AI-Decoded is where we share practical insights from building and operating AI-driven customer experience platforms – beyond the hype and marketing.
In this, our third edition, Neil Mulholland, Principal Engineer, reflects on 2025: a year of rapid change across GenAI, voice automation and cloud infrastructure. The blog looks back at the decisions, challenges and lessons that shaped Syndeo’s product direction and how those learnings are setting us up for 2026. We hope you enjoy the candid view from Neil on what it really takes to turn emerging AI capabilities into reliable, enterprise-grade CX solutions.
2025: A Year in Review
2025 is now behind us. If you’re anything like me, you’re now in that hazy period in January where you’re working your way through the last of the Christmas chocolate and wrestling with New Years resolutions. I’m not a fan of a New Years Resolution, but I am a fan of reflecting back on the year just finished. It’s nice to get out of the day-to-day grind and look back with a longer-term view.
Within Syndeo’s Product & Engineering teams, 2025 was both exciting and challenging. Looking back, the difference from where we started the year and where we finished is striking. Many of the initiatives that felt exploratory at that time have since resolved into a number of strategic goals that now underpin our platform, our roadmap and our approach to AI-driven customer experience.
So, what were those strategic goals?
- Delivering a Voice experience that’s just as powerful and effective as our Digital experience.
- Developing our Flow-Guided approach to Agent development.
- Leveraging that approach to create a suite of ready to go AI Agents.
- Partnering with Google and becoming a GCP first platform.
Let’s take a look at each of these goals in turn and see how we progressed them:
Voice
Our first and most important strategic goal was to establish Syndeo as a powerful CX platform for Voice. This goal was driven directly by the huge demand we’ve seen in the market, and we’ve been lucky to be able to work collaboratively with our biggest clients to add next-generation Voice capability to their solutions. The opportunities in Voice are, if anything, even more convincing than Digital. Despite a widespread “channel shift” philosophy in recent years towards Digital, Voice remains the primary contact method in most Contact Centres.
With the emergence of Gen AI models like Chat GPT, Gemini and Claude, the opportunities are plentiful to reinvent the voice experience away from old school IVR-style solutions. Digital has demonstrated that these technologies can be deployed safely and effectively, delivering huge increases in containment when supported by appropriate security controls and guardrails. The challenge was to bring that same quality of experience to Voice.
Throughout 2025, Voice became our primary area of investment. Over the course of the year, we’ve gone from being able to provide a single Voice solution model to now being able to partner with different telephony and CCaaS providers. This allows us to offer our customers a range of options best suited to their existing tech stack. We’ve also dug deeper into the nuances of the Voice experience and have started to develop tools to reduce latency and other issues that frustrate users and lead the interaction astray.
One of our biggest issues has been keeping up with the pace of change. Partnering with Google (more on that later!) has given us early access to cutting edge GenAI models, built from the ground up for live Voice experiences. Working on pre-GA technologies has been demanding, but it’s positioned us strongly for 2026 where we expect Voice agents to become a central pillar for many customer deployments.
Flow-Guided
The first half of 2025 was, in many ways, a period of evaluation. As Gen AI models advanced rapidly, it became increasingly clear that traditional “deterministic” solutions would soon become defunct as they struggle to compete in terms of the quality of the experience delivered. We looked at competitor offerings and the tooling provided by the GenAI providers. While many of these solutions were impressive, our team’s 20+ years’ experience in the CX industry highlighted some fundamental issues.
In particular, we saw that while new GenAI tools could build great demos, they’d quickly lead to maintenance “hell” in larger solutions. It’s simply not realistic to build a best-in-class, reliable, problem-solving agent that can deal with 20+ distinct types of queries in one gigantic single prompt instruction. Even if teams could overcome the huge cognitive weight of trying to parse and understand all that text and its effect on the end experience, Customer Service teams simply don’t work like that. Multiple people can’t all edit the same single instruction at the same time; they just get in each other’s way.
