Every assignment cycle, military personnel teams balance mission needs, member preferences, family considerations, funding and policy across thousands of positions. Permuta set out to give them one connected system for that work, built on the DefenseReady platform and delivered as SaaS at Impact Level 5 (IL5) on Microsoft Power Platform and Dataverse.

The result is three wins in one program: a working enterprise talent-management solution, a proven model for secure SaaS delivery and an AI capability that reached agentic maturity more than a year ahead of our own roadmap.

In 11 sprint releases between March and October 2026, the team turned a complex, multi-office assignment process into connected, role-based tools for every person in the chain, from the individual service member to the commander.

Next on the same foundation: retention. The shared member record, career-path and routing services were extended into retention, with career counselor roles, unit counselor assignments and interview-request workflows in development.

This program proved that a full mission application, AI included, can be built, secured and delivered as SaaS at IL5 on standard Microsoft Power Platform and Dataverse services. For any agency that wants modern capability without standing up and accrediting its own infrastructure, that is the headline.

• Native Dataverse security. Role-based personas for service members, assignment managers, command staff and family-support reviewers, with Privacy Act and system-of-records notices at sign-in.
• Automated, repeatable delivery. 54 DefenseReady solution packages moved onto CI/CD pipelines, with static application security testing (SAST) wired directly into the pipeline.
• Security by default. Five batches of code-scan remediation and manual review of security hotspots before release, not after.
• Built for real data. Production-pattern data loads for member records, billets, staffing goals and related tables, backed by a source-to-target mapping of roughly 300 fields.
• Accessible by design. Accessibility labeling and role-specific usability fixes handled inside normal sprints.

Permuta’s December 2025 DefenseReady AI roadmap placed Level 3 agentic products in March 2028, with agentic services starting in November 2027. In this program, the team achieved working Level 3 agents in October 2026.

Level 1, in-context experts. Chat assistants are embedded in the mover queue, billet vacancy, unit management and career-path tools, with context engineering refined in the most recent release. Platform-level Level 1 chat now ships in the standard DefenseReady release across about 20 of the most-used features.

Level 2, machine learning. An August 2025 prototype combined five ML models trained on service data to simulate a member’s career path: total service, next rank, next duty title, time in position and recommended specialty. Recommendations can follow member preferences or policy, by hand or automatically. 2026 work extended the approach to retention models for contract risk, fitness and assignments.

Level 3, agents. Agents that reason across tools, query live data, follow policy and learn from the people who use them.

Inside the Level 3 agent

Level 1 and Level 2 testing showed that most AI shortcomings trace to three problems: too little transparency, no access to the data the agent needs and incomplete instructions. The Level 3 architecture addresses all three.

Picture an assignment manager checking which aviators in a population hold flying-duty status. The agent reads the governing policy from its knowledge base, queries live Dataverse records and shows every step in its reasoning log. If a data field has changed, the manager corrects the agent in plain language. The agent asks permission to save that lesson, and once a human approves it, applies the same logic in every tool and keeps it for the next manager assigned to that population.

Why it matters: assignment managers rotate. With approved, shared memory, hard-won institutional knowledge stays with the mission instead of leaving with the person. Agent instructions also require evidence, check whether policy applies to officers or enlisted members and keep a human in the loop for any record the agent creates.

• Volume at pace. 325 work orders completed and more than 8,500 engineering hours invested, with 11 consecutive sprint releases on a three-week cadence from March to October 2026.
• Four levels of testing. Each release passes unit, integration, baseline and requirement-validation testing before it reaches the customer. All 17 items in the most recent release cleared all four levels.
• AI-accelerated quality. Our QA engineering uses AI to generate test cases and importable test suites, building a library of more than 1,000 baseline test cases and speeding full-baseline regression.
• Depth behind the work. One integrated team spanning development, testing, data integration, deployment, security and documentation.

• Built for every service. The capabilities follow processes common across the military: positions and billets, assignments, orders, routing approvals, family readiness and retention.
• A shorter path to secure cloud. A proven SaaS IL5 pattern on Power Platform and Dataverse, with security scanning built into delivery.
• AI you can govern. Transparent reasoning, policy grounding and human-approved memory, on the same platform that holds your authoritative data.
• A partner who knows the mission. For primes and integrators, a product team that pairs a mature COTS platform with deep military personnel domain knowledge and a delivery record you can measure.