Contents Link to heading

A Partnership in Motion Link to heading

A few years ago I had the opportunity to talk to a client about transforming their customer experience and leveraging technology to make the overall experience more efficient for all parties - customers, agents, and other parts of the business.

Workshop Link to heading

It started by diving into the weeds to fully understand what was happening. We held an on-site workshop with key stakeholders from all aspects of interaction. Before touching any technology, we spent time with various roles within the business, understanding their actual pain points from multiple perspectives. What was important to the business? And how could agents better help customers faster?

There were a few core components and through the process, leadership prioritized being innovative. This allowed us to consider a new tool for the types of interactions the callers were having: conversational AI. The product we looked at was maturing, but the key was that they could modernize their interactions with their existing on-premise platform and future-proof their journey, as many of the major Contact Center as a Service (CCaaS) players were integrating with the platform for AI capabilities. This was one of our first tangible wins and helped build the relationship across many levels. The most important part was that it provided immediate impact, raw turn-by-turn analytics, and a good way to build momentum with the client while other adjacent and bigger engagements were in the works.

Gathering Data Link to heading

We also had another engagement to gather more data from a historical perspective. With the implementation of natural language understanding (NLU), we wanted to understand from the call data what the actual calls were about and what agents were talking to customers about. We pulled over 20,000 conversations, converted the audio, and generated transcripts so we could dig into the data. This turned up trends and call drivers that had seasoned veterans wondering, “how is this type of call, going through that flow, in those volumes?”

Voice Security Link to heading

Somewhere in all of this, we also worked with a voice biometrics solution to help tackle fraud, both in the legacy on-prem interactive voice response (IVR) that was being phased out, and the new intelligent virtual agent (IVA). Fraud detection has a lot of moving parts, including carrier considerations, CRM tie-ins, and IVR components. It also had to work across very different environments concurrently. On top of that, we integrated with multiple leading CRM platforms and multiple instances of each.

Considering a Future Link to heading

Finally, we took a hard look at the core platform and what moves we could make there. We considered a few different solutions. We took many factors into consideration such as the platform vision, existing components, roadmaps, and ease of migration. From the existing platform, the main components that were important were the existing conversational AI we had built and were improving, and the quality management (QM) and workforce management (WFM) platform.

With that decision came the design and implementation, done in a rolling fashion. Consolidating multiple platforms and legacy integrations took a lot of planning and tight execution to come together. The environments didn’t move to a modern contact center platform overnight, and they came with all the usual joys of vendor coordination, data migration, and making sure nothing broke for a company handling millions of customer accounts across a series of cutovers.

Moving with the Winds Link to heading

But wait! In the middle of the project, the GenAI revolution started taking place, and platforms and bolt-on startups were coming out of garages everywhere with the newest AI solution. Every step of the way, we worked with key stakeholders to evaluate the solutions out there and where they stood relative to our existing conversational AI investment.

In the end, two solutions really stood out: the existing conversational AI solution was slowly growing and incorporating generative AI and agentic capabilities but they were not fully there yet, then we met a new player that offered a future forward, agentic solution and was making waves and receiving some major investments.

After careful consideration, we pivoted and leveraged a mix of the two solutions until we could fully consolidate onto the agentic AI solution. This wasn’t a case of the first solution failing but rather enabling the client’s evolving ambitions as agentic AI matured and needing to adjust the architecture to match. This approach meant we didn’t throw the baby out with the bathwater either, at points we were running calls through both the on-prem and cloud platforms, integrated with agentic AI, and still leveraging some of the conversational AI components for tactical flows.

Native Data Link to heading

There were a few other things we did along this journey specifically around QM and WFM. The key ones that stand out were implementing bots and migrating and consolidating multiple on-premise systems into a single cloud instance. Building on the initial data insights project, the client understood the importance of being able to export transcripts and mine the data. That project allowed us to look for simpler, more native solutions that complemented what they had and where they were looking to go. Since their existing QM solution was already doing the call recording, the client had access to tools that would allow the generation of the full transcript for data mining. This allowed them to leverage other tools across the existing platform as well as provide the analytics groups with access to the raw data that could be pulled from other systems.

Nearing the End Link to heading

As we near the finish line on the original game plan, I’m humbled by the effort that everyone put in along this journey. None of it would have been possible without the incredible team on this engagement. If I had to sum up the team’s ethos, it’s a philosophy a mentor of mine embraced, “there’s no such thing as stuck, just not solved yet!”

And now that we’re creeping up on the “finish line,” it’s really just another way to ask “what’s next?”

A Few Lessons Learned Link to heading

  • Listening matters more than the tech stack. The workshop shaped every decision that came after it. Skipping that step to jump straight to tooling would have cost us more time later, not less.
  • Data reveals problems you didn’t know you had. The data analysis surfaced call topics that didn’t make sense and shouldn’t have been happening — a good reminder that the data itself is often more valuable than the AI model built on top of it. This directly validated the value of the transcription bots we ended up deploying, which could be easily embedded within their existing solution set.
  • Migrations and new capabilities don’t have to be sequential. Running the conversational AI quick win alongside cloud contact center migration kept momentum visible to the client while the slower architectural work happened in the background.
  • Fraud and security can’t be an afterthought. Building voice security into the legacy platform and the new cloud stack meant designing for two environments at once and building a solution versatile enough to handle both.
  • Being willing to pivot mid-project is a feature, not a failure. Moving from a conversational AI approach to layering in a separate agentic AI allowed us to match the solution to where the technology and the client’s ambition had moved.
  • Be curious. The most memorable moments came from someone asking “what if” or “why not.” Each question opened up new possibilities and answers that fueled curiosity to keep exploring and accepting new ideas.