Bespoken LLM Benchmark: Does ChatGPT know more than Google and Amazon?
Learn more
January 17, 2022 in Blog

The Bespoken Orchestrator

TL;DR The latest addition to the Bespoken Platform builds continuous quality monitoring & improvement right into your conversational IVR through an easily adopted runtime with pre-built integrations for a diverse collection of conversational AI tech stacks.

The Bespoken Orchestrator allows contact centers to rapidly and radically improve the accuracy of speech recognition and NLU in their systems. It acts as a drop-in replacement for existing, monolithic NLP platforms, enabling teams to adopt best-of-breed tools and models for their specific use-cases. This means higher containment, increased customer satisfaction, and overall higher ROI for customer service teams. Learn more about the Bespoken Orchestrator here.


  • Continuously monitor, capture, and tag spoke customer interactions
  • Conduct real-time performance measurement and experimentation across multiple ASR, TTS, and NLU technologies
  • Create and manage multiple runtime configurations
  • Makes utterances available to incorporate as test data in Bespoken accuracy testing & monitoring.
  • Route traffic to providers with fine-grained control based on language, customer intent, customer segment, and/or recognition need


  • Tap into the Best-of-Breed across the constantly evolving AI industry
  • React quickly to new vendor releases and future proof your IVR against inevitable change 
  • Achieve optimal performance in every situation, for every caller, in your specific IVR
  • Minimize risks associated with vendor lock-in
  • Objectively compare performance between platforms without major effort and investment

Catch the Hybrid Wave!

  • Conversational Accuracy Matters – it drives real business results.
  • Hybrid Architecture for Conversational IVRs reduces risk and unlocks the best possible customer interactions.
  • Bespoken’s Orchestrator and AI Experts drive major improvements in <3 weeks.

Advance your Conversational IVR to Hybrid-Level Accuracy
An industry sea change is underway as ever-greater limitations are identified in single-vendor approaches to AI-based IVRs.  No single vendor can be best at every possible recognition, and a hybrid, the multi-vendor approach is finally manageable with help from the Bespoken Orchestrator (see details on reverse).

An Architecture for the Conversational IVR Road Ahead

How It Works

  1. Your team expresses an IVR call flow in one of the supported design tools from which Bespoken Orchestrator can import and execute.
  2. Bespoken Orchestrator provides an IVR with improved speech accuracy rates and dynamic allocation of individual conversational turns to the optimal STT, NLU, and TTS engine for that point in time with that caller.
  3. Once the call is ready to be terminated or transferred to an agent, a disposition code is made available to the contact center platform, if applicable.
  4. On-going adjustments and tuning in the Bespoken Dashboard can happen at any time to experiment with the latest developments in STT, NLU, and TTS technology.
  5. The outcome of the recognition is stored as part of the metadata in an utterance repository, which can be managed by Bespoken or your own team.
  6. Available as a separate Bespoken offering, the Bespoken Validator lets you reuse these utterances to do Functional Testing, Load Testing, or Monitoring with virtual customer interactions delivered over the telephony network (PSTN) just as any customer call would be delivered

Leave a Reply

Your email address will not be published. Required fields are marked *