Human in the Loop AI Systems: Insights from Senthil Kumaran

Artificial intelligence is rapidly reshaping the foundations of modern marketing, but its real impact lies not in replacing human ingenuity, but in augmenting it. As organizations race to embed AI into their workflows, the challenge is no longer just about adopting new technologies – it is about building systems that can scale intelligently while preserving creativity, trust, and control. The future of digital ecosystems will be defined by how effectively businesses integrate data, automation, and human judgment into a cohesive, adaptive model.

In this conversation, Senthil Kumaran, Chief Technology Officer at CreatorIQ, underscores a clear message: successful AI adoption is rooted in strong data foundations, thoughtful system design, and a firm commitment to keeping humans in the loop.

By addressing real-world challenges, from architectural trade-offs to governance and trust, he offers grounded, actionable insights into building AI systems that are not only powerful but also responsible and sustainable.

AI and an innovation mindset

With AI at the center of your organization’s roadmap, how do you foster a culture of experimentation while maintaining enterprise-grade reliability?

Kumaran: My approach has always been to build safety mechanisms and guardrails in the tooling, infrastructure, and through processes to allow for safe and rapid experimentation. Effectively balancing quality and reliability with the required pace for iterative building is a key goal for us at CreatorIQ.

As I noted above, this area of infrastructure and tooling is one of the areas that we are rethinking. And a desired effect of that is the increase-in-pace unlock. The ML pipelines must support deploying and tuning many models while quickly extracting signals from experimentation and enabling broad feature engineering across capability segments.

The ability to test across a controlled user segment, both experiential and non-experiential elements, is of high importance.  

At a global scale, data is fragmented across platforms like Instagram, TikTok, and YouTube. In your AI tool, what architectural patterns have proven effective in normalizing and unifying this data?

Kumaran: The core challenge is that every social platform structures data differently. And this is one of the places where complexity comes in.

We have built sophisticated at-scale pipelines and a centralized intelligence layer, the Creator Graph™, which processes 123 million posts daily. From there, we standardized how we measure performance, suitability, and engagement.

So that brands can compare results across platforms in a consistent way.

How do you approach latency, cost, and scalability trade-offs when deploying AI features across enterprise customers?

Kumaran: Every large feature or capability we develop has to go through evaluating these tradeoffs. When large-scale features are built, the cost is a key factor, and performance/latency has an impact on that cost, along with the scalability tradeoffs.

Shifting gears to AI features, we now have a new set of parameters and financial models to evaluate. But it is no different from the muscle we have already built as an organization. Choice of models, token pricing, and decisions on whether to build custom pipelines or not are all factors that shape these conversations with AI as a backdrop.

We are not building foundational models. Instead, our focus is on unlocking and delivering value to brands and creators and making sure our core systems are performing at scale, in the new world of AI and agents.

Human in the loop: Automation vs human creativity

There’s a fine balance between automation and authenticity. How do you ensure AI enhances creative workflows and decision-making without diluting originality?

Kumaran: Creativity is not easily replicable. And so fundamentally, we believe we need to build tooling and capabilities to support our creator population.

There are so many dimensions to the creative process that are stylistic and nuanced. And one of our AI focus areas is on understanding patterns in data, not creating the content. We use complex ML models and pipelines to analyze creator, content, and performance data at scale. But the creative work always stays human.

AI should empower and enable humans, not replace them. Even though 95 percent of brands use AI in workflows, relationships, and creative direction, we believe it will remain largely human-powered for years to come. This balance is critical to maintaining authenticity, but with the added bonus of improved velocity around the creative process!

How do you design ‘human-in-the-loop’ AI systems that give marketers control and visibility? How do you enable them to guide or override AI-driven recommendations and campaign decisions?

Kumaran: Our goal is to build human-in-the-loop AI systems that empower people to do more, deliver greater impact, and unlock performance at scale.

Control and visibility are two important factors shaping our product decisions. We build systems that let marketers review, adjust, and override recommendations through clear workflows and rapid feedback loops. The system brings forward insights and intelligence from large-scale data. However, humans decide what’s right for the brand and control when and how they wish to leverage this intelligence.

This becomes even more important as brands move from one-off campaigns to long-term creator partnerships.

Operational complexity and enterprise adoption

What organizational changes, across skills, governance, and workflows, are needed for enterprises to effectively adopt and scale AI?

Kumaran: Flexibility and control over workflows are essential for organizations to effectively adopt AI.

They enable teams to evolve internal processes and build the skills needed to leverage them at scale. CreatorIQ’s platform and tooling are built to progressively meet our customers where they are in their AI journey. We can work with different structures and workflows at enterprise scale. For customers who are already far into the agentic journey, governance is important.

