UponAI Solutions & Features

26 calls inside UponAI: highlights from AI Contio's hands-on review

An independent Voice AI reviewer built an agent on UponAI from scratch and put it through 26 test calls. Here is what stood out, and the tools that make it possible.

By UponAI TeamNovember 9, 20264.0 min read
26 calls inside UponAI: highlights from AI Contio's hands-on review

This article is part of the local UponAI blog library and focuses on practical AI voice and communications workflow lessons that matter in live business environments.

Independent review · AI Contio, Behind the Call

26

Test calls in three rounds

47 min

First login to published agent

700-900 ms

Response per turn on Claude Haiku

Why this review matters

Michael Oeth runs Behind the Call at AI Contio, a series where he builds real agents on Voice AI platforms and tests them hard.

In late October he built a front-desk agent for a fictional medical practice in our reseller portal, timed the setup, and ran three rounds of test calls. He has no commercial relationship with UponAI, and he built everything on his own, without our onboarding, to see what a new reseller can do alone.

His verdict on the platform underneath was clear.

“This isn’t a thin wrapper around an LLM with a phone number attached.”

Michael Oeth, AI Contio

What stood out in testing

  • The best call logs in the series. Per-turn latency split into speech recognition, model and speech synthesis, token counts, and the exact arguments and result of every tool call. He could prove every finding from the logs alone.
  • Fast, natural conversation. On Claude Haiku, responses held around 700-900 ms per turn. Interruptions worked, mid-call corrections stuck, and in 26 calls the agent always ended the call cleanly on its own.
  • A 47-minute first build. From first login to a published agent with a working custom function, well inside our 60-minute promise.
  • The custom function editor. He named it the best screen in the product: plain-form parameters instead of JSON, and a Test Request panel that shows exactly what was sent and received.
  • Real version control. Named versions, drafts kept separate from what’s live, and phone numbers pinned to a specific version.
  • Telecom-grade instincts. Function replay against a different endpoint without placing another call, and even a packet capture tab, built by a team that knows SIP and routing.

Already live on UponAI

The review also put a spotlight on tools resellers can use today to build agents that are fast, safe and dependable.

1. Agent IQ

Analyzes an agent’s prompt and functions and suggests improvements, a big help for resellers who aren’t prompt engineers.

2. Voice Simulation

Runs scripted test callers against an agent and scores them against weighted criteria with mocked tools, before a real customer ever calls.

3. Current Time Awareness

Gives the agent today’s date and time, so it books inside real office hours.

4. Live human takeover

Step into a live call from the browser whenever a caller needs a person.

5. Live context injection

Feed new information into a call while it’s running.

6. Weighted A/B routing

Split traffic between agents on the same number to test what works best.

Built for resellers

The review’s conclusion matches why we built UponAI. For telecom providers and MSPs who know SIP, routing and carrier relationships, it called our observability, versioning, simulation and live takeover tools a strong foundation for selling AI voice under their own brand.

Our partners’ customers, the dental office, the HVAC shop, the clinic, just want a phone that’s answered and work that gets done. UponAI gives resellers everything they need to deliver that, with their name on it.

Thanks, Michael

Thank you to Michael Oeth for the time and the rigor. Read the full review on AI Contio.

Building on UponAI? Our free onboarding sessions walk you through Agent IQ, Voice Simulation, time awareness and the prompt habits that get the best out of every agent.

Book your free onboarding →

What This Means

UponAI content is built around production use, not generic AI positioning. The goal is to help teams understand how routing, call handling, automation, and human handoff behave once the system is part of daily operations.

AI Voice Demo

Want To See These Workflows In Practice?

Book a demo to see how UponAI turns inbound demand, AI voice handling, and human escalation into one operating workflow.

We use cookies to improve your experience, analyze performance, and support our marketing. Declining won't affect your use of the site. Privacy policy.