MarTech360 Interview with for Marc Hutchinson, Vice President, Solution Architecture at TTEC Digital

“AI can increasingly analyze processes and recommend improvements, but it cannot replace the perspective that comes from repeatedly building real systems and seeing what works.”

Marc, can you tell us about your professional background and your current role at TTEC Digital?

I’ve spent nearly 30 years working at the intersection of customer service and emerging technology, beginning with engineering and solution architecture at companies such as Lucent Technologies, Bell Labs, Avaya, Aspect, and Genesys, then spent 12 years at Salesforce, where I helped create Service Cloud Voice and its partner telephony architecture, and later moving into product leadership at 1440. Today, as Vice President of Solution Architecture at TTEC Digital, I’m focused on shaping how we architect solutions around Agentforce and AI, while TTEC Digital more broadly helps strategic customers navigate the transformation now taking place across CRM, contact center, and customer experience. What makes this moment especially exciting is that AI is dramatically lowering the cost of building sophisticated software and integrations, putting capabilities and highly personalized service experiences that were once available only to the largest enterprises within reach of almost any organization.

Marc, you’ve spent more than two decades watching customer-service technology evolve from Bell Labs and enterprise telephony through cloud, CRM, omnichannel engagement, and now agentic AI. When you look across that journey, what is one assumption about customer service that has stayed remarkably persistent, even as the technology around it has changed, and do you think AI is finally forcing the industry to rethink it?

One assumption that has persisted throughout my career is that truly differentiated customer service is expensive. Historically, the best experiences required both exceptional people and highly customized software, which meant only the largest or most sophisticated companies could afford them at scale. AI is changing that equation, but the really exciting part is not simply doing yesterday’s customer service more efficiently; it is dramatically lowering the cost and time required to build software, integrations, and new experiences, which lets us create journeys that even six months ago might have been dismissed as impractical. We are already hearing customers say things like, “We may have purchased our last standalone CCaaS platform,” reflecting a broader shift toward increasingly composable customer-service capabilities within platforms such as Salesforce and the major cloud ecosystems. As those technology barriers fall, competitive advantage shifts from who can afford the best technology to who can imagine and orchestrate the best customer journey.

Also Read: MarTech360 Interview with Tim Ringel, Founder and Global CEO at Meet The People

Your career has moved between very different sides of the same problem, from engineering and solution architecture at Bell Labs, Avaya, Aspect, and Genesys to Service Cloud innovation at Salesforce and product leadership at 1440. How has seeing the same customer-service challenge from the technology, product, and customer sides changed the way you judge whether an innovation is genuinely ready for the enterprise?

One thing I’ve learned is that a great demo tells you much less than it used to, because AI makes it remarkably easy to put an impressive veneer over something that may not yet be durable or enterprise-ready. I tend to look at three things: the probable half-life of the technology, the pace at which I expect it to improve, and ultimately the financial return it can create. “Technically possible” only tells me whether something belongs in the lab, and “productized” tells me what someone has packaged for sale; neither tells me whether customers actually need it now or whether the vendor will still be advancing it three years from now. That is one reason I have confidence in platforms like Salesforce: I know the foundation, I know the decades of investment behind it, and I have watched Agentforce move from an exciting early technology into increasingly credible production deployments at the pace I would expect from a major Salesforce platform investment. At TTEC Digital, our role is to bring that same pragmatic lens to customers: not just asking whether an innovation is impressive, but whether it is relevant, durable, and capable of producing measurable business value.

At TTEC Digital, you’re now working at the convergence of Salesforce, voice, AI agents, data, and contact-center technology. From where you sit, what is changing most fundamentally about the architecture of customer service as AI moves from being another capability in the stack to becoming part of the operating model itself?

