AI built natively, not bolted on

"AI was the missing link needed to replicate human behavior on real devices. Witbe built it natively, starting with the Agentic SDK, then extending across the REC and Smartgate."

That sentence framed almost every demo at booth W2051. Not AI as a feature on top of an existing tool; AI as the infrastructure underpinning testing and monitoring. Four layers, one platform, running on real consumer hardware:

1 · Connect to the Witbox

Real-device execution. The same Samsung TVs, Apple TVs, Fire Sticks, Roku players, Android phones, and browsers your viewers use. Connected via HDMI, infrared, and network, not emulators.

2 · Operate in REC AI

The Remote Eye Controller. A virtual NOC where teams see, control, and troubleshoot real devices from any browser, and now, drive the fleet through natural language.

3 · Advanced automation with Agentic SDK

Hybrid automation: algorithmic logic plus agentic AI, applied where it adds value. Three agents — Designer, Runner, Analyst — generate, execute, and analyze tests on real devices.

4 · Analyse in Smartgate AI

The observability layer. AI-driven correlation across devices, ISPs, and regions, with full video evidence linked to the exact timestamp of every issue.

Choose how much AI you want to use in your tests at runtime

The most-asked question on the booth: how much AI is in this test? The Agentic SDK’s answer is a dial. Teams choose how much AI drives each test, at runtime, in the same environment.

Mode: Full AI self-driving
AI usage: 100%
What it is for: Exploratory testing, non-technical teams, overnight UI-redesign validation. Zero scripting. Just describe the goal.

Mode: By-design AI steps
AI usage: ~10–20%
What it is for: Non-regression QA. 1,000+ tests across 20+ models weekly. AI handles dynamic content selection and human-like assertions; deterministic logic measures KPIs.

Mode: Recovery & error qualification
AI usage: ~1–2%
What it is for: Where most customers land. Happy path runs algorithmically; AI activates only on the unexpected events: pop-ups, profile screens, OS prompts.

Mode: Pure algorithmic
AI usage: 0%
What it is for: Channel change tests, repetitive zap-time measurement. No AI tokens consumed at runtime.

The economics matter

Token usage is tracked per test, per device, per application - pre-provisioned, no surprise costs.

Exclusive preview of two demos from the booth

Scenario 1. Prompt the REC

An operator types into the REC AI: "Show me a cluster with OTT devices, mobile, and Smart TVs." The REC builds it. Then: "Launch YouTube."

From there, the AI runs the action on every device in the cluster. Different operating systems. Different UIs. Different languages: English on one device, Spanish on the next. Each instance handles its own boot state, dismisses its own pop-ups, picks its own profile. No script behind it. No specific training for YouTube. The same approach works for any application across any device.

Scenario 2. Unbreakable tests

When Canal+ reskinned their homepage for the Superman launch, the entire UI changed overnight. Under traditional automation, every script that referenced the old layout would have broken; every menu, every tile, every navigation path. With the Agentic SDK, the tests adapted at runtime. Same intent. New interface. Tests kept running.

A+E Global Media: from rework to coverage

The clearest production proof showcased at the show came from A+E Global Media, a client willing to speak publicly, with a joint webinar already on record.

Before the Agentic SDK

The scripting team hit a plateau. One engineer was spending most of their time maintaining captures across 5 Smart TVs and 7 applications as the UIs updated underneath them. New coverage was not advancing.

After the Agentic SDK

Broke through the ceiling. 4x'd their test coverage in two months, with no maintenance. Team shifted entirely from rework to expanding coverage and adding new tests.

In the wider Witbe client base, the pattern repeats: around 80% less QA maintenance on Smart TVs in full black-box conditions across any UI change. This was the headline number from Yoann Hinard's BEIT Conference presentation on the day before the exhibition opened.

Recognition: TV Tech Best of Show 2026

The TV Tech jury came to booth W2051 during the show, attended a live demo, and recognized Witbe's AI-native infrastructure with a TV Tech Best of Show 2026 award.
The award covers the full four-layer platform, with the Agentic SDK at its core.

AI in testing and monitoring is most useful when it is hybrid, measurable, and anchored in reality.

What's next?

NAB Show 2026 is over, but the conversations aren't. The team is back with a list of follow-up demos mapped to client setups, and the Agentic SDK continues to roll out across new clients and expanded use cases in the coming weeks.

Thank you to everyone who stopped by W2051, sat through a demo, and asked the hard questions. See you on the road or in one of our offices!

See Witbe on your own setup

A 20-minute demo, mapped to your stack. Pricing, lead time, and the ROI math included.

Book a demo

FAQ

What did Witbe unveil at NAB Show 2026?

An AI-native infrastructure for testing and monitoring streaming video services. A single platform connecting four layers: Witbox (Connect), REC AI (Operate), Agentic SDK (Automate), and Smartgate AI (Analyse). Demonstrated live at booth W2051, West Hall, April 19–22, 2026.

What awards did Witbe win at NAB 2026?

A TV Tech Best of Show 2026 award for the AI-native infrastructure, with the Agentic SDK at its core. The TV Tech jury came to booth W2051 during the show to see the platform live before granting the recognition.

What productivity gains does Witbe’s Agentic SDK deliver?

Clients report around 80% less QA maintenance on Smart TVs in full black-box conditions, and roughly 100x fewer captures compared to traditional script-based automation. A+E Global Media, an established client who discussed implementation in a webinar earlier this year, broke through a scripting plateau and 4x'd their test coverage in two months with no maintenance after adopting the Agentic SDK.

How much does the Agentic SDK cost to run?

The pricing model is token-based with a hard ceiling. Tokens are pre-provisioned and tracked per test, per device, per application.

How does the Agentic SDK handle UI changes that would normally break scripts?

The Agentic SDK operates on intent rather than fixed UI coordinates. For example, when Canal+ reskinned its homepage for the Superman launch, the entire UI changed overnight, so traditional automation would have broken across every screen. The Agentic SDK adapted at runtime, maintained test intent, and continued running tests on the new interface.

Who uses Witbe's AI-native infrastructure in production?

Witbe is trusted by more than 300 video service providers across 50+ countries. Public customers include Comcast, Cox, Verizon, Sky, Peacock, A+E Global Media, Canal+, BBC, Channel 4, Altice, Orange, YouTube and many more of the biggest players in the industry.

When was the Witbe Agentic SDK launched?

It launched on February 18, 2026, as part of Witbe Suite version 41. It is available on all existing Witbox deployments with no new hardware required, just AI token provisioning and Agentic SDK licensing.

How can I see a Witbe demo after NAB 2026?

Book a session mapped to your own stack at witbe.net/contact-us. Demos cover the same scenarios shown at booth W2051, including the AI dial, the REC AI Assist, and the Smartgate AI root-cause analysis flow, adapted to your devices and applications.