AI Agent Architect’s Monday Morning Briefing 🌟

Week Commencing March 19th 2025

May 19, 2025

Happy Monday!

As always, here’s what caught my eye in the last few days that’s worth sharing.

Today's Digest (TL;DR) 📌

1️⃣ AI Agent Adoption Surges – Read more
2️⃣ Klarna's AI Customer Service Shift – Read more
3️⃣ Strategic AI Talent Trends – Read more
4️⃣ AI Unlikely to Replace Jobs en Masse – Read more
5️⃣ The Difference Between AI Agents and Agentic AI – Read more

On My Radar Over the Last Seven Days 🛰

A paper from Cornell University goes into fascinating depth on what separates genuine AI Agents from the flood of overhyped automation tools flooding the market.

"AI Agents vs. Agentic AI: A Conceptual Taxonomy" cuts through the promotional fog to establish clear boundaries around what constitutes true agency in artificial intelligence.

Where no-code platforms merely stitch together predefined actions with colorful drag-and-drop interfaces, authentic AI Agents demonstrate legitimate cognitive capabilities—they actively reason about their environment, make decisions autonomously, and adapt their strategies over time.

The researchers methodically document how proper agents integrate sophisticated architectural components like causal understanding, persistent memory, and dynamic planning mechanisms that simply don't exist in conventional automation tools.

For business leaders and technologists trying to separate revolutionary technology from incremental improvements, this paper provides an essential framework to evaluate claims about "intelligent" systems.

If you're navigating investment decisions or implementation strategies around AI, this research offers the clarity needed to distinguish between transformative agent technologies and merely repackaged automation.

Strategic AI Talent Trends - AI Adoption Driving Talent Shortages

The demand for AI talent is surging.

63% of employers struggling to find qualified candidates. This shortage is particularly pronounced in non-high-tech firms, where AI adoption is reshaping workforce needs and creating a skills gap.

As companies increasingly rely on AI, the pressure to upskill existing employees and attract new talent will only intensify. 👉 Source

Emerging AI Roles & Career Paths - AI Solutions Architect

This role is gaining traction, with salaries ranging from $110k to $160k, reflecting the high demand for professionals who can design and implement AI systems.

Candidates typically need a strong background in cloud computing and AI frameworks, making this a lucrative path for those with the right technical skills.

The growth potential is significant, as businesses increasingly seek to integrate AI into their operations. 👉 Source

Critical AI Skills in Demand - Prompt Engineering

This skill is rapidly becoming essential, with a growing number of job postings highlighting its importance in AI development.

Professionals skilled in crafting effective prompts can significantly enhance AI tool performance, leading to better business outcomes.

To develop this skill, individuals should focus on practical applications and continuous experimentation with AI models. 👉 Source

Contrarian Corner - Are We All Doomed? 😐

This last week, the prevailing narrative suggests that AI will seamlessly replace a significant portion of the workforce, particularly in corporate roles, leading to widespread job displacement.

Proponents of this view often cite the rapid advancements in AI capabilities and the increasing automation of tasks as evidence.

However, a recent analysis indicates that while AI can enhance productivity, it is unlikely to replace jobs en masse; instead, it will lead to a shift in job roles rather than outright elimination.

A report from last week highlights that 85% of AI projects fail due to poor data quality and unrealistic expectations, suggesting that the hype around AI's capabilities often overshadows the practical challenges businesses face in implementation. This implies that AI leaders should focus on integrating AI as a tool to augment human capabilities rather than a wholesale replacement, ensuring that workforce transitions are managed effectively. 👉 Source

Key Trends of the Last 7 Days 📈

AI Agents Not Just Hot Air, Businesses Already Cashing In

Looks like AI agents aren't just fancy tech talk anymore. A whopping 88% of management are planning to splash more cash on them this year, with most companies already getting stuck in. Two-thirds say these digital helpers are boosting productivity, though many firms are still just dipping their toes rather than diving headfirst.

