The Structure of “Instruction Seeking” at WorkWhy Capable Adults Freeze and How to Build Self-Driving Teams

The Structure of “Instruction Seeking” in the Workplace

Most organizations have teams led by managers who carry formal accountability. In knowledge work, it is common to assign development or operations to an individual, then watch progress stall when bugs appear or dependencies change. Eventually, the manager gathers scattered facts, makes decisions, replans, realigns stakeholders, and protects the deadline.

In many cases, the person is not passive from the start. They can begin tasks and ask questions. Yet the moment uncertainty appears, their communication collapses into “I am stuck,” and the next move disappears from their own control. This article calls that pattern “instruction seeking.”

The hard part is that this pattern is not explained well by raw intelligence alone. It appears among adults with normal learning histories who can perform in short cycles but fail to self-complete tasks over multi-week or multi-month horizons containing unknown variables. The issue sits at the intersection of psychology and organizational design.

The common intuition, “I could study autonomously in school, so why not at work?” follows the same logic. Exams are bounded and stable. Work missions are not. External specs, stakeholders, and constraints keep moving. Plans are hypothesis bundles, and once assumptions break, plans revert to unfinished notes. What is required is not only execution, but assumption updates and redesign loops run by the individual.

Psychological Mechanisms Behind Freezing

1. Self-Efficacy (Bandura, 1977)

Self-efficacy is the belief that one can organize and execute actions required in a specific situation. It affects persistence and strategy adaptation under difficulty. Stajkovic and Luthans (1998) reported a weighted average correlation of r=0.38 between self-efficacy and work performance.

The key is that self-efficacy is not generic confidence. It is a forecast that “I can redesign and move forward when conditions break.” People who go silent under disruption may be trapped less by the problem itself than by the belief that their redesign attempts will fail.

2. Learned Helplessness

Research on learned helplessness, starting from Overmier and Seligman (1967), shows that repeated exposure to uncontrollable aversive events can stop future attempts even when control becomes possible. Maier and Seligman (2016) reframed this with neuroscience: perceived control is something learned, not default.

In workplaces, if supervisors repeatedly rescue every breakdown, individuals fail to accumulate evidence that their own intervention can change outcomes. Behavioral shutdown then becomes a rational adaptation.

3. Motivation Quality (Self-Determination Theory)

Self-determination theory argues that autonomy, competence, and relatedness sustain internalized motivation. Gagné and Deci (2005) summarized how autonomy-supportive environments improve workplace motivation and performance.

Autonomy here does not mean neglect. It means purpose and decision latitude are aligned. When tasks are assigned without authority, or accountability is imposed without control, instruction-seeking behavior is reinforced.

4. Personality Differences Matter, but Do Not Determine Fate

Personality traits show moderate heritability in meta-analytic evidence (Polderman et al., 2015). Research also links conscientiousness to job performance across roles (Barrick & Mount, 1991; Hurtz & Donovan, 2000).

But these are population-level tendencies, not destiny for an individual. Traits influence outcomes, yet environment design and training still leave substantial room for change.

How Organizations Accidentally Reinforce Dependency

Instruction seeking becomes chronic because organizations reward it in the short term. A member escalates a problem without options or recommendation. A manager jumps in, triages, replans, and saves the project. Everyone survives this cycle, so both sides learn the wrong lesson.

• Member learns: “Escalate when stuck, someone else will decide.”
• Manager learns: “If I do not intervene immediately, this will burn.”

Daily micro-goal cycles can make this worse. If work is constantly sliced into next-24-hour tasks, people can execute near-term assignments but fail to learn 30-day planning, risk architecture, and cross-team negotiation. Goal-setting research (Locke & Latham, 2002) supports specific, challenging goals, but goals alone are insufficient; strategy generation must be learned by the owner.

Psychological safety is often misread as comfort. Edmondson (1999) defined it as interpersonal safety for speaking up, not exemption from standards. Safety and accountability are a paired system. Safety without standards increases problem reporting but not ownership. Standards without safety increase concealment and delay warning signals.

A frequently missed metric is boundary spanning: obtaining information, negotiating constraints, and aligning non-team stakeholders. Many people do not freeze on technical complexity alone; they freeze when coordination friction appears.

Job Design for Self-Driving Execution: Mission, Authority, Guardrails

The fastest path is design change, not personality debate. Convert knowledge-work tasks from “a list of activities” into “an outcome contract.” Define 30- to 60-day outcomes first, agree on completion criteria in writing, and specify authority boundaries.

Never hand over responsibility without decision rights. When responsibility and authority are split, paralysis is a predictable result.

Then add guardrails. Higher autonomy exposes uncertainty, which triggers managerial fear and over-intervention. That intervention recreates dependency. Checklists and pre-failure design help break this loop. Gawande et al. (2009) demonstrated in high-complexity settings that structured checklists reduce omission risk.

Premortems (Klein, 2007) are also effective: assume failure has already happened, enumerate causes, and place countermeasures in advance. For people who freeze at disruption time, prebuilt branches work better than real-time improvisation.

Operationally, design “no instant answer” windows. Except for critical incidents, do not answer escalation immediately. Require a second report after two hours in a fixed format, and reject escalations with no options. That window is not abandonment; it is option-generation training.

Development Protocol: Change Reporting Format, Change Thinking

Behavior changes faster through format than through motivation speeches. Convert reports from raw facts into decision material.

Use a minimum five-point template:

1. Current state
2. Impact
3. Options (at least two)
4. Recommendation (one)
5. Next deadline

Do not punish a wrong recommendation. Do not accept no recommendation. That single boundary operationalizes safety plus accountability.

