The explosion of AI adoption has led individuals and enterprises to consolidate around a small number of frontier models. OpenAI’s ChatGPT and Google’s Gemini dominate day-to-day use. Microsoft’s Copilot is embedded across 70% of Fortune 500 companies. Anthropic’s Claude Code excels in augmenting coding, while Musk’s Grok leverages X’s huge array of content, yet their primary use cases remain confined to text and conversation. This dominance risks narrowing innovation.

Companies that stop at generative AI risk stalling at the surface. The next frontier is agentic AI – systems that plan, reason and act autonomously across workflows. Many of the most advanced and commercially viable agentic platforms are emerging outside the “big-model” ecosystem, where innovation moves faster and integration runs deeper.

What Is Agentic AI?

Agentic AI describes systems that don’t just respond to prompts - they think, plan and execute tasks with limited supervision. In practice, that means an AI agent can break down complex goals into smaller steps, interact with other tools and adjust its strategy based on feedback and context.

These systems still rely on large language models (LLMs), but with an added orchestration layer – often called the agentic mesh – that enables coordination, monitoring and adaptation. Instead of a single prompt-and-reply cycle, agentic AI operates in plan–act–observe loops, continuously refining outputs to achieve user-defined goals. In simple terms, AI evolves from a reactive assistant into a proactive delegate.

The risks of the mono-model

Relying on a single foundation model creates blind spots. Each model has its own strengths, weaknesses and biases. Some handle reasoning or summarisation better, while others excel at coding or visual tasks. Depending on one model increases the risk of flawed data, vendor lock-in and rising costs as prices and licensing terms change.

The UK Competition and Markets Authority warned in 2024 that a small cluster of firms – Google, Apple, Microsoft, Meta and Amazon – are forming an “interconnected web” of over 90 partnerships that could limit competition, data access and user choice. A “Coke versus Pepsi” dynamic may soon define enterprise AI, where switching providers or validating results becomes prohibitively difficult.

The advantages of being model-agnostic

A model-agnostic approach uses multiple models within a single platform, allowing organisations to harness the unique strengths of each. Modern cloud infrastructures already make this practical. AWS Bedrock, Azure AI Foundry, and Google’s Vertex AI Model Garden all expose a variety of proprietary and open-source models under one roof, enabling teams to compare, evaluate and switch as needed.

Model-agnostic doesn’t mean costly. Research from LMsys and IBM shows that routing and ensembling – intelligently sending each task to the most appropriate or cost-effective model – can maintain around 95% of a top model’s quality while significantly reducing cost. In other words, you can optimise for both performance and budget.

There’s also a governance benefit. Multi-model setups provide natural checks and balances: independent models can verify one another’s outputs, reducing the risk of hallucinations and bias.

To see this in action, consider how a global media organisation might use a model-agnostic agentic platform: one model could generate summaries of breaking stories, another could verify sources against databases, while a third drafts SEO-optimised headlines – all coordinated within a single workflow. The result is intelligence in motion: content created, verified, and delivered with precision.

From generation to orchestration

For many companies, poor ROI from generative AI isn’t failure – it’s a signal that they’ve reached the limits of phase one. The next phase is about integration and orchestration: deploying agentic systems that act, plan and learn across data sources and applications.

A model-agnostic approach unlocks flexibility, resilience and faster returns. Choose partners who specialise in multi-model agentic AI and prioritise intuitive, human-centred design. Non-technical teams adopt these systems more easily, and leadership can see measurable impact sooner.

The time to act is now. Generative AI was the beginning. Agentic, model-agnostic AI is the next leap – one that’s redefining how every business thinks, builds, and competes.

For more startup news, check out the other articles on the website, and subscribe to the magazine for free. Listen to The Cereal Entrepreneur podcast for more interviews with entrepreneurs and big-hitters in the startup ecosystem.