Meet Alan: Your New CRE Agent Who Works 24/7, Never Makes Mistakes, & Costs A Fraction of Human Labor

Andrew Ng, a prominent figure in the field of Artificial Intelligence, recently made a public presentation where likens AI to “the new electricity,” a general-purpose technology with the power to unlock countless new applications. While much of the recent buzz in AI has centered on foundational models and semiconductors, Ng emphasizes that the true value and biggest opportunities lie in the application layer. Among the myriad advancements in AI, Ng is most excited about a specific technical trend: agentic AI workflows, which he considers the single most important AI technology to pay attention to.

Beyond Single-Shot Prompts: The Agentic Revolution

The typical way most users interact with large language models (LLMs) today is through “zero-shot prompting”, asking the AI to generate a complete output in one go, much like writing an essay from start to finish without backspacing. While LLMs perform surprisingly well this way, Ng highlights that humans don’t do their best work like this, and neither do AIs.

This is where agentic workflows come in. According to Ng, an agentic workflow is an iterative process where an AI system breaks down a complex task into smaller steps, performs research, drafts, critiques its own output, and revises, often looping through these stages multiple times. This approach, akin to a human thinking, researching, and revising, takes longer but results in significantly better and more robust outputs. For example, a legal team could use an agentic workflow to process complex documents, or a healthcare system could leverage it for diagnostic assistance.

The performance gains are stark. Ng cites the HumanEval Benchmark, which measures an LLM’s ability to solve coding puzzles. While GPT-3.5 achieved 48% accuracy and GPT-4 a much improved 67%, GPT-3.5 with an agentic workflow could achieve up to 95% accuracy. This demonstrates that the improvement from using an agentic workflow can “dwarf” the improvements from moving to a larger, more advanced foundational model alone.

The Four Pillars of Agentic Design

Ng identifies four major design patterns that builders are using to construct agentic workflows in their applications:

1. Reflection: This pattern involves an LLM critiquing its own output. For instance, a “coder agent” might generate code, then be prompted to review and critique that code, even suggesting improvements. This iterative self-correction, often combined with unit tests, significantly boosts performance from a baseline level.
2. Tool Use: LLMs are prompted to decide when to make API calls, such as searching the web, executing code, or performing specific actions like issuing refunds or sending emails. This expands the capabilities of agentic workflows by allowing LLMs to interact with external systems and data. Ng notes that LLMs are now often explicitly tuned to support tool use, which creates a higher ceiling for what agentic workflows can achieve.
3. Planning: For complex requests, an LLM can be prompted to devise a sequence of actions or steps to achieve the desired outcome. This allows the AI to break down a large task into manageable sub-tasks and execute them in a logical order, often involving different models or tools for each step.
4. Multi-Agent Collaboration: Instead of a single LLM performing all tasks, this pattern involves prompting an LLM to play different roles at different points in time, simulating multiple agents interacting with each other to solve a task. Ng draws an analogy to multi-threading on a CPU – while it’s still one processor, abstracting tasks into different “agents” can help developers break down and solve complex problems more effectively, often leading to significantly improved performance for various tasks.

The Dawn of Visual AI Agents

Ng is particularly excited about the rise of large multimodal model (LMM) based agents, extending the power of agentic workflows beyond text to include image and video data. Just as with LLMs, LMMs can perform better with an iterative, step-by-step agentic approach compared to zero-shot prompting.

He demonstrated this with a “vision agent” capable of complex visual AI tasks. For example, by uploading an image of a soccer game and asking it to count players, the agent generated Python code to perform object detection and counting, then provided the accurate result. This capability allows businesses with vast amounts of visual data (images, videos) to extract significant value that was previously difficult to obtain. Other demonstrations included splitting video clips to find specific events (like a goal being scored) and generating metadata for video content, enabling searchable video databases. This significantly lowers the barrier to building complex visual AI applications.

Impact on the AI Stack and Development Practices

The emergence of agentic AI is not just changing how applications are built but also evolving the AI stack itself. Ng points to a new, emerging agentic orchestration layer (like LangChain or LangGraph) that makes it easier for developers to build these complex applications.

This rapid pace of development driven by generative AI means that while prototyping machine learning models has become much faster (days instead of months), other parts of the software development process – like product design, software integration, DevOps, and MLOps – still take time. However, the speed of ML prototyping is putting pressure on organizations to accelerate these other pieces as well. Ng advocates for a new mantra: “move fast and be responsible,” emphasizing the ability of smart teams to prototype quickly and evaluate robustly without shipping harmful products.

A critical bottleneck in this accelerated process is evaluations (evals). In the past, collecting test data was a small additional cost to training data. Now, with LLM-based apps often not requiring training data, collecting thousands of test examples becomes a significant bottleneck. Ng notes that building and collecting data are often done in parallel rather than sequentially, and there’s still much innovation needed in how evals are built. He suggests that many teams delay systematic evals, but even quickly thrown-together, imperfect evals can be immensely helpful in complementing human judgment and incrementally improving systems.

