Work About Resume Contact

ehsAI Case Study

AI Document Analysis Made Fast, Intuitive, & Simple

ehsAI - application dashboard using AI to Deconstruct documents, showing list of documents and upload options

introduction

ehsAI is a B2B SaaS platform using AI to enable EHS professionals to process complex compliance documents with speed & accuracy

Full Time UX/UI Designer
2020–2023
ehsAI https://ehsai.com/
Startup acquired by Intelex/Fortive

User Research, User Flows, Information Architecture, Wireframing, Prototyping, Mockups, Usability Testing, Collaboration. Figma, XD, Photoshop, Illustrator

ehsai design for sustainability and saving lives

The Challenges

Digitizing documents for EHS (Environment, Health, and Safety) veterans to work with AI deconstructing & analyzing important compliance data

Disrupting paper and pencil with an AI web app

All EHS professionals deconstruct documents by hand with paper and pencil. Moving to digital would be disruptive as most EHS professionals were afraid of or unfamiliar with AI.

Many types of compliance documentation from all levels of governments

One huge challenge was digitizing compliance documents. EHS legal requirements can come from all levels of government (local, state, federal) in all types of formats. The document formatting can be diverse ranging from digital to typewriter-written to hand-written, including tables, diagrams, and images.

ehsAI - Environment, Health, and Safety document deconstruction by paper and pencil

Strategy

Designing strategy around understanding the challenges of Environment, Health, and Safety compliance data

Three key questions that informed my design strategy:

Research

Discovering user preferences & contextual dependencies

Personas - who will need this product?

Research indicated four primary EHS personas. Most would want to use the ehsAI app and some would want to purchase the ehsAI app for their EHS team:

ehsAI - a collection of 4 user personas from EHS compliance industry
ehsAI - User Experience, Research phase, User Feedback showing varied terminology and interpretations

User feedback discovery reveals varied terminology and interpretations

While interviewing customers and potential customers it became clear that deconstructing data could be difficult because:

AI algorithm modules have sequential dependencies

The AI was broken down into modules that depended on the previous module. This meant having the users work with the AI in steps module to module.

Competitive analysis forewarns of failure

Competitive analysis showed that what we were doing had been attempted before, although all attempts had been abandoned due to the complexity of the problems. Friction points were:

Our team believed we could forge ahead with innovation and creativity using AI to solve many issues.

Planning

Giving the users a clear path to their most used feature

Single document deconstruction is the red route

Red routes are the primary tasks the users want to complete the most often. For the ehsAI app the main red route was the single document upload and analysis flow. EHS professionals preferred to process one document at a time from start to finish. Other tasks like document management were secondary in importance.

A user flow to keep the task simple

Client feedback affirmed the single document upload and analysis flow to be the most important. Later more tasks and features were added to the user flows.

ehsAI - a collection of 4 user personas from EHS compliance industry

Ideation, Problems, Iteration

Establishing design direction and solving the challenges of working with AI

Four primary design patterns

After generating different ideas about grid and screen structures, I settled comprising the user interface of four primary design patterns. This informs the design of each new template and keeps a consistent feel for the user as they navigate throughout the app.

ehsAI four main design pattern templates

Wireframes to inform the placement of features

Wireframes were designed to help visualize the information architecture. I prioritized the red routes and organized the workspace of the review steps in way to keep as much of app UI out of the way of the compliance document view.

ehsAI - wireframes

One difficulty was marking the classification words, phrases, and sentences

The document journey went from a customer upload through an OCR (Optical Character Recognition) text extractor into the ehsAI review tool stepper. This meant the ehsAI app was rebuilding the entire document from layout and text to formatting in order for the app, the AI, and the customer to work together analyzing the data for compliance requirements and other important details.

Creating UI for marking the classification of the document sentence text was challenging since we needed a way to show the various text classifications. Labels off the page were used to help with accessibility for the colour blind. Initially we used highlighting to show text classification but as we merged AI steps within the app we settled on underlining for sentence classification and highlighting for word and phrase classification.

ehsAI - Early Iterations on Classification and Sentence Distinction Ideas and a solution

Finding creative ways to interact with columns

For the most part the AI could accurately detect column text structures. However, for the times when it could not, I needed to come up with UI that helped define columns and accurately portray the reading order of the columns. I created a cutting tool to allow the user to slice up columns both vertically and horizontally and numbering UI to determine the reading order.

ehsAI - AI column detection, suggestions, and user editing UI

User Testing and Feedback

Talking to users & getting feedback

Remote user testing

Remote user testing was done using online meetings using screen sharing that were recorded. Some key insights revolved around the Hierarchy Step where the users were assessing the AI evaluations of the document's citation hierarchy. Three main pain points were synonymous across the testing observations and user feedback:

ehsAI - User Feeback for citations
ehsAI - User Experience, Research phase, User Feedback showing varied terminology and interpretations

Quantitative usability testing

Our team gathered quantitative usability statistics to observe various tasks throughout the ehsAI app. Some key metrics involved the length of time on each of the Review Steps where the user works with the AI to define layout structures, text classifications, citation hierarchy, and citation conjunctions. The statistics corroborated the remote user testing that users were spending the most time on the Hierarchy Step. I worked with the customers and our team to come up with hierarchy lines to help visually show citation depth relationships.

