A new way to manage invoices
When invoice volume grows, manual processing and approval bottlenecks make it harder for logistics startups to scale. I translated user feedback into an MVP in one week, connecting the needs of analysts, admins, and approvers through invoice validation, configurable automation, and mobile approval.
A process under strain
The challenge was connecting validation, routing, and approval across three roles while keeping the workflow clear.
Understanding the problem
I mapped roles across the invoice lifecycle and prioritized features by value and effort to fit the one-week scope.
Exploring through drafts
I sketched the first version in Codex, using established AP patterns to explore layouts and interactions directly in code.
AI-assisted MVP
By the end of the week, I had a functional MVP with features designed to address the pain points identified during research.
01 / Discover
Market research
Having never used an AP platform before, I started by exploring how existing tools supported everyday workflows. I compared the reviews of four leading AP platforms on G2, a business and software review aggregator, and used ChatGPT to analyse the feedback and identify recurring themes in invoice management.
Keep invoice reviews together
What I observedStampli and Airbase bring approval activity into the invoice view, keeping reviews in context.
Stampli
Airbase
How it shaped SennderBring the invoice list, selected invoice, and approval status into one workspace.
Invoice view should be easy to scan
What I observedYooz and Airbase automate matching and duplicate detection, helping reviewers focus on potential issues.
Yooz
Airbase
How it shaped SennderShow validation issues beside the invoice, so analysts can fix them before approval.
Make approval rules configurable
What I observedBill supports custom approval and escalation steps, adapting reviews to company rules.
Bill
How it shaped SennderBuild workflows with approver groups, amount thresholds, and approval steps.
02 / Define
Turning insights into a clear plan
Building on the market research, I anonymously surveyed AP analysts working at mid-sized logistics companies to better understand their daily workflows, the challenges they faced, and which features would be most useful in an AP automation platform.
Clarity and approvals on the go led the priorities.
A clearer interface led the requests, followed by mobile invoice review and simpler approvals.
- 32.4%
- Clearer interface
- 24.6%
- Mobile invoice review
- 17.7%
- Simpler invoice approvals
A clearer interface was the top request.
Approvers wanted to review invoices on the go.
Simpler invoice approvals were a top-three priority.
- Clearer interface32.4%
- Mobile invoice review24.6%
- Simpler invoice approvals17.7%
- Automated invoice processing11.3%
- Role-based approver groups9.9%
- Custom approval workflows9.9%
- Clearer notifications7.1%
- Team collaboration5.7%
- Flexible invoice management5.0%
- Faster performance2.1%
Mapping the invoice processing flow
AP platforms typically support three roles: admin, analyst, and approver. These roles shaped the three personas that guided my design, together covering the full invoice lifecycle.
- Workflow setup03
Admin
Sets up the workflow.
- Approver groups
- Assign approvers to the appropriate groups based on their roles.
- Invoice automation
- Configure automation rules so invoices can be processed securely.
- Invoice processing01
Analyst
Validates invoices.
- Invoice capture
- Uploads invoices for data extraction or receives them directly in the inbox.
- Invoice validation
- Reviews purchase orders and transaction history.
- Invoice approval02
Approver
Makes the approval decision.
- Invoice approval
- Reviews each invoice, then approves or rejects it with a clear reason.
- Secure access
- Confirms identity with two-factor authentication before access.
Choosing what to implement
By using a Value–Effort Matrix, I could prioritize features by their impact and implementation effort, focusing on those that could deliver the most value in the shortest amount of time.
Quick winsQuick wins
- Clearer interface
- Simpler invoice approvals
- Clearer notifications
Major projectsMajor projects
- Custom approval workflows
- Role-based approver groups
- Mobile invoice review
- Automated invoice processing
Fill-insNice to have
- Flexible invoice management
- Team collaboration
- Integration with chat apps
Time wastersLower priority
- Faster performance
03 / Design
First iteration
I quickly sketched the first version of the platform in Codex. This gave me a working draft to experiment with and explore different layouts. Compared with Figma, I could move much faster from idea to working feature. Because I was working directly in code, each iteration brought the app closer to a full release.
04 / Build
Learning from established patterns
To give those iterations a clear direction, I relied on ChatGPT to research UI patterns in AP dashboards and used what I found to shape the platform’s layout and interactions around established conventions. My goal was to create a familiar but simple interface that analysts could transition to with minimal effort.

Experimenting with approval rules
AI also made it easier to explore ideas that might otherwise have been too time consuming to implement. Although I ultimately chose a fairly standard approach to automation and approval groups, I could quickly explore alternatives with little to worry about beyond how many AI credits I had left.

Refining the inbox through AI conversations
Working with AI meant I could point to any element, describe what felt wrong, and see the adjustments made in real-time. The inbox evolved through these small exchanges, and my role became increasingly editorial: judging each result, correcting it, and deciding when it felt resolved.

Approvals beyond the desk
With Codex, I could easily translate the desktop experience to mobile. My research had made the priority clear: users wanted a lightweight mobile app they could use to quickly review and approve urgent invoices. This helped me keep the mobile app focused on the invoice inbox and its approval features.

05 / Key Takeaways
Conclusions
Building a functional MVP in a week felt achievable with AI, though testing it with users remains an important gap in my process. The experience showed me how radically AI has changed app creation, giving designers more freedom to explore, build, and refine ideas than ever before.
What the MVP delivers
Invoice validation
A focused workspace for analysts to review invoice details.
Configurable automation
Approval groups and workflow rules based on company hierarchy.
Mobile approvals
A dedicated experience for making invoice decisions on the go.
Functional platform
A one-week project connecting research, design, and development.