Next-generation AI requires a next-generation approach

The latest AI models, such as the new Fable 5 and GPT-5.6 Sol, possess reasoning capabilities that far surpass their predecessors. However, most users continue to interact with them using simple, single-shot prompts, effectively treating them like slightly better versions of older models. According to the AI Daily Brief, this approach misses the true potential of these systems. To capture their value, businesses must shift from giving simple instructions to defining high-level goals and engaging in iterative, collaborative loops with the AI.

The AI Daily Brief recently explored a growing gap between the capabilities of new frontier AI models and how people are using them. The episode highlights that while models like the hypothetical Fable 5 and GPT-5.6 Sol represent a new paradigm of AI—one with sophisticated reasoning and planning abilities—most users are stuck in old habits.

The core issue is that people are applying interaction patterns developed for older models (GPT-3.5, GPT-4) to this new class of technology. This results in underutilisation and a failure to realise the potential for significant productivity gains.

Shifting interaction patterns

The fundamental change required is moving from a transactional model of interaction to a collaborative one. Instead of providing a detailed, one-off prompt and expecting a final answer, the new approach involves treating the AI as a reasoning partner.

This table illustrates the difference:

For business owners and operators, continuing to use next-generation AI with last-generation methods is not just inefficient—it's a strategic risk. As competitors adapt their workflows to leverage true agentic capabilities, those who don't will fall behind.

The productivity plateau

If your team uses AI simply to write emails or summarise text a bit faster, you are hitting a productivity plateau. The real gains come from automating entire multi-step workflows, not just accelerating individual tasks. These new models are designed to handle complex chains of tasks, such as conducting market research, analysing the data, synthesising a report, and creating a presentation from it. Sticking to simple prompting leaves this enormous value on the table.

ROI on AI investment

Frontier models come with higher costs, whether through API calls or premium subscriptions. Using a a highly capable, expensive model for tasks that a cheaper, older model could handle is a poor return on investment. The justification for investing in state-of-the-art AI is its ability to deliver an order-of-magnitude improvement in output or to automate work that was previously impossible to automate. This only happens when you change how you work.

The shift to agentic workflows

This evolution in interaction is the practical foundation of agentic AI. An AI agent is a system that can understand a goal, make a plan, and use tools to execute that plan autonomously. The new models are built to power these agents. By failing to adapt, businesses are effectively ignoring the most significant development in AI since the launch of ChatGPT. The future of work in many sectors will revolve around teams of humans managing fleets of AI agents to execute complex business processes. Getting started with this new collaborative interaction model is the first step on that path.

Adapting to this new paradigm doesn't require a complete overhaul overnight. It's about a progressive shift in mindset and process. Here are four practical steps for business leaders to take:

1. Audit your current AI usage. Identify the top 5-10 ways your team currently uses AI. Are these simple, single-shot tasks (e.g., "summarise this email") or are they more complex? This benchmark will clarify where your organisation currently stands on the AI maturity curve.

2. Train for goals, not just prompts. Start training your team to think differently. Instead of asking "what prompt should I use?", encourage them to ask "what is the final outcome I want to achieve?" Frame AI tasks as projects to be delegated, not questions to be answered. Run small workshops where you practice giving the AI a high-level goal and then refining its plan.

3. Pilot one complex workflow. Select a single, high-value, multi-step business process. This could be lead generation research, creating sales proposals, or analysing customer feedback. Dedicate a small team to reimagine and automate this entire workflow using a new-generation model. Document the process, the time saved, and the quality of the output compared to the old manual method.

4. Shift your metrics of success. Move beyond measuring the speed of individual task completion. Instead, focus on the value generated by the entire automated workflow. Key metrics should include:

   • Hours of skilled labour saved and reallocated to higher-value work.
   • Reduction in errors or increase in quality and consistency.
   • Creation of new insights or outputs that were not possible before.
   • Overall impact on the business objective the workflow supports.