Customer service is moving from an architecture organized around applications and channels to one organized around AI agents, data, and orchestration. Technologies such as Salesforce Headless 360 and Claudeforce make that shift very tangible: an AI agent can now operate against the same data, identity, workflows, and security model that historically lived behind purpose-built user interfaces, and the ease with which we can equip those agents with new MCP-based skills is forcing us to rethink assumptions that have shaped contact-center UX for more than a decade. We are still discovering whether tomorrow’s primary interface will be a traditional console, Slack, a conversational experience, voice, or some combination of them, but I am convinced it will look very different from today. These re-platforming cycles only come along every 10 or 15 years, and they give innovations that were gradually absorbed into SaaS and CCaaS an opportunity to be reinvented for an AI-first world. Agentforce Contact Center is particularly important because it finally brings voice, still one of the most human and important service channels, natively into that same Salesforce architecture, allowing it to inherit the agentic, workflow, integration, and automation capabilities that digital channels have benefited from for years.

TTEC Digital’s first live Agentforce Contact Center deployment with Compass Working Capital went from kickoff to production in six weeks and is expected to eliminate roughly 6,000 hours of staff work annually. Having been close to that journey, what did the experience reveal about the difference between an AI agent that performs well in a demonstration and one that is trusted to operate inside a live customer-service environment?

What surprised me most was how little time we had to spend on the foundation. Voice has historically been one of the most specialized, expensive and time-consuming parts of a contact-center implementation, particularly when you are integrating telephony with CRM data, workflows, and services. With Agentforce Contact Center, most of that complexity was already handled by a Salesforce platform the team knew and trusted, so we could spend the majority of the six-week project designing better customer experiences and employee-facing AI automation rather than building plumbing. We still tested rigorously, maintained human checkpoints, and worked through details such as fine-tuning the voice agent to verbalize data naturally, but those were experience-design problems rather than infrastructure problems. For the users, trust grew because the technology consistently made their jobs easier and improved the customer experience; for management and IT, it came from safeguards, visibility, and control. That combination of faster time to value and better CX is what made the project so compelling.

What makes the Compass deployment particularly interesting is that the goal was not to automate the human relationship away. AI took on work such as real-time transcription, data capture, task creation, and appointment scheduling while financial coaches remained central to the customer relationship. When you evaluate an agentic AI use case, how do you determine where autonomy genuinely creates value and where human judgment should remain at the center?

At Compass, the question was never simply, “Can AI do this?” These are first-time homebuyers, often navigating the most consequential financial decision of their lives, and success depends on having a coach they trust as an advocate through credit building, savings, mortgage qualification, and everything in between. AI can be extraordinarily capable, and in some cases may even surface better information than a human, but it does not yet share accountability or have “skin in the game” the way a trusted coach does. So we focused autonomy on work that was labor-intensive, verifiable and distracting, such as coordinating appointments across multiple calendars, while using AI to make the coaches more informed, available, and effective. I also think this exposes a weakness in traditional measures like containment, transaction volume, and handle time: the outcomes that really matter are things like homes purchased, credit scores improved, and savings increased. The harder challenge is getting business and financial leaders comfortable tying those outcomes to measurable economic value, but that is where the industry ultimately needs to go.

Gartner’s 2026 research found that only 24% of service and support leaders had demonstrated positive financial returns across their AI use cases, even as AI investment continues to accelerate. From the enterprise implementations you’re seeing, what separates organizations that are turning AI into measurable business value from those that are simply adding more AI capabilities to their technology stack?

The organizations producing real financial returns from AI tend to make two decisions well: where to place the architectural center of gravity for AI, and what measurable business outcome they are prepared to own. Building primarily around established CCaaS or other legacy platforms can be the lower-risk path, but it often produces more incremental returns because those environments are already highly optimized; building directly on frontier AI technologies can create extraordinary innovation, but it also asks the enterprise to assume much more architectural, operational, and vendor risk. I see Salesforce as a compelling balance between those extremes, combining the pace of AI innovation with the data, workflows, governance, and enterprise foundation businesses already depend on. But architecture alone cannot create ROI: IT and the business have to agree on what success means and share accountability for delivering it. In fact, consumption-based AI pricing can be a useful discipline, because if the token cost of a use case feels frightening relative to its expected value, that may be a signal that you are automating the wrong thing. At TTEC Digital, we can take that alignment a step further through managed and hosted offerings tied to business outcomes, so in examples such as BDR-style lead follow-up, our incentive is not to deploy more AI but to make sure the AI actually helps convert more leads.

