# BotBot lets brands experiment on voice-AI behavior and tie changes to business outcomes

Canonical: https://mudpie.ai/companies/botbot/
Breadcrumb: [Home](https://mudpie.ai/) / [Companies](https://mudpie.ai/companies/) / [BotBot lets brands experiment on voice-AI behavior and tie changes to business outcomes](https://mudpie.ai/companies/botbot/)
Author: Ali Abouelatta (https://mudpie.ai/authors/ali-abouelatta/)
Published: 2026-09-19
Updated: 2026-09-19
Research type: Company profile
Method: Company and accelerator sources checked 2026-09-19. Product claims are attributed to their sources; this is research, not a hands-on product trial.

BotBot is a behavior-experimentation layer for voice AI. It fits a brand that already has a voice agent but cannot tell whether its greeting, pacing, upsell timing or recovery language is helping the business.

## What it does

BotBot says teams can change voice behavior in plain English, run variants against live conversations and measure outcomes such as order value, booking completion, handle time or containment. The company can either plug into an existing voice stack or supply the agent, while engineers keep the runtime in code and non-engineering teams work in a UI. [BotBot homepage](https://botbot.com/)

That is a more useful wedge than another transcript dashboard. The product is designed to connect a behavior change to a business metric and build a record of what works for one brand’s customers. It names Pipecat, Deepgram, ElevenLabs and OpenAI as integration paths, which matters to a team that does not want to replace its current model, voice provider or runtime.

The public homepage shows experiment cards with percentages, ticket values and store counts. They read like product examples rather than identified customer case studies, so I am not treating them as verified outcomes. The real buying question is whether BotBot can expose the assignment, traffic split, guardrails and denominator behind a live test without disrupting a revenue-bearing call flow.

The founding context supports the problem. The [Speedrun profile](https://speedrun.a16z.com/companies/botbot) describes Steve Mayernick’s prior AI, UX and growth work at RJMetrics, Guru and Pryon, and Michael Mayernick’s engineering leadership at McDonald’s AI labs and IBM Watson Orders.

## What to check

Before rollout, a brand should define which behavior is allowed to vary, which calls are excluded, how human escalation works and what counts as a successful conversation. Voice experiments can change several outcomes at once: a warmer greeting may affect duration, satisfaction and conversion, while an upsell change can alter both revenue and abandonment. BotBot’s promise only matters if those tradeoffs remain visible.

The site does not publish standard pricing. Procurement should also ask about transcript retention, access controls, model/provider responsibility and rollback speed. “No engineering ticket” is useful only if the organization has a safe approval path for changes to customer-facing behavior.

## My editorial take

I would shortlist BotBot for a company with enough voice volume to learn from controlled changes and a clear owner for call quality. I would not buy it to decorate an unmeasured voice bot. Start with one narrow behavior and one primary metric, then review the full conversation impact before expanding to promos, personas or regional variants.

## Quick facts

| Field | Sourced detail |
| --- | --- |
| Buyer | Brands running voice AI for ordering, booking or service |
| Product | Behavior controls, live experiments and outcome measurement |
| Integrations | Pipecat, Deepgram, ElevenLabs and OpenAI are named |
| Pricing | Contact-led; no standard public price found |
| Main fit question | Do you have enough call volume and instrumentation to learn safely? |

## Sources checked

BotBot’s Speedrun company profile, homepage and careers page were checked on 2026-09-19.


## Author disclosure

I cofound Lazyweb and publish Mudpie. This is an owner-written publication, not an independent testing organization. Research notes distinguish observations, sourced reporting and editorial judgment.