Skip to content
Article reader Listen + reading controls
LISTEN + READ YOUR WAY

Article reader

Preparing the reader…

0:00 0:00
Reading settings
Text size
100%

GIDE and the Organizational Work of AI Adoption

It is tempting to describe the Global Information Dominance Experiments (GIDE) as a technical program: connect data, apply analytics and artificial intelligence, and accelerate decisions across the joint force. That description is accurate but insufficient. The more difficult problem GIDE confronts is organizational.

Defense institutions are not short of promising prototypes. They are constrained by the distance between a technical possibility and an operational capability: fragmented authorities, incompatible incentives, acquisition boundaries, security processes, data ownership disputes, uneven skills, and a rational reluctance to alter mission workflows on the strength of a demonstration.

Recurring experiments matter because they can make those dependencies visible and negotiable. Properly structured, GIDE is not only testing technology. It is rehearsing a different way for the Department to build, govern, and absorb digital capability.

The Chief Digital and Artificial Intelligence Office (CDAO) characterizes GIDE as a quarterly series involving the services, all combatant commands, the Joint Staff, and international partners. Its role is to mature data and software capabilities incrementally and connect them to acquisition pathways in support of Combined Joint All-Domain Command and Control (CJADC2).

That description contains an organizational design hiding in plain sight: a common cadence, cross-boundary participation, concrete mission threads, incremental delivery, and an intended route to sustainment.

AI Adoption Fails at the Seams Between Organizations

An AI capability is usually discussed as if it belongs to a program office or application team. In practice, its performance depends on a network of actors:

  • operational users who define what matters and interpret ambiguous conditions;
  • mission owners who hold decision authority;
  • data stewards who control access and meaning;
  • platform teams that provide compute, identity, networking, and observability;
  • cybersecurity and test organizations that evaluate risk;
  • acquisition professionals who create contractual and funding mechanisms;
  • vendors who build components and maintain proprietary knowledge;
  • and senior leaders who resolve conflicts that no technical team can adjudicate.

Each participant can perform well within its local mandate while the overall capability fails. A data steward can correctly restrict access, a security team can correctly demand evidence, an acquisition office can correctly enforce a contract, and an operator can correctly reject a workflow that increases risk. The failure emerges from the interfaces among their decisions.

This is why “culture” is often invoked in digital transformation conversations—and why the term can be unhelpfully vague. The operative question is not whether people favor innovation in the abstract. It is whether organizational structures allow them to make coordinated changes without violating their responsibilities.

GIDE creates a bounded environment in which those conflicts become attached to a mission outcome rather than debated as general policy.

Recurrence Changes the Incentive Structure

A single demonstration rewards presentation. A recurring experiment can reward learning.

When teams know they will return to the same mission problem in roughly 90 days, an unsuccessful result does not have to be hidden or rhetorically converted into success. It can become a prioritized input to the next iteration. The cadence creates several useful pressures:

  1. Small enough increments to finish. Teams must choose changes that can be integrated and evaluated within a bounded period.
  2. Evidence over aspiration. Claims encounter users, data, interfaces, and mission constraints.
  3. Continuity of responsibility. Known defects can be tracked across iterations rather than rediscovered by a new initiative.
  4. Earlier cross-functional engagement. Security, data, acquisition, and operational stakeholders cannot remain late-stage reviewers if the team intends to show measurable progress at the next event.
  5. A forcing function for decisions. Leaders must resolve which mission threads, interfaces, and constraints matter most.

The Department’s 2024 Innovation Fact Sheet later described GIDE as an every-90-day process to test, measure, optimize, and field CJADC2 solutions. The cadence is not merely a schedule. It is a governance mechanism.

Shared Mission Threads Become Boundary Objects

Large organizations struggle not only because their systems use different formats but because their communities understand the mission differently. A target, track, risk, priority, or course of action may carry different meanings at different echelons or within different functional communities.

A shared mission thread can operate as a boundary object: something concrete enough for multiple groups to work on together while allowing each group to contribute its specialized knowledge. Operators can discuss decision consequences, engineers can trace system behavior, data professionals can identify semantic conflicts, security teams can model threats, and acquisition officials can identify transition constraints—all against the same operational sequence.

