Test Like the Real World
Welcome back to The Bluejay Times !
I have spent most of this week going around downtown San Francisco trying different coffee shops and doing deep work. Different neighborhoods and vibes, but the same thing everywhere I went.
Espresso machines running in the background, music playing and people having conversations at the table next to me, doors swinging open every few minutes letting in the sound of the street outside.
Sitting in one of those shops this week, something crossed my mind. The people calling your voice agent are never calling from a perfect situation and setting where everything is quiet. They are calling from exactly the kind of places I have been sitting in all week.
- A customer calling a coffee shop to ask about catering is probably in their car with the windows cracked.
- A patient calling to book a medical appointment might be walking down a busy street trying to multitask.
- A driver calling a logistics dispatch line could be on the highway with wind noise coming through the cab.
The background noise is not an edge case that occasionally shows up. It is the default condition that most real calls happen in. And yet most simulations never account for any of it. They test whether your agent works in perfect conditions and then send it out into a world that looks nothing like the test environment. The gap between those two things is exactly where agents start to quietly fail in ways that are hard to diagnose because nothing looked wrong in testing.
That is the problem that custom background audio was built to solve.
With Bluejay you can now upload your own background audio and layer it directly into your simulations. Street noise, restaurant chatter, an espresso machine running in the background, whatever your real customers are actually calling from. Your simulations now sound like the real world because they are built from it.
Read about it in our feature spotlight below!
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
Heres what happened at Bluejay last week:
- Rohan sat down with Kwin Kramer, co-founder of Pipecat and Daily, to talk about the infrastructure layer powering the next wave of voice AI!
- Bluejay has officially moved into its new office in SOMA, The Sky. More on this soon!
- The team pushed 72697 lines of code this week to make Conversational AI more reliable!
Skywatch Podcast Episode
Inside Pipecat: Kwin Kramer on Building the Infrastructure Layer of Voice AI
This week Rohan sat down with Kwin Kramer, co-founder of Pipecat and Daily, to talk about where voice AI infrastructure is actually headed and why most people are still underestimating it.
From building gestural interfaces that ended up in Minority Report, to creating the most widely used open source framework for voice agents, now standardized by NVIDIA and AWS, Kwin shares what it actually takes to bet on real time infrastructure before anyone else believes in it.
A few things from this conversation worth paying attention to:
- In summer 2023 Kwin was showing voice agent demos to anyone who would watch. Every single person had the same three reactions. Whoa. That cannot be real. And then I do not know if I want to talk to an AI. He thinks we are still not fully past that third one.
- The big labs are building reasoning models that require too many thinking tokens to work for real time voice. The next wave is open, customizable, low latency models built specifically for these use cases.
- Continual improvement is the 2026 problem. Closing the loop so your agents get better automatically from production data is the next frontier nobody has fully solved yet.
Full episode on Youtube and Spotify now!
Behind the Build
- Red teaming is now in closed beta. It lets you simulate adversarial callers attempting to break your agent through social engineering and prompt injection, so you can find the weaknesses before someone else does.
- We also shipped a cloned voice feature this week. You can now upload your own audio samples and use a cloned version of that voice directly on your digital human. Head to the same Behavior section, upload your samples, and your cloned voice will be ready to assign. The more specific your digital humans sound, the more realistic your simulations get.
Feature Spotlight: Custom Background Audio
Bluejay simulations are built to replicate the real world from the start. Real customers do not call from quiet rooms. They call from cars, restaurants, waiting rooms, and streets. The background noise is always there, the audio is never perfect, and the conditions are never controlled. That is the reality your agent is being deployed into and that is exactly what your simulations should reflect.
Custom background audio gives you even more control over exactly what that sounds like. You can now upload your own audio and layer it directly into your simulations so that the conditions your agent gets tested to match the conditions it will actually be used in.
Whatever environment your real customers are calling from, whether that is a busy hospital waiting room, a highway with wind coming through the cab, or a restaurant with music and conversation in the background, you can now replicate it inside Bluejay.
Head to Digital Humans, then Behavior, then Add Custom Audio and you are good to go.
Give it a try on your next simulation and let us know what you think!
That's all for now. I'll see you next time!
Azfar Khan
Storyteller @ Bluejay