Key Trends and Future Challenges

Ng identifies several crucial trends supporting the agentic AI revolution:

• Faster Token Generation: Agentic workflows require reading and generating a lot of text (tokens). Efforts to speed up token generation through semiconductor and software advancements will make agents much more efficient.
• LLMs Tuned for Tool Use: Modern LLMs are increasingly optimized not just for answering human queries but explicitly for supporting tool use and fitting into iterative agentic workflows.
• Rising Importance of Unstructured Data Engineering: With generative AI’s prowess in processing text, images, and video, managing and deploying unstructured data to create value is becoming a major effort for businesses.
• Image Processing Revolution: While text processing is here, the image processing revolution is rapidly advancing, significantly increasing the range of possible applications.

He also touches on challenges and underrated areas:

• Bridging Business Needs to Agentic Workflows: It’s still difficult for businesses to break down existing processes into the right granularity of micro-tasks for agentic workflows. This requires a rare skill set to define steps, branches, and effective evals.
• Tactile Knowledge for Debugging: Building agents often requires “tactile knowledge” – the ability to quickly diagnose issues by looking at output traces and making informed decisions on what to do next. This skill is built through practice with various AI tools and understanding their limitations.
• Underrated Opportunities:
  • Voice Stack Applications: Ng sees massive, often underestimated, opportunities for voice-based agentic applications in enterprises. While real-time voice can be challenging due to latency, agentic voice workflows offer more control and significantly reduce user friction, as people feel more comfortable speaking than typing for many applications.
  • AI-Assisted Coding: Ng firmly believes that AI-assisted coding makes developers significantly faster and that everyone should learn to code, as it enables better instruction to computers across all job functions. He dismisses the idea that AI will automate away coding jobs, comparing it to past fears when programming languages became easier.
  • MCP (Model-Client Protocol): Ng sees MCP as a “fantastic first step” towards standardizing the interface for agents to plug into various tools and data sources, greatly simplifying the “plumbing” of data integrations. While early and still evolving, it promises to reduce the integration effort from N*M to N+M.
  • Agent-to-Agent Communication: Though very early, the concept of agents from different teams successfully interacting is a future frontier. Currently, multi-agent systems primarily work best within a single team due to protocol understanding.

In this presentation, Andrew Ng is immensely optimistic about the future of AI agents, believing they are expanding the realm of what’s possible and opening up countless new applications. The ability to experiment and build faster than ever before makes this an exhilarating time for builders in the AI space.

The landscape of work is shifting beneath our feet. For decades, the pace of business has been steadily accelerating, pushing human capacity to its limits. Now, a new force is emerging, one that promises not just to keep pace but to redefine what’s possible: Artificial Intelligence. This isn’t just another tech upgrade; it’s a fundamental rewiring of how we work, as profound as the Industrial Revolution or the advent of the internet. We are standing at the dawn of, what Microsoft refers to as the Frontier Firm.

The Dawn of a New Reality: Intelligence On Tap

Imagine a world where intelligence is no longer a scarce resource. Picture a scenario where you can access “intelligence on tap”, abundant, affordable, and available on demand, just like electricity. This is the promise of AI and, more specifically, AI workflow agents. These digital colleagues are fundamentally transforming knowledge work, allowing businesses to operate with unprecedented agility and scale.

The Frontier Firm is the blueprint for this new era. It’s an organization that seamlessly blends human judgment with machine intelligence, creating a powerful synergy. In a Frontier Firm, every employee takes on a new, critical role: the “agent boss.” This isn’t a futuristic concept; it’s here now. Leaders are already seeing the writing on the wall, with 81% expecting AI agents to be integrated into their company’s AI strategy within the next 12-18 months. The era of small-scale pilots is over; broad adoption is not just an option, it’s a necessity for survival and growth.

The “Why Now?”: Bridging the Capacity Gap

Why the urgency? The answer lies in what’s called the “capacity gap.” Businesses today face escalating demands, yet human capacity remains finite. Consider this: 80% of the global workforce reports lacking enough time or energy to do their work effectively, even as 53% of leaders demand increased productivity. This creates a chasm between what needs to be done and what humans alone can achieve.

This is where “intelligence on tap” becomes the game-changer. AI and agents offer a new lever for growth, providing “digital labor” that can scale capacity as needed, filling that critical gap. It’s about more than just doing existing tasks faster; it’s about redefining knowledge work entirely. The days of “I send emails” or “I create pivot tables” are giving way to “I create and manage agents.” This shift isn’t about replacing humans but rather freeing them from drudgery, allowing them to focus on higher-value, more strategic tasks that truly move the needle.

The Three Phases of AI Transformation: A Non-Linear Ascent

According to Microsoft, in their 2025 survey, the journey to becoming a Frontier Firm isn’t a single leap, but a progressive, multi-phased transformation. While organizations may find themselves operating in all three phases simultaneously, understanding each stage provides a roadmap for strategic integration:

Phase 1: Human with Assistant – Boosting Personal Productivity

At its core, this phase is about empowering every employee with a personal AI assistant. Think of it as a super-powered digital sidekick that handles routine tasks, allowing individuals to work faster and more efficiently.

