I had the pleasure of speaking remotely with Nitesh Kumar of Flatworld Solutions. I was excited to have this conversation because I’m intrigued by how organizations are integrating AI and its capabilities into their current processes and technologies, as well as how AI affects people and improves the experiences of agents and customers. That’s what drives Nitesh: finding ways to make AI work within existing environments. Please enjoy our conversation.
NITESH KUMAR is Vice President at Flatworld Solutions, where he leads the Contact Center and Healthcare verticals, along with the company’s AI product line. He brings extensive experience in the outsourcing and offshoring industry, with expertise in sales, customer relationship management, team building, and operational management.
Q. Nitesh, Flatworld Solutions has an amazing story to share. Please introduce yourself, your organization, and what you do.
I’m Nitesh Kumar, Vice President at Flatworld Solutions, where I lead our Contact Center and Healthcare verticals, along with our AI product line. My work sits at the intersection of two things I care about: building technology that genuinely changes how an operation runs, and making sure the people behind that technology are set up to do their best work, not replaced by it. Flatworld Solutions has spent over two decades helping organizations run leaner, smarter customer operations, and today that mission looks like pairing a large global delivery footprint with AI tools we design and build ourselves, rather than license off a shelf.
Q. Please share with us any “stats” about your organization. We would love to know a little about you. How are you structured? Where are you located? Types of businesses you work with in the contact center space? How many centers/agents? Types of communications and how many handled?
Flatworld Solutions runs a global delivery network of 5,000+ agents across 10+ delivery centers, with multiple locations in India, two sites in the Philippines, and additional centers in Bolivia, Colombia, and Vietnam. That footprint gives us real follow-the-sun coverage and genuine language diversity — English, Spanish, Tagalog and Vietnamese-language support operating side by side under one delivery model, not as bolted-on regional add-ons.
On the industry side, we support clients across most of the verticals that depend most heavily on contact center excellence: healthcare, e-commerce and retail, insurance, mortgage and financial services, banking, travel and hospitality, telecom, technology/SaaS, utilities, and logistics. Across voice, chat, email, and social channels, our teams handle well over 200 million customer interactions a year — and that number keeps climbing as clients shift more complex, judgment-heavy work to us rather than just overflow volume.
Q. Your mission is to use technology to drive business efficiency and growth. What technologies do you incorporate and how do they drive efficiency and growth?
The technologies driving our growth fall into three buckets: automation, intelligence, and governance. On automation, we use AI-driven routing, RPA, and self-service to take repetitive volume off live queues before it ever reaches an agent. On intelligence, our proprietary agent-assist tools surface the right answer, the right script, and the right next action in real time, mid-conversation, not after the fact. And on governance, our in-house quality monitoring platform analyzes 100% of interactions instead of the industry-standard 2-5% sample, using sentiment and compliance scoring to catch issues as they happen rather than in a post-mortem review weeks later.
That shift — from sampling to full-coverage monitoring — is, in my view, one of the most underrated AI unlocks in this industry. It turns quality from a backward-looking audit function into a real-time coaching and risk-management tool, and that is where efficiency and growth actually compound.
Q. Please talk with us about how you see integrating AI into customer service centers.
The biggest lesson we’ve learned is that AI succeeds in a contact center when it’s woven into the workflow agents already use, not bolted on as a separate tool they have to context-switch into. We’ve built this out most deeply in regulated, high-stakes environments like healthcare and insurance, where an agent-assist tool doesn’t just need to answer faster, it has to get the compliance language exactly right every time. That same architecture — agent assist, automated quality monitoring, error tracking, and a unified reporting layer — translates directly into e-commerce, mortgage, and other verticals where speed and accuracy both carry real financial consequences.
Done well, AI in the contact center isn’t a chatbot bolted onto a website. It’s a layer that sits underneath every interaction, guiding the agent, scoring the conversation, and flagging risk, so that the channel a customer happens to be in becomes almost irrelevant to the quality of experience they get.
Q. What do you see as the biggest AI challenges, beyond the tech itself?
Honestly, the technology is the easy part now — the models and tools have matured faster than most organizations’ ability to absorb them. The real challenges are structural and human, and I’ll be candid about where I think our industry is getting it wrong.
