Speko GTM brain

category

market profile

• text: Voice agents combine speech recognition, language models and speech synthesis, and the best model can differ by language and objective.
• text: Speko offers a hosted routing API, a customer-side gateway and a separate hosted agent and telephony platform.
• text: The market includes full voice agent platforms as well as individual model providers and open runtimes.

personas

• title: Voice AI engineer
  pains:

  • Benchmarking and integrating separate STT, LLM and TTS vendors is ongoing work
  • A single vendor may perform unevenly across languages
• title: Engineering leader at a voice agent startup
  pains:

  • Voice quality, latency and cost change as providers release models
  • Provider outages or switches create operational and engineering risk
• title: Customer operations or product leader
  pains:

  • Phone agents need testing against difficult callers and monitoring after launch
  • Running multilingual inbound services requires reliable telephony and escalation

icp

• text: Voice AI software teams building production phone agents or embedding speech in applications, especially those serving multiple languages.
• text: Technical teams using LiveKit or Pipecat that want routing while retaining control of the conversation runtime.
• text: Higher-volume enterprises with committed usage and custom operating requirements can evaluate the enterprise tier.

message pillars

• text: Unify multiple speech models behind one integration.
• text: Select models based on language-specific performance.
• text: Support the full lifecycle of voice agents.

objections

• objection: Why not use one speech vendor directly?
  response: Compare the published language benchmarks and test whether routing improves your own calls enough to justify another layer.
• objection: We need to control credentials and runtime.
  response: The customer-side Gateway is described as keeping provider credentials in the customer process and working alongside LiveKit or Pipecat.
• objection: What will calls actually cost?
  response: Model routing has a provider-rate markup; the bundled infrastructure rate and telephony charges are separate pricing paths and should be modeled per use case.

competitors

• name: Vapi
  why: Voice agent platform competing for developers building phone agents.
• name: Retell AI
  why: Voice agent platform with overlapping build and telephony use cases.
• name: Bland AI
  why: Voice calling agent provider listed in Speko's own comparison navigation.

influencers

• name: Bek Abdikamalov
  url: https://www.ycombinator.com/companies/speko/jobs/UqVS4xf-founding-head-of-growth
  why: Founder and CEO with previous production voice AI experience.

keywords

• voice AI model router
• speech to text model benchmark
• multilingual voice agent API
• voice AI latency comparison
• voice agent platform pricing
• TTS routing API
• Vapi alternative

voice

• technical and measurement-led
• direct
• developer-oriented

signals

• text: A voice agent team is expanding into languages where its incumbent speech model underperforms.
• text: An engineering team is evaluating LiveKit or Pipecat and comparing STT and TTS providers.
• text: Call volume is growing enough that latency, model pricing and failover have become material.

stack

• LiveKit
• Pipecat
• SIP

hiring

• Founding Head of Growth

pricing

motion

channels

• developer documentation
• SEO benchmarks
• LinkedIn
• X
• Discord
• sales demos

confidence

Founders

• name: Beknazar Abdikamalov
  title: Founder/CEO
  role: CEO
  bio: Founder and CEO of Speko (YC S26). Previously co-founder and CTO of Hupo (voice AI; $14M raised; backed by DST Global and Meta), where he spent four years shipping production voice AI for fortune 500s in 10+ languages. Earlier a software engineer at Amazon; MS in CS from UNIST, Korea.