Good morning, and thank you for the time.
Daniel and I are here this morning to give you a readiness update on Project Amber, and to ask the Board for approval to launch.
A quick note on naming. Project Amber is the SIG’s broader AI effort. The product you’ll see today, the member-facing assistant, is Chapter One of that effort.
——- SLIDE 2 OF 14 AGENDA ABOUT 20 SECONDS
STEVE:
Here’s where we’re going. I’ll give you the headline, then walk through what Amber is, the evidence that it’s ready, the business value, and what comes next. We’ll close with the decision we’re asking the Board to make.
——- SLIDE 3 OF 14 EXECUTIVE SUMMARY ABOUT 60 SECONDS
STEVE:
The headline is this. Amber has met every readiness target we set, and we believe it is ready to launch to all members in Milan.
Three proof points.
One. The system is production-ready. Clean Azure environments. Independent security testing has passed. Infrastructure has been load-tested for scale.
Two. We met our composite launch readiness gate of ninety-five percent. Our expert member review group has voted unanimously that Amber is ready.
Three. The business case is real. We expect Amber to offset roughly one hundred thousand dollars a year in contractor support costs, while accelerating onboarding for both staff and members.
With that context, the ask, which we’ll come back to at the end, is Board approval to launch Project Amber to all members at the Milan F2F on June third.
——- SLIDE 4 OF 14 WHAT AMBER IS ABOUT 50 SECONDS
STEVE:
Before we get to the readiness story, a quick reminder of what Amber actually is.
Amber is a single, always-available AI expert for the Bluetooth ecosystem. Every member company can use it.
And this is the important part. Amber retrieves, and Amber cites. Amber does not invent answers. It is grounded in our adopted specifications, our test documents, our qualified product data, and our SIG processes. Every answer shows where it came from.
It also connects live to the Qualification Workspace, so members get answers grounded in their own qualification data. That’s something no general AI tool can do.
It runs on Azure, built in close partnership with Microsoft.
——- SLIDE 5 OF 14 PROCESS GUIDANCE AND QUALIFICATION MAPPING ABOUT 45 SECONDS
=====================================
DANIEL (suggested):
This is where Amber earns its keep.
It walks members through the SIG qualification workflow, step by step. It integrates directly with the Qualification Workspace APIs. So when a member asks, what test cases are required for my product, Amber answers against their live data. And it links ICS items to their required test cases through the TCMT, accurately.
[BEAT]
This is where Amber delivers value no general AI tool can match. It understands SIG-specific processes, and it understands member-specific data.
[CLICK]
=====================================
SLIDE 6 OF 14 INFRASTRUCTURE AND SECURITY ABOUT 75 SECONDS
=====================================
STEVE:
Two stories on this slide. Stability, and independent security.
[BEAT]
On stability. We migrated to Microsoft Azure, with clean development and production environments built from scratch using Infrastructure as Code, on the Microsoft Agent Framework. We’ve completed rigorous automated and manual quality assurance. Our API connections are stable and verified.
We load-tested for the Milan launch scenario. Just over six thousand requests, at ten, fifty, and two hundred concurrent users. Ninety-eight point five percent success. Plus live load testing by our QA team and SIG staff.
[PAUSE]
Now the part I want to underline.
On independent security testing. Zero successful attacks.
We ran four red team rounds through Microsoft Azure AI Foundry. Five hundred and fifty-two standardized attacks. A zero percent attack success rate, across every category and every complexity level.
Because this ran on Microsoft’s standardized harness, it’s fully independent and auditable inside Azure Foundry. This is not a self-graded test.
[CLICK]
=====================================
SLIDE 7 OF 14 CONTENT AND ACCURACY ABOUT 90 SECONDS
=====================================
DANIEL (suggested):
On content. We’ve been busy in the run-up to Milan. We added and indexed the Drafting Guidelines. We connected ICS and Test Case Mapping to the Qualification Workspace APIs. We expanded our curated content and few-shot examples. We added support for member document uploads. Diagrams and figures in the specs are now searchable and referenceable. And we’ve added curated answers to anchor responses on the questions members ask most often. Core six point three and HID SCI are being added.
[BEAT]
Now on accuracy. I want to be precise here, because there are several numbers and they each mean different things.
[PAUSE]
Our primary go, no-go metric is the Composite Launch Readiness Score. It combines three things. Sixty percent SME acceptance, thirty percent answer quality, and ten percent head-to-head win rate. The gate is ninety-five percent.
We are at ninety-five point three. We have cleared the gate.
