AI Era Power Shift: Decision Rights to Frontline Nodes, Governance to Core

The article opens with a concrete scenario: an equipment manufacturer uses AI to aggregate fault data, contracts, engineer schedules, and spare parts to suggest dispatch decisions. Previously, a service manager needed multiple calls to clarify the situation; now the information appears instantly. However, a field lead notices a small-revenue client with a downed machine is deprioritized behind a high-revenue client whose equipment is still running. The system sorts by a fixed "customer value" standard; changing the order requires an exception request. The author observes that while information preparation accelerates, the decision wait remains — only now one must find who can explain the sorting logic, approve exceptions, or modify the standard. Visible approval steps may decrease, but power moves into less visible places.

AI Redistributes Knowledge and Work, Making an Old Problem Concrete

When more people can propose solutions and more actions execute automatically, who decides what gets priority and who can change the shared arrangement? The article cites a 2024 field experiment at Procter & Gamble published in Organization Science (June 2026): 791 professionals worked individually or in pairs on product innovation tasks with or without AI assistance. AI-assisted individuals produced solutions matching the quality of unaided two-person teams, and AI helped cross-functional participants create proposals balancing technical and commercial concerns. This shows certain cross-disciplinary analysis capabilities can reach individuals more easily. However, the experiment did not test changes in budget or hiring authority, nor does it prove professional teams lose value. Moving from generating better proposals to committing the enterprise still requires authorization, resources, and accountability.

Scarcity Must Be Re-identified

An analysis can be regenerated infinitely, but an on-site engineer, a batch of spare parts, or a client's trust cannot necessarily scale. Those who control such scarce conditions may gain influence, yet scarcity alone does not constitute power: it depends on whether others depend on it, whether substitutes exist, and whether the holder decides who may use it. A technician's ability to change system configuration gives operational influence, but whether the company authorizes them to alter customer priority is a separate governance question. Professional capability, actual influence, and formal decision rights must be distinguished; otherwise an operational permission may silently shoulder strategic trade-offs that belong in open debate.

AI Can Also Strengthen Central Control

When leadership gains direct access to more complete business information and can compare more options, decisions previously delegated may be recentralized. Those holding budget, hiring, and policy-adoption authority can also steer technology investment and permission scopes, potentially reinforcing existing power. Technology simultaneously expands both frontline autonomy and central control possibilities. The "mycelium organization" model chooses to delegate more rule-based judgments to capable frontline people while building the shared conditions that support those judgments. Whether an organization has made this shift is visible in actual authority and work practices.

When Information Becomes Easily Accessible, the Basis of Power Changes

A person influences organizational decisions through formal authority, unique knowledge or resources, trusted coordination relationships, or risk-acceptance judgment. In the past, a manager who knew every engineer's capability, the client's real pain points, and which department would cooperate held coordinating value. If that information now reliably resides in shared systems and routine work can be arranged within rules, influence maintained merely by passing information and approving standard items needs a new foundation.

Deciding What Is Worth Doing and Influencing More Actions

As option generation accelerates, the organizational challenge shifts from lacking options to allocating limited attention and resources among abundant options. The equipment firm can ask AI for more dispatch plans, but cannot give all clients the same expert simultaneously. Whether to honor contract commitments, reduce downtime losses, or protect key accounts still requires human judgment. These goals sometimes align, sometimes conflict. Goal-setting has always been a core management power; AI changes how goals enter action. A vague requirement once needed a manager's interpretation and could be adjusted during execution; when the same goal becomes a system sorting rule, it continuously affects thousands of tasks. A small weight change alters what many people work on, wait for, or leave unaddressed long-term.

Traditional business systems already sort and dispatch by preset conditions. AI adds the ability to read unstructured repair descriptions, infer urgency, and generate handling suggestions. Once the enterprise allows these judgments to directly affect dispatch, the configurator's choices — which data to feed, what the model should prioritize, how much scope to automate — shape the interpretive space previously left to humans. AI lowers the cost of bringing more situations into automated handling, making the applicability boundaries of these settings more critical to surface.

Rules Enter the System; Trade-offs Made During Rule-Making Get Repeatedly Executed

If customer value is measured mainly by historical revenue, smaller-revenue clients are structurally disadvantaged. If after-sales performance is measured by ticket closure speed, engineers may gravitate toward easily closed tickets. When such goals are embedded in AI-involved workflows, they influence more concrete actions; whether the goals cover what the enterprise truly values requires separate judgment. This gives those who define data definitions and evaluation standards more consequential influence. For example, "equipment restored" could mean a simple restart or stable operation under agreed conditions for a period. The chosen definition affects when a task ends, whose workload increases, and who owns subsequent failures. Defining a field can already be allocating responsibility.