Rotating daily-meeting facilitation can help if scoped correctly. It teaches people to read dependencies and turns status meetings into planning meetings. But do not force facilitators to own final decisions. Their role is to standardize questions, collect five-point inputs, and hand decision packets to decision-makers.

To sustain 30-day execution, keep four lightweight artifacts continuously updated:

• Weekly plan update
• Risk register
• Decision log
• Stakeholder map

These can be short. What matters is that updates are scheduled as real work.

Implementation intentions (if-then plans) are another lever. Gollwitzer and Sheeran (2006) reported medium-to-large effects on goal attainment (d=0.65). In multi-week work, surprise is normal. If-then plans convert surprise into prepared branches.

In the AI Agent Era: Dependency Mode vs Thinking Mode

Since 2025, AI has evolved from chat tools into agents that execute tasks through tool use. Major vendors now frame agents around capability plus guardrails, and by 2026 enterprise discussions have shifted to permissioning, evaluation, and context-sharing across multiple agents.

The practical fork is clear: using AI as a secretary tends to increase dependency; using AI as a coach can increase autonomy.

A productive operating model includes four rules:

1. Let AI decompose plans, but keep final decisions with humans.
2. Use AI to widen premortem failure candidates before team selection.
3. Use AI to check report-template completeness, not to replace judgment.
4. Use AI to generate many if-then plans, then let humans adopt only selected ones.

Authority design is critical. The more external write-access agents have, the larger the blast radius of mistakes. The act of defining what AI may and may not do becomes a direct test of organizational maturity.

Hiring for Self-Drive: Measure Evidence, Not Emotional Impression

Fixing chronic instruction seeking after hiring is costly. It is more rational to screen for multi-week planning and disruption-handling patterns upfront.

Meta-analytic selection research (Schmidt & Hunter, 1998; Schmidt, Oh, & Shaffer, 2016) shows that method choice materially affects performance prediction and productivity outcomes. Combinations of cognitive measures, work samples, conscientiousness-related indicators, and structured interviews outperform unstructured impression interviews.

In practice, separate interviews into two lanes:

• Past behavior evidence
• Situational work sample

For the first lane, ask for a 30+ day project where assumptions collapsed midstream and probe for impact slicing, stakeholder mapping, option comparison, deadline renegotiation, and recurrence prevention. For the second lane, provide a broken-spec scenario and ask candidates to produce a five-point decision memo in 20 minutes.

The score is not correctness. It is speed and coherence of decision-material generation.

Personality tests can be supplementary, not definitive. Writing-based work samples are usually more discriminative than interview charm.

When Change Still Fails: Placement and Boundaries

Even with better development and hiring, some instruction-seeking patterns remain. At that point, use role design boundaries rather than moral judgment.

• Roles requiring self-drive: exploration, negotiation, architecture, prioritization.
• Roles where procedural reliability creates value: routine operations, monitoring, repetitive execution.

Too much discretion in the latter burns people out. Too little autonomy in the former collapses outcomes. Role confusion is what turns managers into permanent rescuers.

Manager psychology also matters. Escalation intervention is often a loss-avoidance response, consistent with Prospect Theory (Kahneman & Tversky, 1979). That is exactly why systems, not heroic willpower, are needed: reporting formats, authority design, checklists, premortems, and if-then plans.

Conclusion: As Support Increases, Ownership Becomes Scarcer

Instruction seeking is not just an individual defect. It is a learned product of system design. Teams can reduce multi-week freezing by intentionally building self-efficacy loops, preventing helplessness loops, pairing safety with accountability, teaching strategy generation, and upgrading hiring methods.

AI agents can either harden dependency or amplify autonomy. If AI is used only as a secretary, human strategy muscles atrophy. If AI is used as a coach for option generation, failure anticipation, and structured reporting, self-driving capacity grows.

One hard market fact remains: value is created not by listing problems, but by presenting decisions that move work forward inside uncertainty. As support tools proliferate, that capability becomes rarer, and its reward premium grows.

Key References

• Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change.
• Stajkovic, A. D., & Luthans, F. (1998). Self-efficacy and work-related performance: A meta-analysis.
• Overmier, J. B., & Seligman, M. E. P. (1967). Effects of inescapable shock.
• Maier, S. F., & Seligman, M. E. P. (2016). Learned helplessness at fifty.
• Gagné, M., & Deci, E. L. (2005). Self-determination theory and work motivation.
• Polderman, T. J. C., et al. (2015). Meta-analysis of the heritability of human traits.
• Barrick, M. R., & Mount, M. K. (1991). The Big Five personality dimensions and job performance.
• Hurtz, G. M., & Donovan, J. J. (2000). Personality and job performance.
• Locke, E. A., & Latham, G. P. (2002). Goal-setting and task motivation.
• Edmondson, A. (1999). Psychological safety and learning behavior in work teams.
• Gawande, A., et al. (2009). A surgical safety checklist to reduce morbidity and mortality.
• Klein, G. (2007). Performing a project premortem.
• Gollwitzer, P. M., & Sheeran, P. (2006). Implementation intentions and goal achievement.
• Schmidt, F. L., & Hunter, J. E. (1998). The validity and utility of selection methods.
• Schmidt, F. L., Oh, I.-S., & Shaffer, J. A. (2016). Personnel selection methods and productivity.
• Kahneman, D., & Tversky, A. (1979). Prospect theory: An analysis of decision under risk.