First, there’s a change-management gap. A lot of companies are bolting AI onto existing processes instead of redesigning the process around what AI actually makes possible, and they end up with incremental efficiency gains instead of real transformation — then wonder why the ROI never shows up on the balance sheet.
Second, there’s a trust deficit on both sides of the interaction. Agents are anxious about being replaced rather than augmented, and that anxiety quietly undermines adoption: if your frontline doesn’t trust the tool, they will work around it. Customers, meanwhile, are increasingly wary of AI-driven service that feels scripted or evasive, especially in sensitive categories like healthcare and financial services, where one bad automated experience can do lasting brand damage.
“The biggest lesson we’ve learned is that AI succeeds in a contact center when it’s woven into the workflow agents already use…”
Nitesh Kumar
Third — and this is the one I think gets underestimated the most — is governance and data quality. Everyone wants AI-driven insight, but very few organizations have the clean, structured, compliant data pipelines to make those insights reliable. Globally, we’re seeing a wave of companies that bought the AI tool before they fixed the data and process foundation underneath it. That sequencing problem is going to show up as a wave of stalled or quietly abandoned AI initiatives over the next couple of years — not because the technology failed, but because the groundwork was never there.
Q. Do you think AI is serving to improve accessibility and efficiency? Do we still have more work to do to make this an operational success?
Without question, yes — AI has made customer service meaningfully more accessible. Real-time translation and transcription are breaking down language barriers, voice-to-text and sentiment tools are helping agents better support customers with hearing or speech differences, and intelligent routing means customers reach the right resource faster instead of being bounced between queues.
But accessibility and efficiency gains aren’t automatically distributed evenly, and that’s where the industry still has real work to do. A lot of AI deployments are optimized for the median customer and the median interaction, and they can quietly make things harder for customers with complex needs, non-standard accents, or several overlapping issues in one call — exactly the customers who need the most help, not the least. Getting this right takes deliberate design and continuous human oversight; it doesn’t happen automatically just because a model gets deployed.
Q. As businesses look to combine outsourcing with AI, what does it take to move beyond simply automating tasks and actually redesign contact center operations for greater efficiency, quality, and growth?
Combining outsourcing with AI only pays off when you treat it as an operating-model redesign, not a cost-reduction project. Too many organizations approach this as “automate what already exists” — take the current workflow, the current org chart, the current KPIs, and bolt AI onto the edges. That gets you a 10-15% efficiency bump at best, and it’s the ceiling most transformation initiatives are quietly hitting right now.
“…I don’t think the winning model is “AI versus agents.” It’s AI handling volume and consistency, and agents handling judgment and trust…”
The real gains come from asking a harder question: if we were designing this operation today, with AI-native tooling, what would the structure, staffing model, and quality framework actually look like? In our own operations, we’ve found that consolidating specialized teams into fewer, larger hubs — rather than spreading thin generalist teams across many small sites — produces measurably better outcomes: faster escalation handling, tighter knowledge-sharing, sharper quality calibration. That’s the same principle AI-era contact centers need to apply to themselves. Don’t automate the old structure — redesign the structure around what AI and skilled people can now do together.
Q. As AI becomes more embedded in contact center operations, how do you see the role of people evolving, and what balance between AI and human expertise do you believe businesses need to achieve the best customer outcomes?
I don’t see this as AI replacing agents — I see it as AI absorbing the parts of the job that never needed a human in the first place: looking up information, filling out forms, summarizing a call, scoring a transcript against a QA checklist. That frees agents to spend their time on what people are still demonstrably better at: reading emotional nuance, de-escalating a frustrated customer, and handling the genuinely novel problem that doesn’t fit a script.
Businesses getting this balance right are investing as much in upskilling their agents as they are in the AI itself, turning agents into higher-judgment problem-solvers supported by AI, rather than treating AI as a slow-motion replacement plan. Long-term, I don’t think the winning model is “AI versus agents.” It’s AI handling volume and consistency, and agents handling judgment and trust — and the moment either side creeps too far into the other’s territory, the customer experience is the first thing to suffer.
This article appears in November 2026.
