Case Study
Warmstart.ai
End-to-End Design
Email platform generated by AI to build businesses and manage customer relationships.
PROJECT OVERVIEW
Warmstart.ai
Warmstart.ai is an AI-generated email platform that allows business owners to build their businesses and manage customer relationships. Warmstart.ai was developed to replace cold, generic emails with messages that resonate, drawing insights from social interactions and previous communications.
This platform was created to integrate with a user's Gmail account, analyzing interactions with contacts to predict likely responders based on user-defined criteria. Using this data, the technology generates email drafts for each contact, capturing the user's tone and incorporating insights from social media and past interactions, streamlining the email drafting process.
PROBLEM EXPLORATION
It takes forever to write personalized emails!
Before developing this product, the founder, Dave, created an AI query platform called Questy. Unlike other AI tools, Questy stood out by providing every answer with links to credible references from across the web. He wanted to share this new tool with his network and attract more users, but he faced challenges in identifying the right contacts to reach out to and crafting compelling emails that would prompt a response. All of this took a lot of his time!
That’s when he came up with the idea of using the tools he’s built to create Questy, and create another tool that could solve this issue he was facing. And that’s when he approached me!
RESEARCH
Explore and understand the pain points users encounter when sending emails within and outside of their networks with the objective of eliciting responses of interest from recipients.
We can generally assume that sending emails with the expectation of receiving positive responses—whether it's about a product or service we are promoting, or networking to find new opportunities—can be somewhat nerve-wracking (..at least in my world it is!).
We wanted to explore these feelings more deeply to understand why people might experience them.
Objectives:
✅ Understand general behaviors around sending emails.
✅ What success looks like for people after sending emails.
✅ Understand peoples’ feelings around AI being a solution to enhance the email drafting process.
USER INTERVIEWS
We aimed our interviews towards individuals who frequently use email outreach for their jobs, networking, or building their businesses.
Total number of participants: 13
Industry/Careers:
Tech
Recruiting
Sales
Designers
Job Seekers
Small Business Owners
KEY INSIGHTS
🙅🏻♀️ People generally feel hesitant and anxious when it comes to email outreach. Rejection hurts. 👎
⏱️ Researching who to reach out to takes a lot of time and can be challenging.
♥️ Personalizing emails increases the likelihood of getting a response back from the recipient.
🗣️ The main goal for email outreach is to engage in conversation with the recipient.
🖇️ Linkedin is the primary tool used for research on people.
🤖 People generally like the Idea of AI doing a lot of the research and email draft writing, but in order to trust the output from AI, there needs to be a level of oversight the user has.
PROBLEM DEFINITION
Users often feel anxious and stressed about email outreach due to the time required for research and personalization. While AI can be a solution to this, users need control over the final output to ensure the messages drafted on their behalf sounds like them.
A solution is needed to simplify these tasks while maintaining user oversight and personalization.
As anticipated, research findings validated the proposed pain points. Moving forward with the design of this product, we identified the MVP considering our bandwidth as a small team and timeline to launch. Dave’s role in this small startup was not only founder, but also front and back-end engineering.
Core features for MVP include:
Sign up/New account creation
Contact scoring
Drafting customized email campaigns
Ability to send drafts into gmail drafts
Dashboard to track email campaigns
**One major feature limitation that Dave mentioned at the beginning was that the only email provider that his technology could work with (for now) is gmail.
USER FLOWS
Based on the features we identified, we were able to define 3 major user flows that included: onboarding, email campaign creation, and email draft creation.
ONBOARDING
Onboarding consisted of two sub user flows:
Sign up process
Profile generation - Connecting user’s Linkedin profile to account.
The purpose of linking a user’s LinkedIn profile to their account was to enable the Warmstart platform to quickly generate a biography for the user’s profile. In a future version of the product (MVP 2.0), the goal is to fully integrate the LinkedIn profiles of the user’s contacts.
SIGN UP PROCESS
PROFILE GENERATION
EMAIL CAMPAIGN CREATION
The original plan for this user flow was to provide users with choices for the type of campaign they want to create. We aimed to offer broad options focused on connecting with people to build relationships.
The next part of this process is where the AI technology curates a list of contacts based on the selection and prompt that the user provides.
