V.ISC.2621 - Agentic AI Sprint 1 - Value
Analysis of 1,437 eNVD Jira issues found the typical development ticket clears in about three days, yet the slowest quarter stall for roughly twelve — mostly waiting on clarification rather than active coding.
| Project start date: | 03 June 2026 |
| Project end date: | 12 November 2026 |
| Publication date: | 06 October 2026 |
| Project status: | In progress |
| Livestock species: | Grain-fed Cattle, Grass-fed Cattle, Sheep, Goat, Lamb |
| Relevant regions: | National |
|
Download Report
|
|
Summary
A four-week agentic sprint (15 June – 10 July 2026) moved the eNVD engineering pod from ad hoc AI use to a governed, single-agent workflow, delivering three live features end-to-end. Engineering is now saving around three hours per ticket, with larger pod-wide gains projected once the approach extends to business analysis and QA.
Objectives
To change how the eNVD team builds software by embedding a governed agentic workflow in engineering and proving the model before widening it — targeting capacity and delivery speed rather than a single dashboard metric or cost saving.
Key findings
Engineering pipeline (skills, MCPs, hooks, evals, guardrails) is implemented with three features live. Around three hours saved per ticket is demonstrated in engineering; roughly nine hours per ticket — about 600 working days a year per pod — is projected across all three roles. Early adoption signal is strong but small (nine of ten rated AI confidence 4–5 out of 5, n=10).
Benefits to industry
Faster, more reliable eNVD delivery means integrity and traceability improvements reach producers and supply-chain users sooner. Capacity handed back to the team can be reinvested in further enhancements without added headcount cost.
MLA action
Confirm the value assumptions with the delivery team, run the adoption survey across the whole team, agree a small set of ongoing throughput and time-saved measures, and scope the BA and QA extension as the next phase.
Future research
The next phase would extend agentic delivery upstream into business analysis (dev-ready specifications) and downstream into QA (automated coverage), then scale the pattern across other eNVD pods. Light effort-logging on a sample of tickets is needed to confirm the three-hour saving against a baseline.
More information
| Project manager: | Joel Wilson |
| Contact email: | reports@mla.com.au |
