oGPT
A conversational interface for ad-hoc questions, like “what are my sales forecasts” or “who are my top customers,” so users can get answers without hunting through reports.
Designing AI-powered ERP workflows for growing businesses
TL;DR
I’m shaping AI-assisted ERP workflows that help traditional businesses move from fragmented, manual operations to clearer, digital decision-making.
Disclaimer: This case study is a high-level overview — some specifics are covered by NDA. Happy to walk through the details in a call.
Enterprise software shouldn’t hide complexity—it should help people work with it.
Oboda serves businesses managing inventory, sales, purchasing, and fulfillment, often under real operational pressure: tight margins, manual workarounds, and teams with mixed comfort levels using software.
The design challenge
Bring AI into existing workflows without adding friction. Insights need to show up at the right moment, be easy to trust, and never take the decision out of the user’s hands.
Start with the decision, then design the interface around it.
I partnered with product and engineering to map the jobs behind core ERP screens: understanding what changed, deciding what to do next, and acting with confidence. Rather than treating AI as a separate destination, the direction was to embed it where decisions already happen.
01 · MAP
Identify the operational questions and handoffs around a workflow.
02 · BENCHMARK
Study how leading ERPs approach AI to understand table stakes, differentiation, and gaps.
03 · PRIORITIZE
Separate information users need now from supporting detail.
04 · PROTOTYPE
Test the clarity of recommendations, states, and next actions.
Make the AI legible, actionable, and easy to verify.
- Add AI at the decision points that matter, inside the screens users already trust, not as a bolt-on feature.
- Pair every insight with its source data, so users can verify the “why” before acting on the “what.”
- Use progressive disclosure so dense data stays scannable, while users who want to go deeper still can.
- Design for uncertainty and review, keeping the user, not the model, in control of the final call.
Two ways to bring AI closer to the work.
oInsight
Contextual insights layered onto existing cards, data grids, and charts, turning raw data into clear, actionable takeaways at the point of use.
Building the foundation for human-centered AI in ERP.
- 72% reduction in time-to-decision for creating product orders.
- Adopted as the reference pattern for AI across the entire ERP.
This is ongoing product work. The design system and interaction patterns now give teams a consistent way to introduce AI across the platform, without overriding the operational expertise of the people using it.