Terms of engagement: Agentic AI | Feature | Research live
Language evolves quickly in tech. One term you’ll hear more and more is “agentic AI” — and understanding it in plain English helps cut through the hype and focus on what it can really do.
What is agentic AI?
Agentic AI is an artificial intelligence system designed to act on your behalf: it can plan, take actions, and make decisions autonomously to achieve a goal. Think of it not as a single chatbot that replies to prompts, but as an orchestrated set of software agents that coordinate to move work forward without constant human nudging.
- AI agent: a single software program built to independently complete a specific task (for example, pulling a report or booking a meeting).
- Agentic AI: a coordinated system of multiple agents that collaborate, plan, and execute multi-step workflows end to end.
- Generative AI: models that create content (text, images, code, etc.) in response to prompts, which can be one part of an agentic system.
Why it matters now
AI is rapidly weaving into everyday operations. Organisations are moving beyond simple prompt-response tools and instead deploying agents inside real workflows — to fetch and summarise information, trigger actions in business systems, and follow up when conditions change. This shift is reshaping consumer experiences and commercial strategies alike.
Agentic commerce: AI that shops for you
In retail, “agentic commerce” describes AI agents acting on behalf of shoppers. These agents can compare products and prices, check stock, apply loyalty benefits, monitor for price drops, and even place orders within user-defined rules. For brands, this changes the rules of engagement online: you’re not only marketing to humans; increasingly, you’re persuading their agents — with accurate data feeds, clear policies, robust integrations, and trustworthy performance signals.
What agentic AI can do in research
Across brands, tech platforms, and insights agencies, agentic AI is starting to run distinct segments of the research workflow. Examples include:
- Study design: drafting briefs, proposing methodologies, and aligning them to business objectives using prior learnings.
- Fieldwork orchestration: recruiting participants, launching surveys, monitoring quotas, and adapting sampling in real time.
- Data processing: cleaning datasets, coding open-ends, and generating first-pass summaries and visualisations.
- Analysis and synthesis: connecting internal and external data sources, identifying patterns, and running simulations to explore likely outcomes.
- Knowledge activation: indexing prior research, retrieving relevant evidence for new questions, and nudging teams with timely insights.
The practical benefit is speed and consistency: agents don’t tire, they document decisions, and they can hand off tasks reliably. But the bigger shift is continuity. Rather than episodic projects, agentic systems can sustain a rolling “insight loop,” updating learnings as new signals arrive and prompting action when thresholds are hit.
Expert view: augment, don’t sideline
Louisa Livingston, chair of the MRS Advanced Insights and Analytics Council, describes agentic AI as a step beyond today’s prompt-driven tools. In her view, the technology’s real promise is its ability to progress tasks across tools and data sources, enabling faster, more responsive decision-making. The efficiency gains are significant — yet the strategic edge still comes from human expertise: interpreting context, understanding commercial realities, and framing the right questions. For research teams, the opportunity is to use agents to elevate this strategic work rather than replace it.
Getting started: principles, not just pilots
If you’re considering agentic AI, a few practical principles apply:
- Define guardrails: set goals, budgets, data access, and escalation points so agents know when to act — and when to defer to people.
- Instrument everything: log plans, actions, and outcomes to enable audit, learning, and continuous improvement.
- Integrate with sources of truth: connect to approved data and systems to avoid drift and duplication.
- Keep humans in the loop: assign clear ownership for oversight, interpretation, and final decisions.
- Measure impact: track cycle time, cost to insight, decision quality, and downstream business effects.
The bottom line
Agentic AI moves AI from “answering prompts” to “achieving outcomes.” For commerce, it will influence how products are discovered, compared, and purchased. For research, it promises faster, more connected, and more continuous insight generation. The organisations that benefit most won’t just deploy agents; they’ll redesign workflows so that autonomous systems accelerate the human judgment that still defines great strategy.
