AI features users actually trust.
AI product design is the work of turning a model's output into an interface people understand, trust and act on. I design chat, scoring, recommendation and agent experiences for SaaS teams, and I led design on the AI-driven B2B marketplace named First Runner-Up, Best AI Product at AI Summit UAE 2025.
For teams whose AI works, but whose users hesitate
A model that performs well in testing can still go unused. If people can't predict what it will do, can't see why it gave an answer, or can't fix a wrong result, they quietly go back to the old way. This engagement is for:
- SaaS teams adding an assistant, copilot or chatbot to an existing product
- Founders building an AI-native product from zero to one
- Products where a score, ranking or recommendation drives a buying decision
- Teams whose AI usage is low even though the model looks good in demos
Six AI-native features, designed from scratch
At Zoftware, where I lead product design, I designed six AI-native features from scratch for an AI-driven B2B software marketplace serving enterprise buyers across 15+ countries: Strategy Builder, System Fit Score, RFQ Builder, Smart Comparison Tool, the Zain AI chatbot and a web-based telecaller.
The product was named First Runner-Up, Best AI Product at AI Summit UAE 2025, and was praised by the founder of G2, the world's largest B2B software review platform. I also designed a mixed reality surgical planning tool for Microsoft HoloLens that changed 70%+ of participating surgeons' pre-operative decisions.
How an AI product design engagement works
- Agree the metric. We pick the one number the AI feature should move, such as activation, task completion or conversion, before any screens.
- Map the model's real behaviour. What it gets right, what it gets wrong and how long it takes. Interfaces fail when they are designed for the demo instead of the edge cases.
- Design the trust layer. Explanations, sources, confidence, editable outputs and a clear way to undo or hand over to a human.
- Prototype and test. Clickable Figma prototypes with realistic, imperfect outputs, tested with real users before engineering commits.
- Hand off and measure. States, tokens and LLM error cases documented for engineering, then iteration after launch based on what users actually do.
What you get
- Flows and wireframes for the AI journeys that matter most
- High-fidelity UI in Figma with a clickable prototype
- Loading, streaming, low-confidence, empty and error states
- Conversational UX: prompts, suggested actions and fallbacks
- Human-in-the-loop patterns for review, edit and approval
- New components added to your design system, ready for handoff
Is your AI feature earning its place?
Send your product URL and the one metric you want the AI feature to move. I review it personally and reply by email with three specific, prioritised fixes.
- 01A written review, not a sales call
- 02Three fixes, ranked by impact on your metric
- 03Yours to keep, whether we work together or not
I'll look at your product with that metric in mind and reply to your email with three prioritised fixes.
Questions about AI product design
What is AI product design?
AI product design is designing the interface between people and a model: how users ask, how results are shown, how uncertainty is explained and how mistakes get corrected. Good AI design makes the output useful and trustworthy, not just impressive in a demo.
Why aren't users adopting our AI feature?
Usually because they can't predict what it will do, don't trust the result, or can't see where it fits in their workflow. Showing sources and confidence, starting from a useful default instead of a blank prompt, and letting people edit the output tend to help more than a better model.
Do you only design chatbots?
No. Chat is one pattern. I've designed fit scores, comparison and recommendation views, structured builders that use AI behind the scenes, and agent workflows. The right pattern depends on the task, and often it isn't a chat box.
Can you work with our existing design system and engineers?
Yes. I work inside your Figma files and design system, add the new AI states and components to it, and hand off with annotations and edge cases so engineering can build without guessing.
How do we get started?
Send your product URL and the one metric you want the AI feature to move. I'll reply with a free 3-point teardown, and if it makes sense we scope an engagement from there.
Where are you based?
I'm based in India and work remotely with teams worldwide, including clients in the USA, UK, Germany, Japan, Russia and the UAE, with overlap across US, European and Middle East time zones.