One of the harder design questions in an AI-enabled contact center is what happens when the machine should no longer be the one making the decision. With 87% of customers in recent Gartner research saying access to a human is essential when companies use GenAI for customer service, what does a genuinely intelligent AI-to-human handoff look like, and what have you learned about preserving context and trust when that transition happens?

A genuinely intelligent handoff is not just a transcript and a transfer button; it is the AI exercising judgment about when the situation has changed, what context matters, and how to prepare the human to take over without making the customer start again. Hard-coded rules still have value as guardrails, but customer issues, business priorities, and employee solutions are constantly evolving, so the handoff logic has to become more dynamic and context-aware. As AI context windows expand, we should take advantage of that by giving the machine broad access to history, data, and prior interactions, then compressing that into a highly relevant briefing a human can absorb in seconds while still engaging emotionally with the customer. That is where platforms like Salesforce have an architectural advantage: with Agentforce Contact Center, the AI and the employee are working from the same customer data, workflows, and connected services, which reduces the integration gaps that so often destroy continuity. The goal is not merely that customers never repeat themselves, but that the combined human-and-AI system can keep up with them, learn from how similar problems have been solved elsewhere, and surface answers from volumes of information no individual employee could process in real time.

You’ve spent much of your career taking emerging technologies from concept toward enterprise adoption, whether that was early web and contact-center technology, cloud platforms, Service Cloud innovations, or today’s AI agents. As enterprises become more comfortable putting AI into production, what do you think solution architects will increasingly need to understand beyond technology itself to make these systems work at scale?

Business and financial judgment absolutely matter, but I think the Salesforce ecosystem has sometimes over-corrected toward the business at the expense of deep technical expertise. This has surfaced most recently in the demand for Forward Deployed Engineering (FDE) experts who are equipped to get hands on in taking innovative solutions from design to production. The solution architects who will be most valuable over the next few years are the ones getting hands-on with technologies like Headless 360, MCP skills, frontier AI models, channel APIs, prompt engineering, and voice synthesis, because customers need people who have actually worked through these problems, not just people who can describe their business impacts. A great architect should be able to move from a highly technical conversation with a specialist developer to a business-case discussion with a CTO, then connect the two into a strategy that lets the customer attempt something they previously thought was too expensive or difficult. AI can increasingly analyze processes and recommend improvements, but it cannot replace the perspective that comes from repeatedly building real systems and seeing what works. At TTEC Digital, that is the kind of expertise we are trying to institutionalize through tools and frameworks like ScoreCX: learn quickly, codify what we learn, and use it to help customers take AI to levels they would not have discovered on their own.

You’ve now seen several generations of technology arrive with the promise of fundamentally changing customer experience. Looking five years ahead, what is one thing you believe the industry is likely to get wrong about the future of AI-powered customer service, and what would you want enterprise leaders to understand before that assumption becomes expensive to unwind?

I think the industry will get it wrong if we define success primarily as automating today’s processes, reducing headcount, and approaching parity with the best competitor in your market. That may improve the economics of the current operating model, but it misses what is extraordinary about this moment: we can now build experiences that would have been considered too expensive, too complex, or simply impossible even six months ago. Companies should absolutely automate the work that does not require a person, but then reinvest some of that capacity in creating delightful new journeys and in developing great employees who can deliver the most valuable human moments when customers need them. I would also be very cautious about locking into any particular AI surface or operating model too early; customers are already moving toward shorter technology commitments because the way we interact with AI is changing too quickly to predict with confidence. Over the next few years, today’s explosion of approaches will undoubtedly consolidate, but right now the advantage belongs to organizations willing to experiment broadly, influence how these technologies evolve, and focus less on recreating what already exists and more on imagining what could never have existed before.

Thanks Marc!