ACP: The Amazon Connect Podcast

Send us Fan Mail

This week we welcome Bentley Paton, a Cloud Interact solutions architect, to discuss Amazon Connect’s Agentic CX Designer following AWS’s NLX acquisition and its move from preview to GA.

This isn't just a review of features - it's talking candidly about the good and the bad of the new Agentic CX Designer with someone who is actually building production solutions with it today.

Bentley explains how it combines agentic and deterministic IVR design in a drag-and-drop UI, removes the need to build Lex bots, and is invoked from a single Amazon Connect contact flow block that routes into CX Designer applications and flows. 

We compare prior Connect AI agents and Lex-based approaches, cover when deterministic logic is preferred (e.g., payments) and how AI can invoke deterministic tools, and review integrations (currently REST/API calls often via API Gateway rather than direct Lambda). 

We also talk about built-in chat-based testing, debugging, and automated test replays, note limited live-call audit visibility, discuss environment deployment via aliases, JSON export limits, business-user readability benefits, iteration needs in production, and current pricing understanding, including separate costs for related services and mention of Live Sync.

Find out more about CloudInteract at cloudinteract.io.

Send us Fan Mail

Host Tom Morgan runs a solo episode of the Amazon Connect Podcast to play five unedited recordings of Amazon Connect Customer AI agent calls and explain what’s happening under the hood.

Four clips come from CloudInteract’s multi-industry demo environment and one from a noisy conference floor. Scenarios include airline fare lookup with email capture, retail banking identity verification, transaction clarification, balance and loan payoff (with calculations done via external tools) and a guarded handoff for address change, a retail order address-change handoff showing the value of ID verification and intent capture, healthcare appointment rescheduling with mishearing detection/correction and a safety transfer for medication side effects, and a conference concierge agent that performs well amid background noise.

Tom emphasizes guardrails, tool use, 100% QA via evaluations and rules, AI agent performance metrics, and the GA release of the Agentic CX Designer visual canvas.

Find out more about CloudInteract at cloudinteract.io.

Send us Fan Mail

Join us as we welcome Kevin Kito (President & CEO) and Allan Gaspe (Solution Architect) from 911 Secure to discuss 911 obligations and solutions for cloud contact centers, especially Amazon Connect. 

They explain common misconceptions about 911 location accuracy, the complexity introduced by cloud and nomadic workers, and key U.S. legislation: Kari’s Law (direct 911 dialing, routing to the local PSAP, and immediate on-site notification) and Section 506 of RAY BAUM’s Act (dispatchable location beyond just an address).

They describe how 911 Secure’s Gatekeeper maintains dynamic location using network signals and prompts remote users to confirm location, how Amazon Connect queries their API only during 911 calls, and how shared callback numbers enable PSAP call-backs to reach the correct agent.

They also cover cell-phone 911 alerting via geofencing with RapidSOS and stress the real-life stakes of getting E911 wrong.

Although this is a more US-focused episode than normal, there is definitely something for everyone here, and we end with a discussion about the moral, not just legal, obligations that administrators worldwide have when considering emergency calling.

Find out more about CloudInteract at cloudinteract.io.

Send us Fan Mail

Join us as we welcome Mark Pritchard, a senior platform and automation consultant at Jabra, to discuss Jabra’s move to Amazon Connect starting in 2021 and how the shift improved flexibility, number management, and cloud operations after COVID. 

Mark explains the challenges of running four global Connect instances with a small admin team, the need to standardize IVRs, scripts, and flow design for easier troubleshooting, and practical conventions like embedding related phone numbers in flow attributes.

He describes replacing a third‑party callback solution with a native Connect approach, then using AI tools (nicknamed “Bob”) with Cursor and GitHub Copilot to build internal portals for callback visibility, skills-based routing changes, and a new migration portal to selectively move resources to new instances during a major business separation, while sharing tools via a new LinkedIn group: Amazon Connect Tools Development.

Find out more about CloudInteract at cloudinteract.io.

Send us Fan Mail

This episode of ACP recaps a busy July of Amazon Connect Customer updates.

The headline is a major quality upgrade to Amazon Connect’s AI voices (announced July 20), adding 100 new voice options across 50+ languages, improved turn-taking and interruption handling, fewer gaps, and more control over speed, volume, emotion, and SSML-like markup (including optional laughter 😂). 

We play before/after demos in US English plus Australian and UK variants, noting increased naturalness but limited regional UK accents. 

Additional releases include new supervisor dashboard metrics for agent queues, automatic example evaluations to support agent coaching, workforce planning/forecasting expanded to emails and tasks, a new “interrupt agent” flow block to force routing to specific agents even when unavailable, and rule-based redaction that blurs specified apps or web pages in agent screen recordings to reduce PII exposure.

