Building the Switzerland for Voice and Conversational AI
Welcome back to The Bluejay Times.
Last week, Bluejay opened its doors to everyone. Now, anyone building voice or conversational AI agents can access world-class evals for free without having to enter a credit card.
What happened next was interesting; the teams that showed up were not just the ones we expected. Indie developers and early-stage startups came in, and even enterprise teams decided to hop on and try it for themselves.
That is the thing about lowering the barrier to entry: the teams that need good testing the most are rarely the ones with time to wait for it.
The bar for production readiness in Voice AI and Conversational AI is rising fast. The teams shipping with confidence are the ones testing rigorously from day one.
At Bluejay, we are building a Switzerland, a neutral player that builds trust between humans and AI systems. That mission does not change whether you are a solo developer signing up for the first time or a team running millions of Voice and Conversational AI interactions a day.
The bounty is still live. We built three voice AI agents, each one hiding three of the most common issues we have seen break agents in production.
One person has already found the first issue and will be awarded $2,000. There are still two agents left with $4,000 on the table. Go try it at getbluejay.ai/debug!
As a recap:
- Bluejay is a testing and monitoring platform for Conversational AI agents. Companies ranging from Fortune 10 enterprises to fast-growth startups in the Silicon Valley use Bluejay to make sure their voice and text agents work in production (monitoring) and development (testing) environments.
- Our team, now ten strong, works around the clock to make sure your agent behaves when talking to customers.
- This newsletter is 100% human written. It always has been, and it always will be. Ask yourself about what you are consuming. If the writer hasn't read it, why should you?
Give Humanity the tools to trust Artificial Intelligence.
Announcements
Here is what happened at Bluejay last week:
- Bluejay went self-serve. Sign up, add credits, and start testing at getbluejay.ai!
- The Voice Agent Bounty is still live. One issue has already been found and claimed for $2,000. Two agents remain with $4,000 still up for grabs. Go try it at getbluejay.ai/debug!
- The team pushed 84,314 lines of code to production this week to make Conversational AI more reliable!
Join the Bluejay Community
We are building a Slack community for everyone using Bluejay, whether you are just getting started or running thousands of simulations a month. It is the place to ask questions, share what you are building, get early access to new features, and stay up to date on everything happening at Bluejay.
If you want to be part of it, the link is below. Come say hi 🐦
https://join.slack.com/t/bluejaycommunity/shared_invite/zt-442xk85k5-Qkbq65wVOxK~1FHU5OadEA
Behind the Build
Big week for the engineering team. Here is what shipped:
- Build agents with Bluejay AI. You can now import diagrams and IVR exports as editable Agent Workflows and auto-generate test plans and simulations from reviewed QA checklists.
- A brand new onboarding experience. Signing up now drops you straight into Bluejay AI as the default first experience !
- Metrics app store. Browse a metrics library and wire metrics into your agents from a shared catalog.
- Deep Reports. Agent-generated, verified customer reports plus a redesigned branded four page executive summary PDF for simulation runs.
Feature Spotlight: Red Teaming
Most security testing for voice agents is manual. You think of a few attack scenarios, try them out, and call it done. The problem is that real attackers do not work off a fixed script. They probe, adapt, and exploit what they find in real time.
Red Teaming on Bluejay works the same way.
Point it at your agent, launch a test, and that is the entire setup. It starts by probing for vulnerabilities, then launches attacks based on what it finds, adapting as it goes rather than running through a predetermined list.
Once it is finished, the report covers three things. Pass or fail results on every OWASP LLM Top 10 category, each backed by the real conversations that triggered them. Your MITRE ATLAS mapping, showing which attacker techniques actually landed. And a plain English breakdown of what happened and what to fix.
No security background required. Just run it and read the report.
Run Red Teaming today at getbluejay.ai.
That's all for now. I'll see you next time!
Azfar Khan
Storyteller @ Bluejay