This does not mean metric-setters can arbitrarily manipulate the organization. Most standards exist to make complex work comparable and coordinable. The problem is that every standard has applicability conditions, and users may not see what was omitted. If business leaders only watch final scores, technical teams only implement specs, and the frontline has no correction channel, a local assumption can gradually harden into an unquestioned fact. Power analysis must therefore step back from "who signs last" to see who defines the option set and who sets the evaluation yardstick. The person who makes a particular choice always appear most rational also steers organizational direction.

This influence must enter explicit governance responsibility. Business leaders propose goals, professionals explain metric applicability, system owners implement configurations, and affected frontline people can raise counter-examples. Important rules are formally adopted, interpreted, and revised by authorized people. Shared rules should make trade-offs visible and prevent system configurators from changing business principles solely through technical permissions.

Give Action Rights to Nodes, and Send Support to the Frontline

The article advocates "strong core, active frontline": the core maintains shared direction, professional standards, risk baselines, and public capabilities; the frontline responds to customers and field conditions. In customer tasks, the node bearing business decisions and outcome responsibility is the person or team that can perceive problems, judge, connect needed capabilities, and accept corresponding accountability. One person can keep their professional role while taking different task roles with corresponding permissions.

This authority must be formally established. Qualified business leaders, based on professional assessment, confirm the task owner's judgment capacity, specify decidable matters, resource scope, and exception boundaries, and enable relevant departments to deliver promised support. Authorization adjusts as tasks end or conditions change. Business decision rights should move toward such nodes. If the field lead can judge service urgency, they should get matching permissions: adjust dispatch within agreed service standards, budget, and staffing; choose handling approaches; and make deliverable commitments to the client — without climbing the hierarchy for every in-scope decision. Cross-boundary conflicts go to those with coordinating authority.

Having the right to choose a solution also requires the conditions to obtain the support that solution needs. Resource calling changes: professional departments maintain personnel capabilities and long-term development; task owners obtain services and coordinate schedules via open support rules. Rule-compliant support should be delivered directly; when an engineer's time is already committed, a designated coordinator adjudicates priority and arranges remediation for the affected client. Tasks cannot arbitrarily pull resources; departments cannot refuse agreed support just because "the person reports to me."

Outcome responsibility becomes concrete: task owners own complete client problem resolution and coordinate follow-up corrections; professionals own their technical judgments; resource providers honor committed support; system owners explain digital capability and permission configuration consequences. Power and responsibility land on respective parties so the organization does not pile all consequences on the last person who clicked "confirm."

Shared Rules Must Be Trustworthy

Political scientist Elinor Ostrom's 2002 work on long-lasting common-pool resource governance (irrigation, fisheries) offers inspiration: most people affected by daily operating rules can participate in modifying them; monitors are accountable to resource users; low-cost conflict-resolution channels exist. These are not direct templates for enterprise management but suggest shared order can be sustained through participation, monitoring, and error correction.

The mycelium organization's "protocols" are formally adopted shared agreements for collaboration: who decides what, which standards apply, how disagreements are handled. "Protocol as authority" demands these agreements be consistently and fairly honored, binding rule-makers and interpreters alike. Managers cannot demand employees respect authorization boundaries while personally overriding boundary-respecting decisions on a whim.

The frontline must also gain the right to drive correction: submit facts to correct erroneous records; escalate to authorized handlers when situations exceed rule presets; initiate rule review when similar exceptions recur. The organization must clarify who receives, response timelines, who decides changes, so dissenters know how their input is processed. Professionals hitting stop conditions should be able to halt relevant actions per their authorization.

Raising objections does not mean everyone can rewrite rules; bearing outcomes does not mean replacing all professional judgments. The need is for different powers to mutually support and constrain: frontline gains real action space, public capabilities provide reliable support, professional judgment guards baselines, new facts can change shared arrangements. The core continuously corrects; the frontline no longer depends on occasional favors.

This also constrains system owners. Maintaining models, data, and permissions matters, but system owners must declare capability boundaries and configuration consequences, not dismiss professional dissent as "not understanding technology," nor attribute system-caused problems solely to the employee who clicked confirm.