The episode also reiterates AWS’s April rebrand splitting Amazon Connect into four products under an agentic AI banner.

Find out more about CloudInteract at cloudinteract.io.

Send us Fan Mail

On this episode we welcome Hannah Bloking, Senior Manager for Applied AI Acceleration Solutions Architecture at AWS, to discuss her career path from fine arts into tech, her work across the Executive Briefing Center, Amazon Connect (now Amazon Connect Customer), and leading Amazon’s internal rollout of what evolved from Amazon Q into QuickSuite.

Hannah shares how AWS and partners are adapting to rapidly evolving agentic AI, why the biggest challenge is choosing and understanding the right use cases rather than the technology itself, and how Amazon Connect’s expanded portfolio (including examples like Connect Talent) is being received by customers.

We also cover the intersection of contact center data and BI/analytics with Quick, the importance of focusing on end-customer experience over cost optimization initially, and advice to upskill, embrace change, and not fear making mistakes.

Find out more about CloudInteract at cloudinteract.io.

Send us Fan Mail

This week we deep dive into call recording in Amazon Connect, covering how recording evolved from expensive, tape-based systems into ubiquitous digital recordings driven by compliance needs like PCI, HIPAA, GDPR, and financial regulations. 

We discuss how most recordings historically went unreviewed, but AI transcription and analytics, such as Amazon Connect Contact Lens, now make it feasible to analyse 100% of calls for transcripts, sentiment, topics, summaries, compliance flags, evaluations, and structured reporting via data lakes.

We also highlight risks around governance, third-party AI, and the importance of PII redaction (including new screen recording redaction capabilities). 

The episode clarifies retention: audio in S3 isn’t deleted by Connect, but contact records become unsearchable after two years unless customers build indexing/search solutions. 

Finally, we discuss migration options and new AWS APIs enabling import of legacy recordings into Connect with conversational analytics, including constraints, quotas, and pacing considerations.

Find out more about CloudInteract at cloudinteract.io.

Send us Fan Mail

We recap the two-day Amazon Connect Customer Community event in London, noting that the agenda and partner conversations were overwhelmingly focused on AI, especially agentic AI, rather than traditional contact center topics.

We discuss workshops on setting up agentic AI and automating evaluations/quality monitoring, emphasizing that while building an agent can be fast, ongoing monitoring, measurement, and business change are critical.

Key themes from sessions and keynotes include AI agents evolving from rigid chatbots into collaborative coworkers that may orchestrate multiple agents, the provocative idea that an “AI boss” could arrive before an AI colleague, and “machine customers” where an AI acts on behalf of users and selects the most efficient experience. 

We highlight strong AWS investment signals, the value of in-person community access, and the urgency for organizations to start now.

Find out more about CloudInteract at cloudinteract.io.

Send us Fan Mail

Amazon Connect Forecasting, Capacity Planning & Scheduling (FCS) Explained — Native WFM, Agent Self-Service, Data Lake & Notifications

Tom Morgan and Alex Baker welcome Pavan Dusanapudi, a senior architect on the Amazon Connect team, to discuss Amazon Connect Forecasting, Capacity Planning, and Scheduling (FCS) and how it compares to workforce management.

 Pavan explains how FCS uses historical contact data and ML/AI to forecast volumes, plan staffing, and optimize schedules, emphasizing native, one-click enablement versus lengthy third-party WFM integrations. They cover agent-focused capabilities in Amazon Connect workspaces such as schedule viewing, overtime offers, time-off requests, and shift exchanges, plus a new schedule/activity notification feature using the Connect rules engine with email, tasks, and EventBridge destinations. 

The episode also highlights Amazon Connect Data Lake access for FCS data and use cases like adherence, payroll reporting, shift rotation equity, and scheduling metrics, and recommends piloting FCS in one business unit with available funding options and a 90-day free trial.

Find out more about CloudInteract at cloudinteract.io.

Send us Fan Mail

Tom and Alex welcome John Ing, freelance Amazon Connect consultant and founder of the independent Amazon Connect user group, to discuss the group’s rapid growth over two years, its community-led agenda, and why smaller, Chatham House Rules-style events enable candid sharing and practical problem-solving.

John compares Connect community dynamics across the UK, Australia, and the US, noting similar challenges but greater scale and regional “clusters” in America that may require multi-city events. He outlines the difficulty of funding events while preserving independence and keeping attendance free.

The conversation also covers AWS’s recent Amazon Connect rebranding, speculation on broader strategy beyond contact centers, and where organizations are on agentic AI adoption—mostly experimentation and POCs, with production limited due to testing, observability, and risk concerns—before talking about the upcoming user group even in London on June 3 - register at tacug.org.

Find out more about CloudInteract at cloudinteract.io.