Skip to content
Trace Margin

Video and podcast roundups

Which recordings help enterprise buyers distinguish a pilot from a rollout?

Trace Margin research deskSection 07Page 5 of 12Last updated

4. What should an AI deployment have to beat before expansion?

Listen: Voice AI Is the New BPO. A 25-Year Contact Center Vet Explains

Who: Dom Nicastro interviews Farah Ahmed, director of sales and CX transformation at StorageVault, on CMSWire’s CX Decoded. Published: 7 October 2026. Length: not established from the reviewed page. Evidence: publisher recap and embedded podcast. (cmswire.com)

Key takeaways: Ahmed attributes launch readiness to an organised knowledge base, rather than vendor speed alone. The recap describes adding use cases individually and requiring each to improve on existing resolution, callback and survey measures. It also raises the commercial question of per-seat versus per-resolution pricing. (cmswire.com)

Our buyer interpretation: This is our recommended first enterprise episode because it connects operational preparation with explicit expansion gates.

Action after listening: Create a release decision with four possible outcomes: expand, continue testing, narrow scope or stop. Require operations—not only the supplier—to approve the evidence behind that decision.

Caution: StorageVault is a Canadian operator. Its operating lessons may inform US procurement, but do not import its legal or data-handling assumptions. The interview is customer testimony, not an independently audited performance study.

5. How do you test both completed work and caller experience?

Watch: EVA — A Framework for Evaluating Voice Agents by ServiceNow
Audio alternative: ServiceNow Podcasts episode

Who: Bobby Brill with ServiceNow researchers Tara Bogavelli, Katrina Stankiewicz and Gabrielle Gauthier Melançon. Published: 29 April 2026. Length: 30 minutes. Evidence: official video description, episode notes and chapters. (podcasts.apple.com)

Key takeaways: The session separates task accuracy from conversational experience and discusses errors propagating between speech recognition, reasoning and speech output. Start at 07:57 for names and confirmation codes, 10:06 for accuracy versus experience, and 16:57 for a flight-rebooking failure example. (youtube.com)

Our buyer interpretation: This is the most useful technical briefing in the shortlist for a procurement team that needs to understand why a convincing conversation is not an acceptance test.

Action after watching: Require evidence for both the final system state and the caller experience. In a booking trial, assess the appointment record separately from interruptions, repetition and confusing responses.

Scope warning: The accompanying EVA research release uses a synthetic airline dataset and includes expected final database states. It is a testing reference, not certification of a dental, HVAC or financial-services workflow. (huggingface.co)

6. How should testing continue after the demonstration works?

Listen: Building Trustworthy Voice AI: The Evaluation Playbook with Brooke Hopkins

Who: Brooke Hopkins, founder and CEO of Coval, on ODSC’s Ai X Podcast. Published: 16 January 2026. Length: 41 minutes. Evidence: detailed publisher-supplied episode notes. (podcasts.apple.com)

Key takeaways: Hopkins discusses simulation across accents, languages, background noise, interruptions and workflow variations. The notes distinguish pre-production simulation from production monitoring and emphasise evaluating multi-turn conversations rather than isolated answers. They also cover loops, abandonment and compounding errors. (podcasts.apple.com)

Our buyer interpretation: Pair this with EVA when writing an AI voice agent pilot plan. Use EVA to organise measurement and this episode to challenge test coverage and ongoing operating responsibility.

Action after listening: Ask who owns the regression suite, who labels disputed failures, what changes trigger retesting and how production incidents become new tests. Require a named owner after the implementation team leaves.

Commercial lens: Hopkins leads an evaluation supplier. The notes disclose sponsorship by ODSC AI East 2026. Treat the proposed testing approach as input to procurement, not proof that any particular testing product is necessary. (podcasts.apple.com)