Opening a New Wave of AI Agents? Anthropic’s New Skills Hint at a Smarter Future for Automation

With AI models now twice as fast and a third of the cost, the era of specialised, single-use agents may already be over before it truly began.

For years, the technology sector has been captivated by the promise of AI agents—autonomous systems designed to execute complex, multi-step tasks on our behalf. From booking holidays to managing complex software deployments, the vision was a world populated by a diverse array of digital assistants, each a specialist in its own domain. However, a recent series of quiet but powerful updates from AI research company Anthropic suggests a radical departure from this paradigm. With the launch of the lightning-fast Claude Haiku 4.5 and the introduction of a new ‘Skills’ framework, Anthropic is signalling a future where we don’t need an army of different agents. Instead, we will have one highly capable agent that can be taught the specific skills required for any given task. This is a profound shift, and for professionals in project delivery, it represents a pivotal moment. The focus is no longer on finding the right agent for the job, but on teaching the agent you have the right skills. This is the dawn of true AI-driven workflow automation.

The Haiku Revolution: Speed Becomes the New Frontier

On October 15, 2025, Anthropic released Claude Haiku 4.5, a model that redefines the trade-off between performance, speed, and cost [1]. What was considered state-of-the-art just five months prior with the release of Claude Sonnet 4 is now available at one-third of the price and more than double the speed. As one industry observer noted, “just six months ago, this level of performance would have been state-of-the-art” [1]. This is not merely an incremental improvement; it is a step-change that unlocks entirely new possibilities for real-time applications. For project managers, whose days are a constant juggle of low-latency tasks—from responding to client queries to providing instant feedback to development teams—Haiku 4.5’s combination of high intelligence and remarkable responsiveness is a game-changer. The ability to have a near-frontier model that operates at this velocity means that AI assistance can be woven seamlessly into the fabric of daily work, rather than being a cumbersome, time-lagged process.

Furthermore, Anthropic has demonstrated that its more powerful model, Sonnet 4.5, can act as an orchestrator, breaking down a complex problem into a multi-step plan and then deploying a team of multiple Haiku 4.5 models to execute the sub-tasks in parallel [1]. This mirrors the very essence of project management—deconstructing a large, complex goal into manageable work packages and assigning them to team members. The potential for project professionals to leverage this capability to automate complex workflows, manage dependencies, and accelerate delivery is immense. As one developer commented, “Speed is the new frontier for AI agents operating in feedback loops” [1].

From Agents to Skills: A New Philosophy of Automation

The most significant strategic shift, however, comes with the introduction of ‘Skills’ for Claude, announced on October 16, 2025 [2]. A ‘Skill’ is essentially a folder containing a set of instructions, scripts, and resources that Claude can load on demand to become proficient at a specific task. This could be anything from analysing an Excel spreadsheet according to a company’s financial reporting standards to drafting marketing copy that adheres to specific brand guidelines. Instead of building a bespoke ‘Finance Agent’ or a ‘Marketing Agent’, you teach Claude the skill of finance or marketing. As Brad Abrams, a Product Lead at Anthropic, explained, “the thing that’s interesting to me about Skills is basically about agents… it’s not about meeting arbitrary benchmarks — it’s about being able to do the task you need at your own company” [3].

This is the core of the Project Flux perspective: the future of work is not about replacing humans with a plethora of disparate AI agents, but about empowering humans with AI that can learn and execute the repeatable processes that form the backbone of any project. Nearly every aspect of project delivery, from risk assessment to resource allocation, involves a series of automatable workflows. The ‘Skills’ paradigm speaks directly to this reality. It suggests a future where a project manager can create a custom ‘Skill’ for their team’s unique change management process, or their specific method for generating client progress reports. This ‘Skill’ can then be shared and executed flawlessly by Claude, ensuring consistency, eliminating human error, and freeing up the project team to focus on higher-value activities like strategic decision-making and stakeholder engagement.

This approach is already gaining traction with major enterprise players. Box is using Skills to transform stored files into standardised PowerPoint presentations and Word documents, while Rakuten and Canva are leveraging the technology to embed their services more deeply into agentic workflows [2]. The message is clear: the value lies not in the agent itself, but in its ability to be customised and taught. This democratises automation, moving it from the exclusive domain of software developers to the hands of the subject matter experts who actually perform the work.

The Project Delivery Professional of Tomorrow

The implications of this shift for project delivery are profound. The role of the project manager will evolve from a coordinator of human tasks to an architect of AI-driven workflows. The most valuable skill will no longer be the ability to manage a Gantt chart, but the ability to deconstruct a complex project into a series of automatable ‘Skills’ that can be taught to an AI. This is not a threat, but an opportunity. It is a chance to move beyond the administrative drudgery that consumes so much of a project manager’s time and to become a true strategic adviser, focused on delivering value and driving outcomes.

The future of project delivery is not about being replaced by an agent. It is about becoming the one who teaches the agent. It is about building a library of custom ‘Skills’ that encapsulate your team’s unique expertise and processes. It is about leveraging the incredible speed and power of models like Haiku 4.5 to execute those skills with an efficiency that was previously unimaginable.

The revolution is here. Don’t just manage your next project—automate it. Start building your skills library today and redefine what’s possible in project delivery.

References

[1] Anthropic. (2025, October 15). Introducing Claude Haiku 4.5. https://www.anthropic.com/news/claude-haiku-4-5 

[2] Anthropic. (2025, October 16). Claude Skills: Customize AI for your workflows. https://www.anthropic.com/news/skills 

[3] The Verge. (2025, October 16). Anthropic turns to ‘skills’ to make Claude more useful at work. https://www.theverge.com/ai-artificial-intelligence/800868/anthropic-claude-skills-ai-agents 

[4] VentureBeat. (2025, October 16). How Anthropic’s ‘Skills’ make Claude faster, cheaper, and more consistent. https://venturebeat.com/ai/how-anthropics-skills-make-claude-faster-cheaper-and-more-consistent-for 

[5] Neowin. (2025, October 16). Anthropic adds composable 'Skills' to Claude for specialized workflow support. https://www.neowin.net/amp/anthropic-adds-composable-skills-to-claude-for-specialized-workflow-support/