[BEAT]
The strongest confidence signal sits underneath that. One hundred percent SME acceptance, on our forty-four-question launch set. Forty-four accepted, zero partial, zero rejected.
We also win one hundred percent of head-to-head comparisons against ChatGPT and Microsoft Copilot. I want to be honest about that number. We treat it as a benchmark, not the gate. Those tools cannot access SIG content, so it is structurally an unfair fight.
On the broader two hundred-question coverage suite, we are at ninety-three point five percent.
[CLICK]
=====================================
SLIDE 8 OF 14 EXPERT MEMBER VALIDATION ABOUT 45 SECONDS
=====================================
DANIEL (suggested):
And the people closest to this content agree.
We granted pre-production access to a growing group of expert member reviewers. Roughly twenty in March. Fifty-two today. One hundred and six unique users in total.
We have a structured feedback loop. In-app thumbs up and down. A triage process that routes accuracy issues, feature requests, and general questions to the right owners.
[BEAT]
And the expert review group has unanimously agreed that Amber is ready for member-wide launch.
[CLICK]
=====================================
SLIDE 9 OF 14 BUSINESS VALUE ABOUT 50 SECONDS
=====================================
STEVE:
So what does Amber do for the SIG, in dollars and in time.
We expect it to offset roughly one hundred thousand dollars a year in contractor support costs.
That comes from deflecting an estimated eighty percent of the routine interoperability queries that Cloud to Ground handles for us today.
It frees SIG staff to focus on escalations and errata routing, instead of repetitive lookups.
And it drastically reduces ramp-up time for onboarding new staff, and for new engineers at member companies.
[CLICK]
=====================================
SLIDE 10 OF 14 THE MILAN LAUNCH ABOUT 50 SECONDS
=====================================
STEVE:
On to the launch itself.
We plan to open Amber to all member companies at the Milan F2F on June third. The launch moment will be a live presentation and a video.
Rachel is leading the go-to-market. That includes a phased soft launch, member messaging, and post-launch monitoring.
[BEAT]
Worth saying clearly. The Milan launch is not the finish line. It is the start of a continuous improvement loop. Member feedback flows back to Daniel and the content team. Daniel remains the authority on every answer Amber gives.
[CLICK]
=====================================
SLIDE 11 OF 14 ROADMAP ABOUT 75 SECONDS
=====================================
STEVE:
Looking forward.
Two things are coming to Chapter One soon after Milan.
First, a feature we call Projects. It lets members scope Amber to a chosen set of documents for focused, deep cross-referencing. We will preview it in Milan, and we are targeting completion by mid-June.
Second, a model upgrade to GPT five point four. Similar pricing, with a reported thirty-three percent reduction in factual errors. We are evaluating it now.
[BEAT]
And the longer roadmap.
Chapter Two is AI-assisted test case and specification authoring. The goal is to lift document quality and reduce the burden on working groups. This depends on Project Blue’s AsciiDoc migration, which I’ll touch on at the end if there’s time.
Chapter Three is a qualification workflow assistant, embedded directly inside the Qualification Workspace.
[BEAT]
And one more piece of the roadmap I want to flag. We’re scoping the work to expose Amber as a connector, using the MCP standard. The simple version. It means staff and members will be able to query Amber’s knowledge from inside the tools they already use, starting with Claude Cowork. Alex has scoped it. We are treating it as a core Chapter Two building block.
[CLICK]
=====================================
SLIDE 12 OF 14 RISKS AND CURRENT LIMITATIONS ABOUT 60 SECONDS
=====================================
STEVE:
No project this size launches without limits, so let me name ours plainly.
[BEAT]
Risk. Pinpointing the single most relevant section for very broad Core Spec queries is hard. We are mitigating with ongoing routing and retrieval tuning, and with curated answers that anchor the questions members ask most.
Limitation. Amber works from adopted SIG specifications. It does not analyze a member’s own proprietary external documents. We are setting clear scope expectations with members. The Projects feature will let them focus Amber on their chosen specs.
Dependency. Member-facing launch is contingent on Terms of Use finalization. That work is on track with General Counsel. We will be asking the Board to approve the Terms of Use in Milan as well.
[CLICK]
=====================================
SLIDE 13 OF 14 DECISION REQUESTED ABOUT 45 SECONDS
=====================================
STEVE:
Which brings us to the ask.
We are requesting Board approval to officially launch Project Amber to all members at the Milan F2F on June third.
Daniel and I are happy to take any questions before the vote.