, batch conversion) agentification costs nothing and only adds convenience. Deprivation scales with skill: novices may welcome the agent as a safety net, while experts value the manual path even when they mostly delegate. Keeping manual channels has maintenance costs, and not every product should pay them.

Applying it

• Identify "process-as-output" tasks in the product—creation, tuning, competition, writing—and keep a full manual channel for each.
• Never let the agent entry replace the tool entry: keep parameterized manual modes alongside one-tap features, and allow entering manual fine-tuning from an agent result.
• Test both channels with expert users: can they reach manual tools without passing the agent, and is the tuning granularity fine enough?
• Watch manual-versus-agent usage trends; if manual use approaches zero, check whether the entry is buried rather than the need gone.

Related

• Same group: B4.16.1 In instrumental interaction the user controls each step; in agentive interaction the user delegates a goal · B4.16.2 The two modes demand different feedback: tools need immediate response, agents need visible process · B4.16.3 Delegation transfers control and responsibility together; its boundary must be stated before the handoff · B4.16.4 The two modes can coexist in one product, but switching should be the user's decision, not the system's
• Nearby: B4.10 Direct manipulation · W1 Game interaction and skill expression
• Search terms: skill expression · automation paradox · creative control