Surveys and feedback sessions

For our longer term users we wanted to hear their thoughts on features they thought would improve their experience. Using Survey Monkey our team was able to gather insightful feedback about feature request. Notable requests like:

The Results

Designing a product where customers gain cost and time savings while growing their trust in working with AI

There were three primary strategies that informed my design to make it successful:

These three primary strategies are the goals derived from the original three key questions.

Screens

ehsAI

The Dashboard

The dashboard features three types of document imports as well as the paginated documents list. Users can also filter their documents list and take action on each document.

The Changes screen

The Changes screen features a preview of new versions of previously imported documents giving the user a way to isolate and evaluate changes made to documents, such as the eCFR (Electronic Code of Federal Regulations).

The History screen

Whenever a previously processed document gets a new version released by law makers a history of versions will be created for it and this will be managed in the History screen. Here the users can clearly see the version history and access any of the document versions for their compliance needs.

The Preferences screen

The Preferences screen allows users to adjust settings for preferences like Desktop Notifications, this allows the users the choice to receive a notification or not when the AI is processing between review tool steps.

The Filter screen

Filtering feature is accessed from the dashboard and allows the user to search their documents list by various filters in any combination. For example, this comes in handy for audits where the user needs to reference documents that have been processed in the past.

The Conjunctions screen

The Conjunctions screen is part of the review steps where the user works through the document with the AI to evaluate document text classifications. In this step the user must ensure each parent requirement has been labeled as desired.

Style Guide

Product design needs a solid foundation of well-defined branding

Although the ehsAI logo was established before I joined the team, I expanded the branding by defining and adding the mission statement, values, tone of voice, character, app colours, typography, and custom icon set to keep the experience of the brand cohesive.

ehsAI branding logo, colours, typography, type styles

Animations

Adding some colour and fun to compliance document processing wait times

We were honoured to have the very talented designer Nikoo Farvardin intern with us while she was finishing her studies at Emily Carr University of Art and Design. She created a library of illustrations that radiated with colour and spunk bringing out the fun part of our brand. We had two points in the user flow of working on a long documents where the processing wait times could sometimes be lengthy. Later I animated her two of Nikoo's illustrations: one for preparing the document for the review steps and a second for processing the document at the end of the review steps.

ehsai loading animation for preparing your document ehsai loading animation for processing your document

Value and Impact

The highlights of going from a concept to happy customers using a functioning B2B AI SaaS product in six months and beyond

From 0–1 in 6 months

Our whole team was instrumental in quickly bringing to life a disruptive technology that upends paper and pen with AI/ML by simplifying complex AI interactions and quickly yielding accurate results. We went from a concept to fully functioning B2B SaaS in six months. From there we continued to improve and add features.

Time and cost savings of 80%

Designs that allow Environment, Health & Safety (EHS) professionals to work with AI to deconstruct regulations and permits into useable compliance information saving significant time and money while being more accurate.

Accomplishing something that had never been successful previously

Although others had tried and failed, the team at ehsAI successfully built an app that could upload and import compliance documents and give EHS professionals a fast and accurate output. Thanks to top notch leadership and a dedicated hard-working team.

Companies using ehsAI

Used by happy EHS veterans at companies like Koch Ag Energy Solutions, ALL4, Mosaic, a big tech company, and more.

Customers growing trust of AI

One of our proudest achievements was having customers who were skeptical and leary of AI grow to the point of trusting and loving to work with AI.

Customers recommending ehsAI

Our question to customers was, "Have you ever discussed or recommended ehsAI to others?"

Yes, to other team members throughout Asia
Yes, told others in the company. I was asked to do demos of ehsAI for the standardization council in the company

User quotes

There's a sense of satisfaction and gratitude when customers validate the goals

Thanks to all our customers for contributing to our journey of designing and building. And thanks to my awesome teammates for the valuable feedback, suggestions, and inspiration. Here is some of the praise we got from our users.

The time-savings is amazing
The whole process feels like a video game, it’s entertaining, it’s fun
Thanks to all the amazing folks in dev/design for making ehsAI such an easy tool to use! Great work everyone!
Love the ability to get back into the app and find it easy to work with and use, even after an extended break
The newly added lines in the hierarchy step make it easy to follow
Intuitive and simple

ehsAI awards for product, business achievement, & innovation

These are team awards

The effort of the whole team contributed to the success of ehsAI as a business and as a software service to our customers.

Top Product of the Year from Environment + Energy Leader (2021). "A panel of independent judges commended ehsAI for its innovative approach to simplifying the complexities of EHS compliance with real-world applicability and proven impact of AI. According to the judges, “This product simplifies the time-consuming exercise of compliance with a myriad of permits. The usefulness of this product is on multiple levels, from reducing risk to saving time and labor.”"

EBJ Business Achievement Award (2021). "ehsAI (Vancouver, BC) for a game-changing year for its ehsAI platform, including global teaming agreements, industry recognition, and being acquired by Intelex, a subsidiary of Fortive. The company’s ehsAI platform in the cloud is revolutionizing the way EHS compliance is managed. ehsAI’s algorithms allow EHS professionals to deconstruct any EHS compliance document, no matter how messy and complex, into succinct compliance requirements in minutes. ehsAI has four patents pending."

Big Innovation Award (2022). For products that are bringing new ideas to life in innovative ways and have made tremendous strides at improving the lives of their community. ehsAI wins for its innovative compliance platform.

ehsAI Top Product Award 2021
ehsAI EBJ Business Achievement Award 2021
ehsAI Big Innovation Award 2022

Thank you

Thank you for taking the time to read through this case study. If you have any comments or questions about the case study feel free to reach out.