EMAIL DRAFT CREATION & PERSONAS
Another component of this process was the option for users to select different personas to adjust the tone of their messages based on the recipient. For example, a user could create a casual, friend-like persona for one message and a formal, professional persona for another. This feature would have allowed users to control and define the tone they wanted to convey in their communications.
LOW FIDELITY WIREFRAMES
USER TESTING
In order to get user feedback early on in the design process, we decided to test an early prototype of our Warmstart user flow.
Dave’s preference in this part of the design process was to get user feedback as early as possible in order to avoid any major design flaws or misunderstandings before we invested even more time into this product.
The user flow we wanted to focus on was the onboarding and creation of an email campaign since that would be the main features of this product.
USER TEST DETAILS:
4 user test participants
2 small business owners
1 senior product manager
1 PhD in environmental design
A low-fidelity prototype was developed using the user testing platform, Balsamiq.
KEY INSIGHTS
👍 Overall user flow was straightforward and simple.
🔐 There is some nervousness about users granting full email access to the platform.
👁️ Emphasis on not wanting anything sent to email recipients without users themselves reviewing the drafts.
💯 Users may want personalized criteria for scoring contacts.
🏷️ Tags for previous email topics could help classify discussions.
🤔 General confusion on using message templates (templates that are pre-written). Some saw it as a great tool to start. Some had the expectation that they’re needing to adjust the template to make it sound like their own.
📨 Users really like the idea of the drafts showing up in their gmail.
INFORMATION ARCHITECTURE
We hit a creative block. We had no idea where to place stuff….Nothing a little card sort can’t fix!
Although we had our basic user flow mapped out for the user test, we got stuck on figuring out general information architecture - what belonged on the dashboard? menu items? There were times where we were going down rabbit holes that made us lose focus on what exactly we were trying to include in the MVP - do we really include analytics as part of MVP? Did we even have the resources from our skillset to pull this off for MVP?
Personally not comfortable moving forward without validation from more data, I suggested conducting a card sort activity on a group of people to steer us in the right direction.
CARD SORT DETAILS:
5 Participants
2 previous user interview participants
1 graphic designer
1 business development consultant
1 leadership coach
Virtual card sort using Optimal Workshop testing platform
KEY FINDINGS
Main menu items identified:
🏠 Home • 📊 Analytics • 📢 Campaigns
✉️ Messages • 👩🦰 Contacts • ⚙️ Settings
DESIGN SYSTEM + BRANDING
We aimed to ensure the branding and design system conveyed Warmstart’s mission to transform business relationships through secure, AI-driven, personalized outreach.
COLOR PALETTE
When I first started brainstorming the color palette, I was heading towards more warm colors that included shades of orange and yellow to emphasize the warmth in Warmstart (hence, having the actual word in the name). However, Dave stressed the significance of the color palette in conveying trust, security, and warmth foremost. He then suggested a blue-dominant scheme with dark and yellow-gold accents for buttons to evoke these emotions.
FONT
LOGO
We aimed to simplify the logo from writing the actual company name out and removing the overlapping “W” icon and switching to something more soft and approachable like a “W” in cursive.
DESIGN SYSTEM
HIGH FIDELITY PROTOTYPE
Prioritizing essential features and being adaptable to resource constraints are crucial for successfully delivering a minimum viable product in early-stage development.
By the time I was designing these screens, we had already attracted interest from our network for beta testing Warmstart. This made delivering a basic product our top priority (a good problem to have!).
We had to scale back some features for our MVP and defer them to a later version, including our analytics page.
MVP still included all the promised features, though with slight adjustments compared to our low-fidelity prototype:
Sign up/New account creation
Contact scoring
Drafting customized email campaigns**
Ability to send drafts into gmail drafts
Dashboard to track email campaigns
✉️ **We've significantly revised our email campaign feature by eliminating the selection of campaign types and email templates. We found that these options were only complicating the process. Our AI technology aims to simplify user experience by minimizing complexity. To achieve this, users now provide a general sample message reflecting their desired tone. The AI learns from this initial input and any subsequent edits users make directly in Gmail drafts.
ONGOING WORK
Warmstart is currently going through beta testing!
Currently, Warmstart has approximately 10-15 beta testers using the product. Since some front-end development is still ongoing, Dave is personally “white-gloving” these testers through the process to ensure that Warmstart emails are successfully delivered to their recipients.
We anticipate receiving results and feedback within the next 1-2 months.
