Cypris Q
The AI Research
Assistant for R&D Teams
TL;DR: Cypris had strong proprietary data, but customers still left the platform to think, compare, and write. I helped design Cypris Q to close that gap: a unified AI workspace for R&D, innovation, and IP teams.
The issue was not a lack of data. It was a broken path from data to decision.
Search required too much operator skill
Users repeatedly said semantic search was too broad, Boolean was intimidating, and narrowing down a landscape required too many manual steps. They wanted to ask naturally, not engineer perfect queries.
AI lived in disconnected surfaces
AI Insights, Report Builder, search, and exports all behaved like separate tools. Users had to choose the right entry point first, then rebuild context each time they changed direction.
Trust broke whenever the evidence trail felt weak
For enterprise R&D work, a good answer was not enough. Users needed to inspect sources, verify claims, understand what corpus was used, and decide whether an answer was safe to share internally.
Real work was happening outside the platform
Customers copied results into Word docs, PDFs, spreadsheets, and ChatGPT to refine questions, compare outputs, and create executive-ready deliverables. Cypris had the data, but not the full decision-making loop.
Once Q clicked, the product stopped behaving like a feature and started behaving like a workflow.
0
Average satisfaction across the beta experience survey
0
Of users who triggered Q went on to send a first message
0
Submitted a second prompt, showing multi-turn engagement
0
Submitted a third prompt instead of treating it like a one-off demo
0
Chats completed by one customer on a single FTO-style workflow
0
Speed and professionalism feedback from an enterprise customer during onboarding
I designed Cypris Q around the moments where customers were losing momentum.
Ask naturally instead of constructing the perfect search
I reframed the first interaction around a single conversational input. Instead of forcing users to decide between search, AI insights, or report builder, Cypris Q gave them one place to start with a plain-language question.
Show the evidence while the answer is being formed
To rebuild trust, I designed visible source flows: inline citations, a side-panel source experience, and clearer transparency around what Q was using. The answer no longer felt detached from the evidence.
Let users control the corpus, not just the prompt
A major workflow unlock was giving users control over what Q should reason over: selected patents, papers, collections, uploads, or the full Cypris database. This was critical for both precision and trust.
Turn exploration into a shareable output
Instead of making users choose a rigid report format up front, I designed a flow where conversation came first and report generation followed. Users could explore, refine, then export a polished PDF once the thinking was done.
The story held up in the market.
Mannington Mills
"Overall this rounds out your offering nicely, and effectively guides users through the discovery process"
Mannington Mills
"I know ChatGPT can do this, but sometimes the references are random or not relevant, and more difficult to scan than Q"
Saint-Gobain
"Wow this really looks like your own development, I can tell instantly how it differentiates from ChatGPT"
Marmon Holdings
"That's my favorite part right there" (About selecting specific papers/patents and toggling the full database button)
Marmon Holdings
"Now that I have access to Cypris Q, I'm going to star it on my browser and use it alongside ChatGPT and Grok to compare answers"
Callaway
"It's similar to chatting with GPT or Gemini in terms of visuals and ease of use"
Halliburton
"Continuing to love Cypris Q- it's the first place that I look when conducting research"
RPM
"Q is a much more user-friendly experience that caters specifically to corporate R&D users overall"
Saint-Gobain
After testing additive radio calculations, process comparisons, and patent discovery: "It did a great job across all tasks"