This is more productive than asking stakeholders to “align on AI.” It changes the conversation from general enthusiasm to specific responsibility:

  • What decision are we improving?
  • What evidence does the decision-maker require?
  • Where does the evidence originate?
  • Which transformations occur before it is displayed or recommended?
  • Who may see it, modify it, or act on it?
  • How will the workflow behave when a dependency is unavailable?
  • What would constitute unacceptable failure?

Over repeated iterations, the answers can mature into shared vocabularies, data contracts, interface standards, tactics, techniques and procedures, and evaluation criteria. That organizational knowledge may prove more durable than any single software component.

The Necessary Change Is Double-Loop Learning

Most improvement efforts practice what organizational theorists call single-loop learning: detect a variance and adjust behavior to meet the existing objective. A data pipeline is slow, so the team optimizes it. A model produces too many false positives, so the threshold is tuned.

AI adoption also requires double-loop learning: questioning the assumptions, policies, and objectives that produced the workflow.

Perhaps the real problem is not the speed of the pipeline but that the organization is collecting data for reporting rather than decisions. Perhaps the model should not be optimized because the decision does not warrant automation. Perhaps a classification practice protects information but also prevents the coalition partner responsible for acting from receiving it. Perhaps the program’s success metric rewards delivery of software rather than improvement of a mission outcome.

GIDE 8 explicitly sought not only to iterate CJADC2 concepts but to develop tactics, techniques and procedures and influence policy. That breadth is essential. If an experiment is permitted to modify technology but not doctrine, authority, funding, or policy, it can optimize only within the constraints that may be causing failure.

Organizational Change Must Survive the Exercise

Experiments can temporarily overcome institutional barriers by assembling senior attention, exceptional technical talent, special access, and urgent support. The danger is mistaking this temporary coalition for a scalable operating model.

For change to persist, GIDE-derived practices need durable homes:

  • Product ownership: A named organization remains accountable for mission outcomes and backlog decisions after the event.
  • Data stewardship: Data access, quality, provenance, and semantic responsibilities are assigned and funded.
  • Platform stewardship: Common services have service-level objectives, support models, and transparent costs.
  • Model governance: Evaluation evidence, limitations, monitoring, and human decision rights travel with the capability.
  • Workforce development: Operators, technologists, security professionals, and acquisition personnel learn together rather than through isolated training.
  • Transition mechanisms: Successful increments can move into contracts, budgets, authorities to operate, and operational support without being rebuilt from the beginning.

The last item is decisive. The CDAO’s current description of GIDE explicitly connects experiments to acquisition pathways. Without that connection, rapid learning can actually deepen frustration: users repeatedly see what is possible but remain unable to depend on it.

A Federated Model for Department-Wide Change

The organizational lesson is not that the CDAO should centrally build every AI system. Mission knowledge is distributed, and combatant commands and services must retain meaningful ownership. Nor can every organization independently recreate data, platform, evaluation, and acquisition machinery; that produces incompatible islands and duplicated cost.

The stronger design is federated:

  • enterprise organizations provide common platforms, standards, reference architectures, evaluation methods, and marketplaces;
  • mission organizations own operational problems, product priorities, and acceptance decisions;
  • cross-functional teams integrate policy, security, data, software, and human factors from the beginning;
  • recurring experiments provide the shared evidence through which local and enterprise decisions are reconciled.

This is institutional architecture, not only technical architecture.

The Broader Lesson

Organizations often assume that adoption follows capability: first build a technically superior system, then manage the change required for people to use it. In complex public institutions, the sequence is circular. The organization must change enough to build and evaluate the capability; experience with the capability then reveals the next organizational change required.

GIDE’s greatest contribution may be to make that loop routine.

The strategic measure of success is therefore not the number of AI-enabled demonstrations. It is whether the Department becomes able to integrate operators, engineers, data stewards, security professionals, acquirers, and mission partners around a measurable problem—and then preserve what they learned in the machinery of ordinary operations.

That is how experimentation becomes transformation.

This essay was substantially revised in July 2026 to replace the original short news commentary with an evidence-based analysis.

My broader work examines these intersections among AI engineering, knowledge infrastructure, governance, and transformation systems. I also welcome practitioner perspectives and constructive disagreement on LinkedIn.

References

READER-NEUTRAL SUBSCRIPTION

Follow Field Notes via RSS.

Copy this address into the RSS reader you already use. New notes will appear there automatically—no account, email address, or tracking required.

Open the feed