Yousign — Building AI Features
Shaped Yousign's AI design strategy, defining how AI integrates into the product experience.
AI without a plan is noise
In 2025, every B2B SaaS player, including Yousign's competitors, started shipping AI-assisted features. The pressure was on. But shipping fast without a clear direction risks inconsistency, mistrust, and wasted team effort.
The challenge wasn't just to build AI features, it was to build the right ones, the right way.
No shared AI vision
Everyone had opinions on AI: product, engineering, legal, design. But there was no shared direction, no principles, and no clear starting point.
Our squad mission: Built Yousign's AI design strategy from zero
4 parallel workstreams
Press release workshop
With my PM we build a workshop session to align product leaders around a shared vision: writing the press release before building the feature forced us to clarify the user value and make early decisions on scope.
Miro support for Press release Workshop
AI Features concept sheets
Each AI concept was documented in a concept sheet covering the problem, the AI intervention, the expected value, and the risks. It was a way to have a fantastic visual support to discuss internally and directly with user during user interviews.
Early concept directions used as conversation tools to align internal teams and validate assumptions with users.
User interviews
We ran discovery interviews to understand how users currently use AI in their process, what are their expectations about AI for Yousign, identify fears or doubts.
These insights directly shaped our AI features priority & recommendation and the way we communicate AI output to users.
User interviews - One pager synthesis
AI Branding workshop
To define how Yousign's AI should look and feel, we ran a internal branding workshop exploring tones, visual directions, and naming.
The goal was to make the AI presence feel trustworthy and coherent with Yousign's strategy: Digital Trust.
The branding direction evolved through multiple iterations before reaching internal alignment and validation.
Signature level recommandation
Based on user research and Yousign's vision for AI and Digital Trust, we identified signature level recommendation as the right first AI feature. Users struggled to choose the appropriate signature level, making it a strong opportunity to improve both clarity and trust.
AI SL reco user flow V1 after several iteration with my squad
Figma prototype to illustrate behaviors and animated effects
Beta launch & AI credit definition
Before the full release, we ran a closed BETA to test the recommendation quality and measure user reception.
We also defined the AI credit model: how many recommendations a user could access depending on their plan as a key lever for both product value and monetization (cost for Yousign).
Consent flow mapping for AI feature activation
Create flow and assets to manage BETA launch and quota mechanism
First AI-powered feature integrated into the product
The AI-powered Signature Level Recommendation was successfully launched and adopted by users, helping them better understand which signature level best fit their document and risk context.
Figures as of April 2026
New upsell entry point in-App
AI-powered signature recommendations created a new in-app upsell opportunity by guiding users toward the right upgrade without disrupting existing flow.
AI Design Principles
The AI guidelines we have drafted serves as a solid foundation that now provides the product team with guidelines for developing AI features within our product.
What I took away
AI is double-edged in trust-sensitive products
AI can be approached in many ways, especially in domains like e-signature: strong recommendations can build trust, while poor or opaque ones can quickly erode it.
Alignment work is the design work
Aligning product, engineering, legal, and design teams early around shared AI principles was a critical part of the project, not a prerequisite to it.
Build AI feature through iteration & feedback loop
We used phased rollouts, closed betas, and continuous evaluation mechanisms to iteratively improve recommendation relevance and user confidence before scaling the experience.