Senior Leadership's Power Center Shifts to Direction and Shared Capabilities

When business nodes gain action rights, management work must change. Someone still decides what the enterprise does and does not do, which capabilities warrant long-term investment, which cross-task conflicts require unified trade-offs, and how much risk the organization can bear. These decisions cannot emerge from each node optimizing locally. These governance duties belong to senior leadership and formally authorized rule, professional, and resource owners; critical trade-offs need a final accountable person. Senior power should concentrate on these issues. Time once spent re-approving dispatches should shift to defining service commitments, building expert teams, arranging public resources, and revising failed standards. Leaders bear overall responsibility, so they must create conditions for frontline action and own the consequences of the rules they set.

For middle management, new value lies in defining complex problems, combining cross-department capabilities, handling exceptions, developing people, and maintaining professional standards. A service manager can simultaneously be a complex client task owner and a builder of after-sales methods; each role has clear permissions. Positions carry long-term responsibilities; specific task decision rights are determined by role, capability, and authorization.

This division must be honored in daily behavior. When the frontline makes a rule-based decision leadership dislikes, leadership lets it execute; when major new risks appear, the responsible person intervenes promptly, explains why, and owns the outcome. Respect for in-rule decisions and takeover of out-of-rule risks together constitute credible delegation.

A Concrete Starting Point: The Dispatch Scenario

The organization can begin with a limited-scope adjustment. Business leaders clarify service targets, resource ceilings, and exception adjudicators; professionals define "restored operation" conditions; the field lead selects plans and coordinates already-committed support. AI aggregates information, compares options, executes routine actions within authorized scope; client commitments, professional conclusions, and major risk acceptance each have a clear owner. After action, someone evaluates full results: Was the downed client helped in time? Did other clients wait longer? Were engineers overloaded by consecutive dispatches? These enter review. If recurring objections show "customer value" is too narrowly defined, the rule owner organizes revision; if frontline capability has risen, authorization expands; if a judgment consistently errs, training, permissions, or methods change. Results feed the next cycle of rule and authorization formation.

Evaluation must also affect rewards and growth. Solving hard problems, surfacing risks, contributing reusable methods all need recognition. Employees taking more complete responsibility should simultaneously gain matching authority, resources, and fair returns; frontline staff acting per rules and reporting issues honestly deserve institutional protection. Only pushing responsibility down consumes the proactive judgment the organization most needs.

Shared direction enters tasks; permissions and resources support node actions; professional review and outcome evaluation find problems; problems drive adjustments to shared rules, capability building, and authorization scope. Power configuration must form such a loop for the organization to continuously self-correct while releasing action.

This gives managers a concrete test: Do in-rule decisions still require private pleas? Do promised resources actually arrive? Can professional opinions stop out-of-bounds actions? Can field-discovered errors change rules? Observe delivery quality, governance cost, and human burden together; an arrangement that adds procedure and makes collaboration harder equally needs correction.

Conclusion

In the AI era, where should organizational power shift? The mycelium organization's direction: Move concrete work decision rights toward nodes that have capability, authorization, and corresponding outcome responsibility; concentrate power to maintain shared direction, public capabilities, and overall risk in a reliable governance core. The two connect through shared rules and accept testing by actual results.

In practice: business decision rights move from layered requests to frontline judgment; professional judgment and veto gain respect based on qualification and capability; resource support shifts from inter-departmental pleading to arrival per shared agreement; rule-making and interpretation have clear responsible parties, constrained by process and dissent. Senior leaders reduce intervention in routine actions, applying judgment to strategic trade-offs, long-term building, and major risks. People affected by decisions gain effective feedback and review channels, enabling power to be corrected.

Under this arrangement, the opening field lead can adjust dispatch within authorized resource range; if the adjustment crowds out another client's earned commitment, a coordinator adjudicates and arranges remedy; if small clients' urgent faults are systematically underestimated, the rule owner organizes a review of sorting standards. Frontline action, conflict adjudication, and rule revision each have owners; the enterprise can track whether commitments are kept and whether experience changes the next work. Power shifts in these concrete relationships. Moving from hierarchy to network, from position to node, requires matching power configuration. Positions carry long-term responsibility; tasks organize concrete action; new insights from action change shared rules and public capabilities — making "strong core, active frontline" a runnable organizational design. Such an organization lets professionals have judgment weight without waiting for promotion, lets the frontline get support without relying on personal connections, and lets managers expand organizational capacity by developing more reliable judges. Power thereby serves more complete value creation and leaves room for human growth. The organization's shared growth gains a sustainable foundation.