One Artifact, Two Learners: Shared Specification for Enterprise AI Agent Onboarding
Soochna Shastra The Mega Journal of Information Technology · 01.06.2026
DOI: 10.59875/2210ssjit1758 ↗Kurz gefasst
Fachartikel in der Zeitschrift Soochna Shastra The Mega Journal of Information Technology, online erschienen am 01.06.2026. Verfasst von Karthick Pandiyan und 2 weiteren Autor:innen.
Abstract (Originalsprache)
Background
Enterprise information-technology teams increasingly encode release and operations workflows as plain-text specifications that software agents execute against work-item trackers, mail systems, and reporting tools. Release readiness is inherently conditional: practitioners branch on item status, product line, and milestone rules. When newcomers are still taught from slide decks, workshops, or scattered expert notes, the training material and the executed procedure drift apart, leaving no single authoritative starting point for onboarding.
Objective
This case report describes and evaluates Shared Specification, a pattern in which one workflow file trains the agent and simultaneously serves as the newcomer's curriculum, eliminating the need for a parallel training pack. Methodology: We applied a single-case industrial design in a telecommunications product organization. The newcomer received only the agent's specification -no presentation, companion document, or workshop -and asked the agent to teach the workflow steps in execution order. The learner recorded every branch the agent added, dropped, or changed relative to the text. We also executed the same specification once against a live work-item tracker at a checkpoint gate and summarized operational measures. Key findings: The trainer reported a clear onboarding starting point because the session began with the file the organization already uses to run the process. One checkpoint execution processed 73 tracker rows, excluded 15, kept 58 in scope, produced five consolidated drafts, and returned a red readiness verdict at 60.0% CAT1 completion against an 80% threshold. Observed failure modes include ambiguous specifications, fluent but non-procedural model answers, teach-back that differs from live execution under time pressure, and tacit knowledge not captured in text.
Conclusion
Shared Specification offers a practical way to align human knowledge transfer with agentic workflow automation in enterprise IT operations. IT managers can adopt the pattern immediately, but comparative learning outcomes against workshops are not yet measured. Supplementary appendices provide the anonymized specification and execution record for independent review.
Zitieren
APA
Pandiyan, K., Nanjappan, D., & Periyaswami, S. (2026). One Artifact, Two Learners: Shared Specification for Enterprise AI Agent Onboarding. Soochna Shastra The Mega Journal of Information Technology. https://doi.org/10.59875/2210ssjit1758
Vancouver
Pandiyan K, Nanjappan D, Periyaswami S. One Artifact, Two Learners: Shared Specification for Enterprise AI Agent Onboarding. Soochna Shastra The Mega Journal of Information Technology. 2026. doi:10.59875/2210ssjit1758
BibTeX
@article{Pandiyan2026,
author = {Pandiyan, Karthick and Nanjappan, Dharani and Periyaswami, Shankar},
title = {One Artifact, Two Learners: Shared Specification for Enterprise AI Agent Onboarding},
journal = {Soochna Shastra The Mega Journal of Information Technology},
year = {2026},
doi = {10.59875/2210ssjit1758}
}
EUVIMED