Project Overview
TruXpense is an AI-powered smart expense tracking platform designed to help individuals and businesses take control of their finances effortlessly. The app uses artificial intelligence to scan receipts, automatically categorize expenses, and surface intelligent financial insights — giving users a clear, real-time picture of where their money goes.
As the lead Product Manager and Project Manager at D'Accubin Technology Limited (UK-registered), I joined the product post-inception and owned its journey all the way through to live production deployment across iOS (Apple App Store), Android (Google Play), and Web — and into the active market entry phase.
My mandate was clear: take a product with a strong idea but unresolved blockers, build the system to deliver it, and ship it to market — on time, to quality, and with a team that could sustain it beyond launch.
The Challenge
- The product had passed inception but had unresolved technical and delivery blockers — no clear path from where it was to a shippable product.
- A cross-functional team of 8 (developers, DevOps, product designers) needed structure, aligned sprint cadence, and clear ownership to function as a cohesive unit.
- AI features — receipt scanning, intelligent categorization, financial insights — required more than coordination; they required product judgment on accuracy thresholds, edge cases, and fallback behavior before they could be shipped responsibly.
- Simultaneous deployment to three platforms (iOS, Android, Web) meant three sets of environment requirements, review processes, and release criteria had to be managed in parallel.
- Stakeholders needed continuous visibility without being pulled into every technical decision — balancing transparency with delivery momentum was a constant tension.
The CALM Framework in Practice
Every product decision on TruXpense was guided by the CALM Framework — Clarity, Alignment, Leadership, Measurable Results. Here's how each pillar showed up in practice.
AI Features — Product Decisions I Owned
The AI layer of TruXpense was not just a technical implementation — it required product judgment at every stage. These were the decisions I made and the criteria I set.
Intelligent Receipt Scanning
Defined accuracy thresholds, edge case scenarios (blurry images, non-standard receipts, handwritten notes), and fallback UX when confidence scores fell below threshold — so the product gracefully handled failure without frustrating users.
Automated Expense Categorization
Set acceptance criteria for categorization accuracy and designed the user feedback loop — allowing users to correct categorizations in a way that improved the model over time. Balanced automation with user control.
Financial Insights Dashboard
Defined what 'useful' insights looked like from a product perspective — not just what was technically possible. Prioritized spending pattern summaries and alerts over vanity metrics that looked impressive but didn't change user behavior.
KPIs for AI Feature Adoption
Established the success metrics: receipt scan usage rate, categorization acceptance rate, insight engagement rate, and active user retention at 7 and 30 days. These became the roadmap's north stars post-launch.
The Approach & Solution
Blocker Resolution & Foundation Setting
Audited all open blockers from inception, assigned ownership, set resolution timelines, and established a single source of truth in Notion. Within the first sprint cycle, the team had clarity on where we were and what needed to happen before the first release.
Product Roadmap & Sprint Structure
Built the full product roadmap from current state to three-platform launch. Broke delivery into phased sprints with clear milestones, AI feature gating criteria, and release checklists for iOS App Store, Google Play, and Web deployment separately.
AI Feature Definition & UAT Leadership
Wrote acceptance criteria for all AI-powered features. Designed and led UAT cycles for each platform, creating test cases that validated AI accuracy, edge case handling, fallback behavior, and cross-device consistency before any public release.
Demo Sessions & Stakeholder Review Cycles
Ran structured product demo sessions at the close of each major sprint milestone. Captured stakeholder feedback systematically and triaged it into the backlog — separating genuine product improvements from scope creep.
Cross-Platform Deployment & Launch
Coordinated simultaneous deployment to iOS (App Store review process), Android (Google Play), and Web production server. Managed release timing, pre-launch checklists, DevOps coordination, and go-to-market communications across all three channels.
Market Entry & Post-Launch Oversight
Transitioned from launch mode into market entry — monitoring user onboarding, triaging live product issues, reviewing stakeholder bug reports, and maintaining product quality while the next roadmap phase was being planned.
Results & Outcomes
TruXpense launched successfully across all three platforms simultaneously — a cross-platform release that many early-stage startups defer or stagger. The product went from a post-inception state with unresolved blockers to a live, publicly available product in both major app stores within the engagement period.
The AI features shipped to spec: receipt scanning, intelligent categorization, and the financial insights dashboard all passed UAT and went live with defined success metrics in place. Post-launch, the product remains active — being maintained, iterated on, and monitored as user adoption grows.
Key Learnings
AI features need product judgment, not just engineering. The hardest decisions weren't technical — they were choosing what accuracy level was "good enough" to ship, and designing the fallback experience so users weren't confused when the AI was uncertain.
Multi-platform launches multiply coordination complexity nonlinearly. Three platforms meant three review processes, three release timelines, and three sets of edge cases. Having platform-specific release checklists was essential — not optional.
Blocker resolution at the start saves multiples of time later. The investment in auditing and clearing post-inception blockers in the first sprint created the breathing room to ship with quality rather than constantly firefighting at the end.
Structured demo sessions change the quality of stakeholder feedback. Unstructured demos produce vague feedback. Structured sessions with prepared questions produced actionable, prioritizable input that genuinely improved the product.
Let's Build Something Together
Whether you're launching an AI product, scaling a remote team, or need clarity in complex delivery — I'd love to connect.
Get in Touch