Chatbot-like assistant that gives customer-facing representatives real-time access to the knowledge and content they need, right where they work.
Spekit AI Assist transforms how teams access critical information, eliminating time-consuming searches and fragmented knowledge bases. This empowers sales teams to focus their energy where it matters most: building meaningful customer relationships and accelerating deal closure. By delivering instant, contextual knowledge access, we're not just streamlining communications—we're fundamentally enhancing sales effectiveness and driving revenue growth.
Drove end-to-end product design by translating complex ML capabilities into intuitive user experiences—from initial model selection and architecture design through iterative testing, design ideation, user research, and successful beta deployment.
Following the general availability (GA) launch of AI Assist, we implemented comprehensive tracking systems to rigorously measure user engagement. Our metrics revealed significant traction, with a compelling 71% increase in questions asked month-over-month (MoM) that validated the feature's value proposition and demonstrated strong market adoption.
1. Adoption metrics 3 months after launch
2. users who sent a message via AI assist at least once per week
Searching for internal information or tracking down colleagues for help with specific tasks makes up nearly20% of an employees workweek.Âą
Traditional knowledge management systems are anchored around the 'destination paradigm' - requiring employees to pause their work, identify a need, navigate to a separate system, and search for answers.
Spekit aims to disrupt this inefficient model by bringing critical knowledge, information, and content directly into employees' natural workflows, eliminating the need to context-switch and ensuring they have instant access to the resources they need, precisely when they need them.
1. The social economy: Unlocking value and productivity through social technologies; McKinsey Global Institute, july 2012
As AI emerged as an essential market feature, we needed to balance speed with strategic intention. During discovery, I analyzed competitor AI tools and rapidly sketched concepts for team review.
We explored ways to seamlessly integrate AI-powered features into users' daily workflows. Our ideas included embedding personalized suggestions directly within Gmail's composition interface and developing an intelligent chatbot to provide contextual recommendations in real-time.
A thorough competitor analysis revealed key differentiation opportunities in the AI tool market. I documented product features through screenshots and identified how to integrate new AI capabilities with existing functionality to maximize user value.
In the solution phase, I rapidly translated discovery insights and feedback into a prototype. This enabled quick validation of our proposed direction and faster handoff to engineering.
After evaluating multiple solutions, we determined that an AI-powered chat interface would deliver the most value. I collaborated with our engineering team to develop a proof of concept, enabling us to quickly gather user feedback and iterate based on real-world usage. This rapid prototyping approach helped us maintain our competitive edge in the market.
The MVP featured a simple chat interface that overlaid on any webpage. The clean UI focused on creating an intuitive, streamlined experience with only essential features.
AI Assist modal
AI Assist response
Using Airtable, I transformed the beta feedback into visual charts that clearly communicated user behavior patterns and product opportunities to stakeholders.
Working closely with the Product Manager, we conducted over a hundred user interviews to gather early feedback on our proof of concept. This process validated our solution while establishing valuable relationships with current and potential users.
I documented all interviews in our research repository, including recordings and detailed notes, enabling efficient analysis of user feedback. This systematic process ensured that we had a consistent framework for research documentation across the full lifecycle of both alpha and beta programs.
Feedback theme
Our interviews revealed that most users struggled to distinguish between the search function and AI Assist feature. Due to this confusion, users defaulted to the more familiar option—search—rather than exploring our AI capabilities.
User are uncertain about the difference in using traditionalsearch vs AI Assist.
When having the search bar and AI Assist button next to each other, users gravitated towards the search bar as their first option.
To push users away from their current search habits, they need a reminder that another tool exists that can better surface what they're looking for.
After meticulously analyzing alpha program feedback, we strategically refined our approach, selecting a focused cohort of 13 highly engaged customers. This targeted beta program was designed to rigorously validate our product and design hypotheses, providing a precise mechanism for measuring our potential success and gathering actionable insights.
User challenges
1. When ASKEDÂ "WHATÂ CHALLENGESÂ DOÂ your customer-facing reps experience that might prompt them to use spekit Ai assist?" in an open-ended format
Feature agreement matrix
Based on the agreement matrix, Spekit admins need to have AIÂ Assist search for answers to sales and/or product questions (85.7%) as well as support or troubleshooting information (71.4%).
They also feel that it would be nice to have AIÂ Assist recommend content to send to a prospect (71.4%) and save chat history/specific answers for future reference (64.3%).
However, they don't need AI Assist to answer questions about a lead, prospect (64.3%) or active deal (57.1%).
Although we received constructive feedback around how we could improve the AI Assist experience, we received a lot of customer love as well. The positive feedback we got helped further validate that we were solving the right problems. Â
"It [AI Assist] is really good for that information that you need quickly when you're in the middle of a conversation."
Account Representative, LocumTenens
"I find myself finding new resources with AIÂ Assist. Now, with AI Assist and the recommend questions and linked sources, I feel like information is more easily accessible within Spekit."
Team Lead, iHeartMedia
"Now, even if I'm not familiar with the content, I can easily put in a question and it [AI Assist] will find something relevant for me or at least guide me."
Team Lead, iHeartMedia
Through research conducted in both the alpha and beta programs, we identified several key areas for enhancement.
We strategically integrated the recommendation feature into our extension's home page interface, optimizing existing space to enhance content discovery. By prioritizing recommendations in the visual hierarchy, we ensured users could immediately access contextually relevant information upon launch, maximizing efficiency while maintaining a clean, intuitive experience.
To gather continuous data on our recommendation algorithm, we implemented a multi-tiered feedback system that enabled users to provide both general input and granular content-level responses. This strategic approach to user feedback collection helped the team determine what was working and what wasn't when deciding what could be improved in the experience.
It's important to recognize that not every design concept makes it into development. I thought it would be fun to showcase some of my ideations that utlimately got moved to the Figma graveyard. Enjoy!
Full extension redesign
Responsive extension overlay
Have an opportunity for me? Get in touch so we can create something great together.
