Case Study · AI Automation Program
AI Automation Program: Operations & Delivery Performance Insights Case Study
Operational transparency for leadership: delivery health and the value of every automation, in one view.
The result: Senior management no longer waits for a status deck. Delivery KPIs and the AI automation program sit in one live view, refreshed every Monday, with every figure traceable to its Jira ticket: 3,410 hours saved a year, 912 tickets resolved and 26% of total work automated.
Results for the AI Automation Program
The tracker changed how leadership oversees the account: reviews now run off one live view instead of a prepared report.
| Area | Before | After |
|---|---|---|
| Leadership updates | Collated by hand from Jira, spreadsheets and email before every review | Live on one URL; no preparation needed |
| Automation value | Savings claimed loosely, hard to verify | Hours saved recorded per automation and rolled up, each tied to a ticket |
| Blockers | Client approvals and access requests buried in threads | Counted on the front page, with the oldest request flagged, so the client can clear them |
| Delivery quality | Reported week to week with no running total | Defects, avoidable errors and security incidents tracked weekly with totals |
| Slow onboardings | A total with no explanation | A timeline showing which team held the work and the delay driver |
- Stalled automation proposals and access requests now get cleared faster, because they are visible to the people who can unblock them.
- Onboarding delays are fixed at their source rather than debated, since each ticket shows where the time went.
- The automation program's value can be defended in a leadership review, because every figure traces back to a Jira ticket.
The customer
A marketing agency that manages data for a portfolio of brand clients, connecting their advertising and commerce platforms into a central data stack. Decision Foundry runs the agency's data operations: onboarding new brands, pipeline monitoring, access management and a steady stream of Jira tickets, plus an ongoing program of automations to cut manual effort.
At a glance
- Customer
- Marketing agency managing data for a portfolio of brand clients
- Decision Foundry's role
- Runs the agency's data operations and an ongoing program of automations
- Solution
- Lightweight web app on a single URL with two workspaces
- Refresh
- Every Monday
- Source of truth
- Every figure traceable to its Jira ticket
The problem
The client's leadership could not see, in one place, how operations and delivery were performing or what the automation program was worth. The facts existed, but they were spread across Jira, spreadsheets, status emails and meeting notes.
- Automation status, blockers, effort and expected savings lived in different places, so nobody could say what had shipped or what it saved.
- Proposals waiting on the client and pending access requests were easy to lose, and quietly stalled builds.
- Delivery quality (tickets resolved, regression defects, avoidable errors, security incidents) was reported ad hoc, with no running total.
- Time to onboard a client had no reliable answer, and no way to see where a slow onboarding had stalled.
- Every leadership update meant hours of manual collation.
The solution: one view of the AI automation program
Operations & Delivery Performance Insights is a lightweight web app on a single URL with two workspaces, each opening on its headline metric so leaders see the story before clicking anything.
Automation Tracker
Each automation with its ticket, business impact, savings, effort and progress.
Lifecycle tabs
What is moving, what is blocked and what is waiting on the client, kept apart.
Delivery Performance Summary
The top delivery KPIs week by week, with a running total.
Drill-down
Any number opens the tickets behind it.
Ticket timeline
A Gantt view of who held each ticket, and for how long.
Inside the AI automation program insights: from overview to ticket
1. Front door: choose a workspace.
The landing page offers two workspaces, the Automation Tracker and Delivery Performance. Each card carries its headline metric, hours saved and tickets resolved, so the story is clear before anyone clicks.
2. Automation Tracker: portfolio view.
The status bar shows hours saved, automations live and in build, proposals awaiting the client and the age of the oldest access request. Tabs split the portfolio by lifecycle stage. Each row links to its Jira ticket and shows business impact, impact category, annual savings, build effort, hours tracked against estimate and progress.
3. Delivery Performance: weekly summary.
A weekly grid of the top delivery KPIs with a running total: tickets resolved, regression defects, avoidable errors, security incidents, average days to onboard a client and the share of work automated. It refreshes every Monday, and every number is clickable.
4. Drill-down to the records.
Clicking the onboarding average opens the tickets behind it, with client, platform, assignee, created and resolved dates and onboarding days. Notes explain outliers, such as a ticket reopened only for clarification.
5. Ticket timeline.
Each ticket opens a Gantt-style timeline showing which team held it and for how many business days: ticket raised by the client, investigation by Data Ops, review by the client, then resolved. Delay drivers are flagged in red, so slow onboardings point to their cause.
Challenges faced tracking the AI automation program
Two very different audiences had to be served by one tool: engineers who live in Jira, and senior managers who want a number and a trend.
Automations move through many states (proposed, awaiting client approval, blocked on access, in build, in production, paused, cancelled), and each state had to be visible without hiding the others.
Savings had to be credible. Every automation needed a stated business impact, an impact category and an annual-hours estimate that could be traced back to a ticket.
Delivery metrics had to be counted the same way every week, so totals and week-on-week columns could be compared.
Onboarding time is distorted by reopened tickets and client-side waits, so the tracker had to show where time was spent, not just the total.
Some automation coverage sits with the client's own team, so the share-of-work-automated figure had to state its own limits.
Get In Touch