This insight led us to develop what we now call Flow-Guided agent design. The idea is simple: break large things down into smaller distinct areas. Combine the power and freedom of GenAI models with the structure and guidance of our Syndeo Flows. Using a Flow-Guided approach, businesses can break down everything they want their solution to do into clear, human readable Flows. Within those Flows, the GenAI tools can flex their muscles and guide customers through the solution. Fixed questions and answers have been re-imagined as goals, and the GenAI can use information it knows about the business (RAG) to help guide the user to complete those goals.
Over the course of 2025, we refined this approach considerably and now have something very powerful on our hands.
AI Agents
Once we cracked our Flow-Guided experience, then the next logical step was to build, build, build. The easiest way for us to assess our approach, was to use it. This also aligned with a clear problem we have seen across the industry. It’s great to have tool kits that can build anything, but very often most people don’t want to start from a blank sheet. Clients in the same industries also end up building largely similar functions, for example to identify & verify a customer, or to check on the status of a delivery.
Having solved those kinds of problems before, often again and again for different clients, we took our years of experience and funnelled it into distinct Agents, each able to deal with a particular problem area. Additionally, we separated personality from the equation: each Agent can be a distinct persona or each one employed as a single actor who can solve many different problems.
By the end of 2025, we built, tested and deployed over 50 such agents. Feedback from customers has been invaluable and, in the future, we expect this library to continue to grow as new patterns emerge.
The beginning of 2025 saw us partnering with Google in the delivery of a PoC for one of our major clients. Through this project, our relationship with Google deepened and we gained a lot of exposure to the GCP platform, particularly their new Conversational Agents and Vertex tools. As the year progressed, we saw the strategic advantages of adopting GCP to deliver the Syndeo platform. Their cutting-edge Gemini models, partnered with their broader infrastructure toolset made them a natural fit and we will seek to work even more closely with Google over time. Our relationship with Google led us to launch our solution on the Google Marketplace, one of the few Conversational AI companies available for purchase via Google.
It’s no easy thing to change the underlying cloud on which our product resides. Along the way, we encountered many interesting peculiarities. In the technical world, what counts as a single “CPU” is vastly different between AWS and GCP. We also encountered unexpected challenges managing the operational complexity of our microservices architecture.
However, we’ve brought with us some of our learnings from previous customer projects. We’re no strangers to multi-stage, global rollouts and one of the most positive outcomes from this project has been the close collaboration of our Customer Success and R&D teams to deliver the migration. Customer Success brought all their customer specific knowledge with them, helping us to understand how customers will be impacted by potential migration plans and to decide upon the best way forward.
Looking Ahead to 2026
Overall, I’m pleased with the substantial progress we’ve made in 2025. While not all of it was immediately visible to our customers, it has positioned us well for the year ahead. Our strategic goals remain largely unchanged as we move into 2026. We will continue to invest heavily in Voice, with particular focus on performance, latency and background noise. This will unlock automation for even more callers. Flow-Guided will continue to evolve as we support a broader range of GenAI models and introduce client-specific guardrails. And for the first time, all Syndeo customers will be operating on GCP.
We are also exploring partnerships with other CCaaS providers, and in 2026 we want to reflect on our own tooling. Syndeo is a powerful, feature-rich platform, but with that comes difficulties in assisting users with how to make the best of it. We want to adjust our user experience, remove defunct tools and better onboard users so that anyone can build a Virtual Assistant, not just the CX experts.
But as always, our goals remain flexible, and we’re ready to adapt as needed. One of the reasons I avoid setting a New Years Resolution is because anything can happen, and the same is of course true in the CX landscape.
Through AI-Decoded, we’ll continue to share what we’re learning as we navigate change: openly, pragmatically and grounded in real delivery rather than theory alone.
In our first edition of AI-Decoded, we discovered how Syndeo Co-Founder and CTO Alan Beck is transforming customer service through AI innovation, from his early teenage coding days in Northern Ireland to leading one of Europe’s fastest-growing tech companies. If you missed it, you can read it here: How CTO Alan Beck Is Transforming Contact Center with AI Agents
In the second edition, we broke down how AI Agents can help retail teams when volumes spike. Here’s the read: How to use AI agents to assist your Christmas rush.
If you would like something more practical, don’t miss our latest FREE e-book.
Download “Designing Your AI Agents Workforce for Customer Service” for practical frameworks, implementation strategies and real-world examples you can start applying today.