And we have frameworks for safety and data privacy as key underlying anchors that will facilitate that journey.  

How do you ensure interoperability with existing martech stacks while introducing AI-native capabilities?

Kumaran: We focus on integrating with the existing systems organizations use to create consistency across workflows and measurements.

When you connect performance data with payments across 80+ markets and 60+ currencies, you get a broad view of ROI. The goal is to reduce fragmentation without forcing teams to change how they operate.

Governance, Trust and Safety

What technical and governance approaches are needed to ensure transparency and auditability in AI-driven selection, measurement, and brand safety decisions?

Kumaran: You need systems that make decisions understandable and traceable. As an enterprise-focused offering, we have built in audit trails, configurable policies, and brand-specific controls.

So that teams can see how decisions are made and adjust them over time. This is even more relevant when AI is inserted into the workflows. As AI analyzes content across text, video, and audio, context becomes increasingly important.

That visibility helps teams scale safely as creator content continues to grow significantly faster than organization-owned content.

Leadership transition and vision

You’re stepping into the CTO role at a pivotal moment. What attracted you to CreatorIQ, and what excites you most about leading its next phase of innovation?

Kumaran: Stepping into this role now is especially exciting given how quickly AI is transforming the industry. What makes this moment pivotal is the combination of our strong position in creator marketing, the depth of our data, which is critical for powering AI, and the maturity of our platform.

What excites me most is the opportunity to build on that foundation. AI is unlocking entirely new user experiences and more programmatic, scalable capabilities.

We’ve already begun integrating large language models and custom ML models into our products (SafeIQ is a great example). And we’re now focused on accelerating that roadmap to deliver more advanced, agentic capabilities and drive greater value for our customers.

As you take on this role, how are you thinking about balancing continuity with transformation? What do you plan to preserve in terms of leadership, and what do you intend to rethink?

Kumaran: Balancing these two is crucial. High-performing, scalable products are essential to sustaining a healthy business. That foundation, in turn, enables investment in new innovation and transformative capabilities.

Our commitment to customers is non-negotiable. So is the pace of delivering high-quality features. As we carve out cycles and invest in transformation, these remain constant. We will continue to preserve and invest more deeply in our data pipelines and APIs. At the same time, we are focused on strengthening data quality and expanding data volume.

Content is crucial to the equation, and the efficiency with which we process and meaningfully extract signals is core to our business. There are other areas that are ready for disruption, such as the ability to run experiments at scale, how we think about automation and workflows, and the underlying infrastructure that must be ready to handle both complex pipelines and workflows.

Future Outlook

Looking ahead 3–5 years, how will AI reshape how organizations discover talent, run campaigns, and measure ROI in influencer marketing?

Kumaran: I believe building human-in-the-loop AI systems will continue to be crucial, so with that as the context, discovery of creators in the future is going to be driven not by using search with qualifiers but by enabling more complex conversational approaches, with retained context, that are understood by an intelligent system to curate creators based on defined outcomes.

This requires so many levels of simplification with the interaction model, including eliminating noise and being precise on many dimensions that must be intelligently matched to the outcomes. The combination of our Creator Graph, our ML pipelines that enable semantic intelligence, and our ability to look back across vast amounts of data will enable us to do this seamlessly.

ROI will come in the form of performance enabled through simplification of workflows and outcomes that amplify the ROI. 

What will differentiate the winners in this new space: proprietary data advantage, AI capability, or ecosystem partnerships?

Kumaran: It’s the combination of all three. Data is the foundation, AI helps make sense of it, and partnerships allow organizations to activate across platforms.

What will really matter is how well companies bring those together into repeatable systems. The next phase is not just about scale, it’s about sustainability. The creator economy only works long-term if value is more predictable and better shared across creators and brands.

Any advice you would like to give to future tech leaders?

Kumaran: With AI everywhere, there is often the pressure to jump into building without thinking through the foundational elements needed to power AI capabilities at scale.

This includes pipelines, data, large-scale infrastructure, and financial modeling that help us understand the true cost of delivering a capability. We make sure the solutions we put forth consider the two most important elements: value and the humans who make it possible.

As tech leaders, while it is tempting to go right into the building phase, having a strategy and a plan to get there is key.

In brief

AI’s real impact comes from how it is operationalized, not just how it is imagined.

The focus is shifting from standalone capabilities to integrated systems – where data, models, workflows, and human judgment work in sync. The organizations that pull ahead will be those that reduce complexity behind the scenes while giving users greater control and clarity.

In that sense, the future isn’t about more AI – it’s about better-orchestrated AI that fits seamlessly into how decisions are made, and value is created.