Work in Concert: Ship a Multi-Agent Accelerator in Six-to-Eight Weeks

A global cloud and technology leader partnered with Persistent to build and maintain GenAI solution accelerators—modular templates that help developers and enterprises adopt AI capabilities without buying a full product. One of these, the Multi-Agent Custom Automation Engine (MACAE), shows how multiple specialized agents can collaborate to execute real, multi-step work at enterprise scale.

This makes an open-by-design, reusable accelerator more valuable than isolated demos.

The Gap: Curiosity Without a Rail

As GenAI gained traction, interest in agentic AI rose—but most enterprises lacked a safe starting point. Teams had no reference models for multi-agent interaction, faced a steep technical barrier to build orchestration from scratch, had unclear first steps, and still ran manual, fragmented workflows (e.g., onboarding, procurement). The result: missed opportunities—fewer experiments, slower operations, inconsistent outcomes.

Co-Create the Rail, Not a Bot

Persistent moved from support to full co-development, collaborating on design, architecture, and delivery. The goal wasn’t to ship a product; it was to publish a practical, reusable reference developers could extend. The accelerator formalized generic task and agent definitions, enabled customizable agent roles (HR, IT, Operations), and supported multi-stage, human-in-the-loop flows—so teams could orchestrate steps, trigger actions, and update systems in context.

Delivered inside the client’s standard six-to-eight weeks cadence (planning → development → testing → deployment), the accelerator shipped with clear documentation and a GitHub release to enable global experimentation and customization with documentation kept clear, updated, and responsive to user needs.

How It Works – Five Moves from Ask to Action

1. Define a goal. A user states a plain-language task (e.g., “Onboard a new employee”).
2. Decompose. The system breaks the task into stages (background check, laptop assignment, handbook share, signatures).
3. Assign agents. Sub-tasks map to specialized agents—HR (verification, handbook), IT (equipment), and a Generic agent for fallback—each tied to backend functions in the codebase.
4. Human-in-the-loop. Users approve, reject, or trigger stages to maintain transparency and control.
5. Trigger actions. On approval, the system can send emails or update systems when customized.

Templates That Travel Like Onboarding or Approvals

Enterprises can add agents, tweak logic, and adapt domain flows (e.g., claims processing, project planning). One team’s blueprint becomes another team’s jump-start—accelerating adoption without reinventing orchestration.

From Experimentation to Adoption

In 2025, 91% of U.S. mid-market companies report using generative AI, signaling a shift from tool trials to operationalized workflows—exactly when governed, reusable accelerators matter. Globally, momentum is broad: 78% of organizations say they use AI in at least one business function (up from 72% in early 2024 and 55% a year earlier), underscoring the need to scale safely.

What changed for the client:

• Accelerated delivery. More accelerators shipped per quarter, each completed in six-to-eight weeks.
• Global accessibility. Publishing to GitHub lowered the barrier to try, fork, and customize.
• Scalable maintenance. Persistent handled issues, documentation updates, and compatibility with evolving cloud/AI models.
• Clear value to the client. Increased accelerator output and usage, greater Azure deployment by end users, and faster innovation via reusable GenAI blueprints.

Open by Design, Built to Last

Impact extended beyond a single launch. The work matured into a co-creation model—publish, learn, iterate—with a public codebase enabling forking, cross-industry reuse, and community-driven improvements. Persistent kept the rail current with tech-stack changes, bug fixes, and ongoing enhancements, turning a single accelerator into a foundation for broader use across teams and business units. The template sparked new accelerators on the same model and cross-team reuse inside the client’s organization.

Why Persistent

1. Proven GenAI delivery. Real agentic implementations, not lab demos.
2. Speed & scale. Agile execution with a consistent six-to-eight Weeks turnaround.
3. A reliable partner. Trusted to evolve from support to strategic co-creation.

Ready to move from chat to change? Ship your first multi-agent workflow in weeks. Scale safely with human-in-the-loop. Compound value with reusable templates.