AI, automation and operational design

Eliminating timesheets entirely.

A recurring three-day administrative process was producing unreliable R&D and CAPEX data while frustrating the people required to complete it. I built an automated reporting system using Jira activity and workflow history. More than two years later, nobody completes a timesheet and the reporting run takes around five minutes.

3 days → 5 minRecurring reporting effort

0 timesheetsCompleted manually for 2+ years

~75%Of the business population needing time tracking covered

Audit testedReporting has withstood formal scrutiny

The transformation

Stop chasing people. Fix the system.

The obvious problem looked like timesheet compliance. The real problem was that the organisation was asking people to manually recreate information its systems already contained.

Before

Manual reconstruction

~3 days chasing and reconciling every reporting cycle

People associated time with dead or outdated epics

Roadmap and initiative attribution was inconsistent

CAPEX and R&D reporting required downstream cleanup

I became the face of an admin request nobody wanted

→

After

System-generated evidence

~5 minutes to initiate a reporting run

No employee timesheet completion required

Jira activity is mapped to initiatives and financial categories

Multiple business groups included in one repeatable process

Structured output ready for finance and audit review

The story

From compliance problem to operating mechanism.

The product mattered, but the key move was reframing the problem from “how do we make people complete timesheets?” to “why are we asking them to do this at all?”

1. The pain was bigger than admin

The process consumed days of recurring effort and created friction across the organisation. Even when people complied, the resulting data was unreliable because work was not consistently associated with the right epics, roadmap items or active initiatives.

2. Jira already contained the evidence

Rather than rely on memory and manual entry, I used Jira activity and workflow history as the system of evidence. The goal was to infer where effort had gone, then map that effort to the categories finance actually needed.

3. I built the missing mechanism

Using AI-assisted development, I built Tediuma: a working application that automates time attribution, supports business-group rules, handles leave inputs and produces exportable R&D and CAPEX reporting.

4. The behaviour disappeared

The strongest outcome is not software adoption. It is that the original administrative behaviour vanished. For more than two years, nobody has had to complete a traditional timesheet while the reporting process continued to operate and withstand audit scrutiny.

Working system

Proof that the mechanism became operational.

The screens below are anonymised mock-data recreations inspired by the production workflow. They show the operating pattern without exposing real people, projects or financial data.

Tediuma — Reports

Reports

Generate and manage automated time-attribution reports

Report Name	Period	Groups	Status	Created	Action
July 2026 Time Report	Jul 2026	Engineering, Product +3	Processing	Today 6m 12s	Cancel
June 2026 Time Report	Jun 2026	Engineering, Product +3	CompleteEmail sent	1 month ago 5m 48s	Download
May 2026 Time Report	May 2026	Engineering, Content +2	CompleteEmail sent	2 months ago 7m 03s	Download
April 2026 Time Report	Apr 2026	Engineering contractors	CompleteEmail sent	3 months ago 1m 41s	Download
March 2026 Time Report	Mar 2026	Engineering, Product +3	CompleteEmail sent	4 months ago 6m 26s	Download

Mock data. This recreation is designed to demonstrate the reporting workflow and operational maturity without exposing production information.

Structured export

Fake example showing the fields used for traceability.

assignee	epic_key	epic_summary	roadmap_item	issue_key	issue_type	issue_status	issue_summary	hours	story_points	project_key
Alex M.	PRG-102	Customer onboarding	Growth	PRG-448	Story	Done	Improve invite flow	18.5	5	PRG
Jordan T.	PLT-220	Platform resilience	Scale	PLT-771	Task	Done	Service alerting uplift	12.2	3	PLT
Sam K.	AI-084	AI authoring tools	Growth	AI-193	Story	Review	Draft generation workflow	21.7	8	AI
Priya R.	OPS-055	Operational efficiency	Keep lights on	OPS-411	Bug	Done	Fix reporting timeout	6.4	2	OPS

Leadership takeaway

The leadership move: stop forcing people to manually recreate information the system already knows, then build the mechanism that makes the better process operational.