P.
Product Designer | AI-powered web & Excel tool for document analysis and CSRD reporting
Simplifying CSRD complexity with AI
MVP design across web app, Excel plug-in, and mobile
Product discovery and user interviews
Cogny is an early-stage startup building AI-powered tools to analyze documents and generate CSRD reports. I joined to strengthen the UX and design MVPs across both an Excel add-in and a Web App, making the product accessible to different user profiles.
Sector
Challenge
Make complex AI-powered document analysis intuitive and accessible
Role
Product Designer (hands-on) from user research to high-fidelity prototypes
Outcome
Dual MVPs — Excel add-in (operational workflows) + Web App (reporting & adoption)
Impact
Reduced time-to-value and improved early adoption across dual platforms
Pain to solve
Research key insights w/ audit trail
Consultants in large firms like the Big Four often need to sift through hundreds of pages to find specific information in contracts, reports, or audits. Traditional tools like “Ctrl+F” are not powerful enough to handle this workload.
Solution
Blend Agentic AI with an excel plug-in
Cogny was created to solve this problem with AI-powered document search. The first MVP took the form of an Excel add-in, giving analysts a familiar environment to ask questions and instantly retrieve relevant excerpts.
Product
Ready to review reports
To reach a wider audience and simplify adoption, Cogny then expanded into a standalone web app. This version allows users to generate structured CSRD reports directly, using automatically suggested queries to guide their analysis.
As company had already raised initial funding. The priority was to :
Validate a first MVP
Consultants in large firms like the Big Four often need to sift through hundreds of pages to find specific information in contracts, reports, or audits. Traditional tools like “Ctrl+F” are not powerful enough to handle this workload.
Improve adoption
Convert early testers into recurring customers by understanding their main points using the current version of Cogny, i.e. the time to product actionable results
Transform complex AI capabilities into simple, intuitive workflows
Cogny was created to solve this problem with AI-powered document search. The first MVP took the form of an Excel add-in, giving analysts a familiar environment to ask questions and instantly retrieve relevant excerpts.
First 30 days: Onboarding
Immersed myself in the product through user interviews, feedback sessions, and journey mapping
Collaborated with founders and developers to identify friction points in the Excel add-in “core loop”
Small first iterations
Months 1–3: Fix the core loop
Focused on the Excel add-in MVP: simplified the journey so users could generate results more easily, iterate until answers felt right, and check sources directly
Delivered wireframes and prototypes using Microsoft Fluent, tested in iterative 2-week sprints
Contributed to shaping the product vision in close collaboration with the team, before a CPO joined
Following months: Pivoting to a webapp
Identified Excel as a barrier to entry: adoption limited to junior profiles, with added value not visible to senior users
Designed user flows and prototypes for a more accessible web app, centered on CSRD report generation and scalability
1. Excel plug-in
Cogny is an early-stage startup building AI-powered tools to analyze documents. Its initial value proposition was clear: help consulting teams (e.g., Big Four) extract relevant insights from hundreds of pages in seconds — an enhanced Ctrl+F powered by AI.
However, when I joined, the core loop — the journey to reach AI-generated results — was too fragmented. Users had to:
Write their own questions from scratch in Excel cells
Upload documents
Choose a search mode (batch vs per document)
Choose a response format ... before they could see a single result.
Iteration was also painful: if a user wanted to refine a question, they had to regenerate the entire report, creating frustration and friction.
Problem Statement
How might we reduce the number of steps to reach first results, and create a smoother way for users to iterate directly on answers?
Objective
Shorten the time-to-value and deliver a quick “wow effect”
Improve adoption by making the first experience simple and rewarding
Enable iteration at the level of each answer, without regenerating the entire report
My Role
As a hands-on Product Designer, I:
Conducted user interviews to identify pain points
Mapped the existing journey to visualize friction points
Facilitated workshops with developers to explore simplification strategies
Designed successive MVP iterations in Figma, tested in short 2-week cycles
Process & Methodology
Step by step
Started with a detailed journey map of the existing flow
Iterated on simplified flows where results appear sooner
Introduced AI-assisted query suggestions to remove the barrier of writing from scratch
Prototyped interaction patterns allowing users to chat with individual results (cells) to refine them without re-running the full search
Outcome & Impact
Reduced the number of steps to results
Delivered a clearer, more intuitive flow: import documents → suggested queries → run search → results → iterate on answers via chat
Set the foundation for later integration into the Excel add-in and the web app
Key features designed
Redesigning the core loop was not only about reducing steps — it was also about creating new features that balanced usability with AI innovation. Working closely with developers and early users, I designed functionalities that made the experience both intuitive and powerful, from onboarding to reporting.