• Concept: Every employee gains an AI assistant, enhancing personal productivity and removing repetitive, time-consuming drudgery.
• Focus: Individual-level productivity. Humans remain the primary drivers of work, but with AI amplification.
• Examples: AI automating CRM record updates, generating meeting summaries, drafting initial emails, or transcribing notes.

Phase 2: Human-Agent Teams – Elevating Team Efficiency

As organizations mature, AI evolves beyond individual assistance to become an integral part of team dynamics. In this phase, agents join teams as “digital colleagues,” taking on specific, delegated tasks.

• Concept: AI agents become active members of teams, executing specific tasks under human direction.
• Focus: Enhancing team efficiency and collective productivity. Agents equip human employees with new skills, freeing them for more creative and valuable contributions.
• Examples: An agent developing a preliminary go-to-market plan, triaging customer support tickets, managing project timelines, facilitating collaborative brainstorming sessions, or even contributing to content creation.

Phase 3: Human-Led, Agent-Operated – Redefining Workflows

This is the pinnacle of the Frontier Firm transformation, where agents become the primary producers of work, operating entire business processes and workflows. Human involvement shifts to strategic oversight and critical decision-making.

• Concept: Humans define the strategic direction, while agents execute end-to-end business processes and workflows. Humans monitor progress via dashboards and intervene for exceptions.
• Focus: New career opportunities for humans as “agent managers” or “workflow orchestrators.”
• Examples: Agents managing entire supply chain logistics, with humans overseeing the system and managing high-level relationships or exceptions. Agents autonomously prospecting and qualifying sales leads at scale, allowing human sales teams to focus solely on closing deals.

It’s crucial to understand that this progression isn’t strictly linear. Organizations will experience, what Microsoft calls a “jagged frontier” of AI transformation, with different departments or teams potentially operating in various phases concurrently. The key is to embrace this dynamism and strategically guide your organization through each stage.

Actionable Steps to Begin Your Frontier Firm Journey

The time for deliberation is over; the time for decisive action is now. Here are concrete steps you can take to begin your transformation into a Frontier Firm:

1. Hire Your First Digital Employees

Treat AI agents like any other team member. Define clear roles for automation, onboard them thoroughly, assign ownership, and most importantly, measure their performance. Start small, but start with a clear purpose.

2. Set Your Human-Agent Ratio

This is a critical new metric. Identify which processes are ripe for full automation versus those requiring human-AI collaboration. The optimal balance will vary: for repetitive, rules-based tasks, one human can effectively manage many agents. For tasks requiring significant judgment or nuance, more human oversight will be necessary. The goal is to optimize agent efficiency while ensuring robust human guidance and oversight.

3. Get to Broad Scale—Fast

Move beyond pilot projects. Identify high-need areas within your organization – operations, customer service, finance – where AI can deliver measurable impact quickly. Once validated, reinvest those gains to scale AI adoption across the enterprise. The goal is pervasive integration, not isolated experiments.

4. Conduct an AI Skills Gap Analysis

The future workforce requires new competencies. Assess your current AI capabilities and identify the skills needed across three critical categories:

• Foundational AI Literacy: Essential for all employees to understand AI’s capabilities and limitations.
• Intermediate AI Skills: For domain specialists who will work directly with AI tools and agents.
• Advanced AI Expertise: For technical specialists who will build, train, and maintain AI systems.

5. Adopt Microsoft’s “Build, Buy, Bot, Borrow” Model for AI Talent

This multifaceted approach ensures you have the necessary AI expertise:

• Build: Develop internal talent through structured learning programs and hands-on experiential learning.
• Buy: Strategically hire for highly specialized AI roles (e.g., AI/ML engineers, data scientists) when internal development isn’t feasible or fast enough.
• Bot: Deploy AI tools, automations, and agents to augment your existing workforce, handling routine tasks and augmenting decision-making. This is about leveraging AI itself as a talent solution.
• Borrow: Access scarce expertise through strategic partnerships with vendors, consultants, academic institutions, or talent exchanges.

6. Create an AI Learning Culture

Learning is no longer an event; it’s a continuous process. Design targeted learning experiences for executives, managers, and individual contributors. Foster a growth mindset and empower “AI champions” who can evangelize and guide others. Leadership modeling is key here – leaders must demonstrate a commitment to continuous learning in AI.

7. Address AI Anxiety and Resistance

It’s natural for employees to have concerns about job displacement, privacy, or reliability. Proactively address these anxieties by demonstrating how AI augments human capabilities, rather than replacing them. Showcase successful pilot projects, highlight skills enhancement opportunities, provide clear role evolution roadmaps, and celebrate shared successes. Emphasize that AI frees humans for more creative, engaging, and impactful work.

The journey to becoming a Frontier Firm is not without its challenges, but the rewards are immense. By embracing this transformation, you’re not just adapting to the future; you’re actively shaping it, positioning your organization for unprecedented growth, innovation, and human potential.