AI Agent Platform Development

Overview JFLCO is seeking a contractor to design, build, and validate an AI agent platform for Wellman Dynamics, a precision manufacturing portfolio company. The platform will sit on top of Wellman's enterprise data warehouse and deliver department-level AI agents that automate recurring analysis, reporting, and document-driven workflows. The concept is modeled on a comparable production deployment at a peer industrial company built with Claude Code (department agents composed of narrowly scoped subagents and skills, distributed to end users as plugins). Scope of Work • Agent architecture and build: one primary agent for each of 9 departments, each composed of approximately 10–15 subagents (~90–135 subagents total). Each subagent should be narrowly scoped to a specific task with dedicated skills, reference documentation, and context/memory files, with guardrails to minimize hallucination. • Discovery and process capture: structured interviews with department leaders and end users to document current workflows, desired analyses, and output formats. Existing SOP documentation is limited; the contractor should expect to build process understanding from interviews and source documents. • Integrations: connections to the enterprise data warehouse (cloud migration in progress; AWS anticipated) and Microsoft 365 (SharePoint, Outlook). Interim phases may run on static data exports before the warehouse migration completes. • Validation and iteration: hands-on testing cycles with Wellman staff for each subagent, reviewing outputs, incorporating feedback, and iterating until outputs are accurate and usable in day-to-day operations. • Deployment, governance, and handoff: packaging agents (ideally as Claude Code plugins) for end users, access controls aligned with data security requirements, user training, and documentation sufficient for internal maintenance after the engagement. Current Environment Wellman's operational data currently resides across two systems (ARD and Odyssey) and is being migrated to a unified cloud data warehouse. Employees use Microsoft 365. The organization is early in its AI adoption; no internal team exists to build this in-house, which is why consulting capability (discovery, change management, and user validation) is weighted equally with technical capability in our evaluation. Anticipated Phasing Phase 1: Discovery and proof of concept (interviews across departments, prioritized agent use-case roadmap, and a working pilot agent for 1–2 departments). Phase 2: Build-out of remaining department agents in priority order with per-subagent validation cycles. Phase 3: Full deployment, training, governance hardening, and knowledge transfer. Proposals should recommend a realistic timeline and sequencing based on your delivery model. Proposal Requirements Please include: (1) relevant experience and case studies, specifically production platform builds, subagent/skill architectures, or comparable agentic systems (manufacturing, industrial, or mid-market PE experience is a plus); (2) proposed team, engagement model (embedded vs. remote), and your methodology for discovery interviews and output validation with non-technical end users; (3) estimated timeline by phase; (4) pricing structure (fixed-fee by phase, per-agent, or time and materials) with assumptions; and (5) client references.