The real magic happens when multiple AI agents work together across departments - that's when the proper transformation kicks off. While most bigwigs reckon these agents will shake things up more than the internet did, only the forward-thinking lot are actually rebuilding how they work from the ground up. 👉 Source

Klarna's AI Customer Service Shift

Turns out those headlines about Klarna ditching their AI customer service bots were well off the mark. Rather than retreating, the Swedish fintech firm is actually handing more tasks to their digital helpers.

According to CEO Sebastian Siemiatkowski, their AI agent is handling about 1.3 million customer queries monthly - work that previously needed around 800 people.

While Klarna is bringing in some human agents for complex issues they used to outsource, they're simultaneously expanding what their AI can do. The bot now handles work equivalent to an extra 100 human staff compared to last year, manages two-thirds of all customer chats, and sorts problems in just two minutes versus eleven with humans. Siemiatkowski admits he might have been a bit optimistic about timelines - "You could place me in the Elon Musk box of saying everything's gonna happen in three months, when it probably will take three years" - but insists the number of jobs handled by AI at Klarna "will continue to go up."

Seems like they're finding the sweet spot where bots and humans can work together. 👉 Source

AI Deployment Watch: What's Working (or Failing) in the Wild 🚀🔥

AI Agent Use Cases

Companies are successfully implementing AI agents for tasks like customer support and invoicing, leading to measurable efficiency gains. This is exactly what I’m observing talking to clients every week - it’s the Repeatable Expertise pattern. Where it exists, opportunity awaits.

These deployments demonstrate the potential for AI to streamline operations and reduce costs. 👉 Source

Research Papers of the Last 7 Days 📚

Pre-Act: An Important AI Agent Architecture Paper

This recent paper from Uniphore introduces "Pre-Act," a novel approach that significantly improves AI agents' performance by creating detailed multi-step plans with reasoning before taking actions.

Unlike the standard ReAct pattern (reasoning + action), Pre-Act generates a comprehensive execution plan upfront that refines itself after each step, helping agents better handle complex tasks requiring long-term planning.

Their experiments show impressive results - their fine-tuned 70B model outperforms GPT-4 with a 69.5% improvement in action accuracy and 28% better goal completion on out-of-domain tasks. This architecture demonstrates that smaller, fine-tuned models can match or exceed larger proprietary AI systems while reducing latency and cost. 👉 Source

Agent-as-a-Service: A Breakthrough Multi-Agent Framework

This fresh research from Wuhan University introduces "Agent-as-a-Service based on Agent Network" (AaaS-AN), a service-oriented framework that solves a critical problem in multi-agent systems - how to organise and coordinate collaboration between AI agents at scale.

While existing protocols like MCP handle tool integration, AaaS-AN tackles the harder challenge of agent-level collaboration through two key innovations: a dynamic Agent Network that models agents as self-organising vertexes, and service-oriented agents with discovery, registration and interoperability protocols.

Experiments show impressive results, outperforming competing frameworks on mathematical reasoning and code generation tasks.

Most impressively, they demonstrated a system with over 100 agent services working together and released a dataset of 10,000 long-horizon multi-agent workflows to advance research in agent collaboration. 👉 Source

Teaching AI to Play the Market Like Humans

This study tested whether Generative AI agents like GPT-4 can mimic human behaviour in economic markets.

Researchers ran simulations where AI agents made price predictions in dynamic market environments, comparing their actions to those of humans in lab experiments. The AI agents showed human-like traits—especially bounded rationality and trend-following—but had less behavioural diversity and were highly influenced by how much memory and randomness (temperature) they were given.

The findings suggest LLMs could be useful tools for simulating realistic economic behaviours, but they still need refinement to fully capture the richness of human decision-making. 👉 Source

That’s your lot for this week. Keep your wits about you and remember: AI is a tool, not a magic wand.

Until the next one, Chris.

Enjoyed this post? Please share your thoughts in the comments or spread the word by hitting that Restack button.