Agentic AI adoption in insights
Editor’s note: Thor Olof Philogène is CEO and founder at Stravito, headquartered in Stockholm, with experience in enterprise technology, revenue growth and AI-driven insights platforms. He holds a M.S. in business administration from the Stockholm School of Economics.
AI has already reshaped how insights teams search, synthesize and automate. What hasn't arrived yet – despite the headlines – is a world of autonomous AI agents running research workflows end-to-end. The path there is more complex, and more human, than the hype suggests.
According to Deloitte’s 2026 “State of AI in the Enterprise” report, nearly three-quarters of companies plan to deploy agentic AI within the next two years. Yet only 21% report having a mature governance model in place. This growing gap between ambition and readiness helps explain why, for research and insights professionals specifically, adoption will be gradual rather than instantaneous.
From autonomy to augmented agency
Agentic AI is often portrayed as an imminent revolution – systems that autonomously complete multi-step tasks, act on goals and manage workflows across platforms with minimal human input. But this narrative oversimplifies where real progress is happening.
In practice, agentic AI is already delivering real value, but the most effective deployments are not replacing humans. They're augmenting them, pairing increasing automation with guardrails that keep people firmly in the loop.
The reality behind the AI revolution
At its core, agentic AI is designed as a dynamic collaborator, one that doesn’t just respond to queries but can proactively execute tasks aligned with business objectives. But to deploy it effectively, organizations need deep integration with internal systems, robust governance and, most importantly, trust.
Many teams are still constrained by outdated data management pipelines, siloed information and legacy tech stacks not built for intelligent automation. Expecting AI agents to handle business-critical workflows without human input immediately ignores the accumulated complexity of enterprise environments.
The risks of moving too fast aren't only operational. AI that returns fast answers without verifiable sourcing can quietly become a generator of confident-sounding misinformation. For insights teams, where decisions carry real commercial weight, depth and verifiability matter as much as speed.
Agentic autonomy is a ladder
Think of agentic AI as moving up a ladder in which systems earn greater autonomy through proven reliability. Each successful rung proves that it's capable of more. Early stages, for example, focus on assistance like suggestions, drafts, routing and summaries, with humans making all final decisions. As confidence grows, systems can take on larger pieces of work, still requiring review and approval. Halfway up a system might autonomously plan a research workflow, read 50 reports and validate its own outputs before presenting a synthesized answer. The human didn't do the research plan work, but they're the one who decides what it means and what to do next. That division of labor is the point.
Agentic AI: Start small, scale with trust
To move from hype to durable value, leaders need to think of agentic AI as an operating shift rather than a single release. Progress comes not from how quickly autonomy is introduced, but from how intentionally it is earned.
The strongest AI enterprise deployments begin with a clearly defined job to be done – a specific research outcome to improve – and start small. Early applications should deliver value even without full autonomy. From there, responsibility expands gradually, with clear boundaries around what systems can do, when human review is required and how decisions are monitored.
Visibility here means being able to trace an AI's reasoning back to the source – understanding not just what conclusion was reached, but which evidence it drew on, what it prioritized and where gaps remain. That kind of auditability is what separates a system teams will trust from one they'll quietly stop using.
The progression also helps surface deeper organizational issues. Early deployments often reveal gaps in things like data quality, undocumented processes and beyond.
For research leaders, the opportunity with agentic AI is not to remove humans from the equation, but to finally unlock the intelligence their organizations already own – turning years of research that lives in files into answers decision makers can actually trust and act on.