A Singapore Government Agency Website
How to identify
Fragmented email approvals and untracked commitments obscure budget visibility, triggering unmonitored spend, contraventions, and audit failures.
Product Owner/Manager: Emily Li – GDP (CEP)
User Research: Zoey Lim – GDP (CEP)
User Research: Marcus Wu - SCG (FIN)
User Research: Sarah Chua - SCG (FIN)
Software Engineer: Jason Lee - GDP (CEP),
Software Engineer: Kristine Lee - GDP (CEP),
4,000 GovTech employees, primarily Project Managers, Product Managers, and Delivery Managers.
Users need to raise AORs, seek funding from the correct cost centers, and ensure all approvals follow the established authority matrix, then identify the appropriate procurement methods, then follow through to procure, then track the AOR spending statuses.
Currently, users rely on email approval loop by manual checking the authority matrix and cost centers without any AOR templates, then followed by manual Excel trackers or fragmented email chains to monitor spending against approved limits. This occurs throughout the procurement lifecycle, but the lack of centralized visibility often leads to budget overages (contraventions) being discovered only when Finance rejects an invoice at the end of the cycle or wrong level of approvals sorted (contraventions).
The root cause is a lack of standardization and decentralized record-keeping. Up to 80% of approval emails contain inconsistent information, and 100% of records are stored in personal inboxes rather than a systematic database. This manual process results in a 5–10% error rate in data entry and a risk that 20% of requests are approved at the wrong authority level due to the absence of automated validation.
Continuation of significant financial and compliance risks. Without a centralized system, there is a potential $1M–$5M exposure for every $100M spent due to weak controls. This leads to a 100% dependency on manual retrieval during audits, increasing the risk of "Procurement Irregularity" narratives and public scrutiny. Furthermore, the inefficiency results in up to 3,000 man-hours wasted annually—the equivalent of losing 1.5 full-time employees (FTE) just to manual tracking.
A user survey was conducted to explore and validate current pain points and challenges. We received 17 responses from GovTech employees and external agency partners (e.g., PA):
❗ 100% Lack Centralized Tracking: All respondents confirmed they have no reliable system for tracking commitments, relying instead on manual searches through sent email folders, sticky notes, or fragmented Excel files.
❗ 94% Visibility Gap: 16 out of 17 users are unsure of their remaining AOR balances and feel there is currently no dependable way to verify real-time spending. This is the highest-ranked feature requested by users.
❗ 94% Template Confusion: 16 out of 17 users are uncertain about the standard AOR template, leading to low confidence that submissions include all necessary fields for timely approval.
❗ 76% Approval & Follow-up Friction: 13 out of 17 respondents are uncertain about authorized approver levels and reported spending excessive time using Email and Teams to chase approvals.
❗ 76% Manual Rework: 13 out of 17 users have had to re-do AOR submissions due to documentation errors. This indicates a significant loss in productivity.
❗ 71% Policy Uncertainty: 12 out of 17 users are unsure of the correct procurement policies to follow.
❗ 59% Categorization Issues: 10 out of 17 users struggle with selecting the correct spending category for their requests.
❗ 24% High-Risk Critical Failures: 4 out of 17 projects faced significant delivery delays because invoices were rejected by Finance due to AOR limits being exceeded.
This is ongoing process. Will update again.
Integration: The solution can coexist with or integrate into GovTech’s internal enterprise systems(eg. Workday) to manage evolving approval hierarchies, real-time procurement updates, and live invoice tracking.
Reliability: The system can accurately automate the complex Authority Matrix without manual intervention.
Data Security: AI processing of sensitive financial and project-specific data must meet Governmentpolicy and data classifications.
GovTech TMO is interested in exploring these ideas with us to determine if there is synergy with the system they are currently developing.
An AI-powered approval and tracking agent centralizes fragmented email approvals and automatically monitors spending against limits, eliminating the risk of untracked commitments and budget contraventions.
To save users’ time, we enable AI chatbot. Our AI asks the key questions to auto-populate the system fields, so users do not need to manually re-type justification.
(Have yet to conduct user testing with prototype at this point, will add in after)
o Percentage of Contraventions: an extremely low percentage indicates that our automated tracking and follow-up system is successfully eliminating contraventions and non-compliance related to overspending and incorrect routing.
o Percentage of "First-Time-Right" Approvals: A high percentage indicates that our automated validation is successfully preventing incorrect routing and incomplete submissions.
o Time from Request to Approval: A reduction in this metric indicates we are successfully eliminating the manual "chasing" of records and email approvals.
We plan to onboard 5 GDP Programmes for a pilot, with the goal of reducing the time spent on manual tracking and approval follow-ups by 60%, or approximately 2 hours per request. This will provide immediate budget visibility and prevent 100% of contraventions within the pilot group.