Thesis & priorities
Frontier’s core thesis is that live, in-the-moment coaching fundamentally changes sales outcomes in a way post-call review cannot. Delivering truly effective coaching requires solving significant real-time challenges, which form the strategic foundation of our platform and product development.
The Hard Problem: Real-Time Coaching
Section titled “The Hard Problem: Real-Time Coaching”The critical value Frontier provides is real-time sales-call coaching, delivered through a heads-up display (HUD) for sales representatives. This live intervention, where the system surfaces a coaching cue (a “signal”) at the precise moment it’s needed, is far more impactful than reviewing calls after they’ve ended. The ability to react and guide a sales rep during an active conversation unlocks a new level of performance. Achieving this real-time delivery requires overcoming significant engineering hurdles, particularly ensuring extremely low latency in capturing audio, transcribing speech, interpreting intent, and generating relevant guidance. Our current architecture is designed to optimize for this speed, with a design target of sub-one-second delivery for fast answers.
Core Tenets: Moat and Commodity
Section titled “Core Tenets: Moat and Commodity”We distinguish between commodity services and the unique value Frontier builds. Core technologies like Speech-to-Text (ASR) provided by services such as Deepgram or Cloudflare Workers AI, and large language models (LLMs) from providers like Anthropic, Google, OpenAI, and TogetherAI, are viewed as powerful, widely available tools. Our strategic moat, therefore, is not in owning these foundational AI components, but in the sophisticated, real-time orchestration of these services. This includes our ability to accurately segment speech (utterance handling), perform rapid inference on each segment, and deliver precise coaching signals back to the rep’s HUD within milliseconds. This orchestration capability, coupled with a virtuous data flywheel that uses aggregated coaching interactions to continually refine our models and knowledge base, forms our true defensibility.
Next Horizon: Accuracy and Completeness
Section titled “Next Horizon: Accuracy and Completeness”Having established a functional and fast real-time coaching system, our explicit next investment area is in enhancing the accuracy and completeness of the coaching signals. This involves a multi-pronged approach:
-
Evaluation Pipelines: Building robust evaluation processes to objectively measure the quality of our coaching signals against ground truth.
-
Data Pipelines: Improving the data infrastructure to ingest, process, and leverage more diverse information sources for a richer understanding of call context and knowledge.
-
Continually Improving Knowledge Base: Evolving our knowledge system, which is in active evaluation across multiple backends. While we currently use Cloudflare AI Search and Supermemory as an interim bridge, we are actively experimenting with advanced approaches like Graph RAG (Retrieval-Augmented Generation) services, including Amazon Bedrock Knowledge Bases and solutions from Anthropic, to provide more structured and contextually rich answers. This will enhance the depth and precision of the guidance provided.
Our Approach: UX, Speed, then Accuracy
Section titled “Our Approach: UX, Speed, then Accuracy”Frontier’s development journey has followed a deliberate sequence:
- UX First: Initially, our focus was on delivering an intuitive and seamless user experience for the sales rep.
- Speed Next: We then prioritized optimizing the system for real-time performance and low latency, ensuring coaching was delivered quickly enough to be actionable.
- Accuracy & Completeness Now: With a solid foundation in UX and speed, our current strategic focus is on elevating the accuracy and completeness of the coaching itself. This includes refining our knowledge base and evaluation methods, which we believe will unlock the next level of value for sales teams.