1. AI-assisted query suggestions
To avoid the “blank page” effect, users can start from predefined query templates (Explore, ESG Benchmark, Due Diligence, Contract Analysis). Each template acts as a starter kit of questions. Before launching, users can freely edit, add, or remove queries, balancing guidance with flexibility.
2. Streamlined document import
Users can select documents from a public library or upload from their computer. Each document is displayed as a column in Excel and automatically paired with the queries in rows. Documents can be added or removed instantly, with the spreadsheet updating in real time.
3. Iteration via chat on each result
Beyond simply viewing AI-generated results, users can consult the original document and exact excerpt that produced the answer. They can then chat with the AI directly on that cell, refining the output without regenerating the entire report.
4. AI Reviewers
A dedicated “Verify” feature allows users to call on specialized AI agents (Climate, Biodiversity, HR, Supply Chain, Governance). Depending on the selection, these reviewers highlight results that may contain inconsistencies or imprecisions — particularly useful when working with data-heavy documents.
5. Draft report generation
Users can automatically generate a draft report summarizing all table data into a structured narrative. This transforms raw results into a cohesive, CSRD-oriented document, ready for refinement and client-facing delivery.
Excel plug-in
Webapp MVP
2. Web app V2
While the Excel add-in addressed very specific workflows, many users were either not using Excel or did not have the seniority or technical skills to install and operate a plug-in. The team therefore decided to extend Cogny into a standalone web app, making the experience more accessible, scalable, and visually powerful.
Unlike the constraints of Excel, the web format allows Cogny to take advantage of the full screen and to start directly from the end goal: generating a structured CSRD report based on automatically generated queries.
Objectives
Facilitate adoption by targeting a broader range of profiles beyond Excel users
Provide an easier onboarding (no installation, no technical setup required)
Deliver scalability and richer experiences through a dedicated interface
Enable direct CSRD report generation with multiple templates
My Role
Same transverse Product Design role as on the Excel add-in
Hands-on work on user flows, wireframes, and prototypes in Figma
Regular feedback loops with early users through interviews and demos
Close collaboration with founders and developers in 2-week cycles
Methodology
Short design sprints (2 weeks) with weekly syncs with developers
User feedback integrated continuously to validate flows and templates
Iterative design process using a design system for consistency
Key features designed
1. CSRD Report Templates
Users can generate a structured CSRD report instantly, choosing from multiple templates (e.g. ESG, Due Diligence). Each template defines the queries that will power the analysis.
2. Transparent Queries
Behind every report, users can view the queries that were generated and executed. This provides transparency and gives them the option to refine or add their own.
3. Enhanced Accessibility & Scale
The web app removes technical friction (no plug-in install, no Excel dependency), making Cogny easier to adopt across a wider variety of user profiles, from analysts to managers.
Key design learnings
AI needs framing to be useful
Users rarely know how to ask the “perfect” question from scratch. Providing guided queries and templates proved essential to overcome the blank page effect and unlock value quickly.
Balance between automation and control
Automatically generating reports created immediate value, but users still needed transparency on the queries behind them — a balance between efficiency (automation) and trust (control).
Designing hybrid flows requires a new mindset
Building Cogny’s experience meant combining classic product design flows (onboarding, navigation, iteration) with AI-driven interactions (dynamic queries, chat with results). Defining user journeys that mix these two logics was a new design challenge.
Accessibility drives adoption
The move to a web app wasn’t about replacing Excel, but about lowering barriers to entry and making Cogny available to a broader range of users, regardless of their technical setup.
Contact
Collective
Medium