[AI for Business #7] - Managing the uncertainty of AI
Building transparency and calibrating trust in your AI applications
Dear AI friends,
Trust makes or breaks our relationships - whether it’s with other people, businesses, or products. When it comes to AI, trust becomes especially tricky due to the uncertainty and occasional errors baked into AI systems. You may be working your way towards calibrated, well-earned trust — only for a single unexpected mistake to unravel your efforts and send you back to square one.
In my new article, I show how to design AI systems that support trust building and calibration. From strategic use case selection to thoughtful UX patterns, it offers a practical framework for product builders, innovation leads, and business decision-makers navigating the AI landscape.
What you'll learn:
Why overtrust can be just as dangerous as mistrust
How to design AI applications that integrate seamlessly into user workflows
The importance of framing, transparency, and feedback loops
Real-life insights from deploying AI in automotive R&D
If you’re building or buying AI systems, this guide will help you shape trust as a strategic asset and differentiator, not a hopeful outcome.
📖 Read the full article on Medium
Other insights in this newsletter:
A practical cheatsheet for explaining AI systems and enhancing transparency
Vibe coding - The Silicon Valley trend disrupting software development
Let me know what you think — and as always, feel free to share your experiences or challenges around AI trust-building!
Warmly,
Janna
Building and calibrating trust in AI
In this article, I break down the practical steps to building calibrated trust in AI systems — from choosing the right use case to designing transparency, control, and feedback into the user experience. Whether you're building AI tools or rolling them out across teams, this is your playbook for making trust a strategic advantage.
Cheatsheet: Explaining your AI systems
Every time users engage with your AI system, they create or update their mental model of how it works. If that model doesn't align with reality, trust quickly erodes. Clear, contextual explanations help users build accurate expectations, creating the foundation for transparency, confidence, and responsible use.
Download the cheatsheet here!
Trend insight: Vibe coding
Earlier this year, Andrej Karpathy coined the term vibe coding — a new, AI-powered way of programming that’s more about vibes than syntax:
“You just vibe: you describe what you want, copy/paste, run it, see if it works, rinse and repeat.” — Andrej Karpathy on X (Feb 2025)
With tools like Cursor Composer and Vercel’s v0.dev, users describe what they want in natural language, and the AI takes care of the implementation. This enables fast, intuitive, and often surprisingly creative app development — especially suited for early-stage ideas, rapid prototyping, and solo builders. Thus, in the current batch of the YC, a leading startup accelerator, more than 90% of the startups’ code is written on “vibes” (cf. this video).
I’ve been experimenting with vibe coding for a few weeks now — and it’s incredibly fun. You can design, prototype, and iterate on new apps in minutes. But beyond the excitement, it also opens your eyes on just how much of software development isn’t about writing code. Vibe coding lets you skip the syntax, but not the thinking. It makes you shine if you understand product design, architecture, DevOps, and the full software lifecycle. Meanwhile, those using vibe coding because they can’t code get bogged down in lots of spaghetti code, security gaps, and fragile quick fixes.
As someone who’s worked at the intersection of AI and enterprise systems, I’m also fascinated by what vibe coding can’t (yet) do:
Integrate cleanly into complex architectures
Support long-term maintainability
Coordinate with complex team workflows
I believe we will soon approach these challenges, moving to a world where AI amplifies the creativity and craftsmanship of software developers. I hope you are as excited to learn how this movement evolves and eventually connects with larger software practices. If you would like to feel the vibes yourself, check out this new course offered by Andrew Ng!
⚠️ Corporate caution: Vibe coding doesn't yet play well with legacy systems, collaboration, security standards, or large-scale architectures. But it’s worth watching - in my view, it is one of the first steps towards a new, AI-driven way of building software.