WELCOME, EXECUTIVES AND PROFESSIONALS.

Automating knowledge work isn’t new, but with the advent of gen AI, automation has become more expansive, adaptable and faster.

In many enterprises, this has manifested as Agentic AI, along with concepts like agents, AI agents, agentic-powered automation, and more.

I've analyzed dozens of reports to unpack viewpoints.

Here are 19 must-reads:

BAIN & COMPANY

Will Agentic AI eliminate or enhance SaaS?

Brief: Bain & Company Partner Chuck Whitten explores whether agentic AI will eliminate or enhance SaaS. History, he argues, shows technological revolutions tend to expand ecosystems rather than replacing them outright.

Breakdown:

• Leaders like Satya Nadella suggest agentic AI could upend the SaaS model entirely. Others believe AI will simply enhance SaaS.

• Technology revolutions are rarely binary: client/server computing didn’t completely eliminate mainframes, neither did cloud with on-prem systems. Instead, they transformed and coexisted.

• If AI replaces the “business logic” layer of SaaS, agentic tool builders still face a steep challenge: mastering the nuanced use cases SaaS refined for years.

• To succeed, they need to establish trust and transparency with customers. Businesses won’t adopt tools they can’t control or verify, and adapting AI to highly specific workflows is nontrivial.

• SaaS vendors may embed AI deeply, creating hybrid solutions that merge AI agents’ intelligence with specialized SaaS strengths.

• Many incumbents’ technical roadmaps aim for this, but it demands a willingness to disrupt their own businesses from within.

Why it’s important: SaaS disruption could redefine enterprise tech, partnerships, and ultimately those who lead the market. Bain’s Whitten predicts convergence, not collapse, where AI and SaaS both complement and compete. The lesson from history: Transitions expand ecosystems rather than replace them outright.

Full report here.

GALILEO

AI agent evaluation best practices

Brief: Galileo, a company that specializes in AI evaluation, released a 93-page guide on mastering AI agents. It covers agent capabilities, real-world use cases, and frameworks, with a strong focus on performance evaluation.

Breakdown:

• Chapter 1 introduces AI agents, their ideal uses, and scenarios where they can be excessive. It includes real-world cases from Salesforce and Oracle Health.

• Chapter 2 details frameworks: LangGraph, Autogen, and CrewAI, providing selection criteria and case studies of companies using each.

• Chapter 3 explores how to evaluate an AI agent through a step-by-step example using a finance research agent.

• Chapter 4 covers measuring agent performance across systems, task completion, quality control, and tool interaction, with five detailed use cases.

• Chapter 5 addresses why many AI agents fail and provides practical solutions for successful AI deployment.

Why it’s important: As AI agents become more prevalent, ensuring they work correctly and safely is key. This is where evaluation comes in. Galileo’s previous guide focused on "Mastering RAG," building enterprise-grade systems. Now, they’ve taken it further with agents using LLMs to complete broader, more complex tasks.

Full report here.

ANDREESSEN HOROWITZ

AI agents will disrupt outsourced work 

Brief: Andreessen Horowitz (a16z), a leading venture firm, explores how AI will disrupt the Business Process Outsourcing (BPO) market ($300 billion in the U.S., 2024) by in-housing largely repetitive, mundane operations with AI innovations such as voice AI agents and deep research.

Breakdown:

• AI startups face incumbents, BPOs, that recognize the opportunity: Infosys deploying 100+ AI agents and Wipro reporting a 140% increase in AI adoption.

• BPOs have distribution advantages, but AI startups bring radical innovation. a16z believes AI startups have the edge in building AI-native products.

• BPOs often use time-and-materials pricing models, marking up labor by 20-30%. Their core business depends on people and selling their output.

• Transitioning to a product-first, AI-native business model is a monumental challenge, especially for public companies. a16z argues it will compress margins, eliminate cash cows, and disrupt culture.

• Leveraging cutting-edge AI is difficult with new releases coming constantly. Even experts struggle to stay updated, limiting BPOs’ adaptability.

• First-class, AI-native teams stay on top of the latest AI and know how to apply it to use cases. A rare mix often absent in BPOs, a16z asserts.

Why it’s important: AI startups have a limited window to disrupt the BPO space. As foundation models stabilize and become more comprehensible to a broader audience, BPOs will increasingly integrate AI, promote in-house AI products to long-time enterprise customers, and shift to outcome-based models.

Full report here.

ANTHROPIC

Building effective agents

Brief: Anthropic published an article sharing best practices from building effective agents with teams across industries, identifying seven common agentic system patterns in production and when to use them.

Breakdown:

• Anthropic distinguishes between two types of agentic systems: workflows, where LLMs and tools follow predefined code paths, and agents, where LLMs dynamically direct their own processes and tool usage, controlling task execution.

• Anthropic is seeing seven common agentic system patterns in production: Augmented LLM (building block), prompt chaining, routing, parallelization, orchestrator-workers, evaluator-optimizer (all workflows), and agents.

• For instance, in the orchestrator-workers workflow, a central LLM divides tasks, assigns them to worker LLMs, and combines their results, making it ideal for tasks like complex coding and data searches across multiple sources.

• Agents are emerging as LLMs advance in understanding complex inputs, reasoning, planning, tools, and error recovery. Ideal for open-ended problems where the steps are unpredictable and a fixed path can't be hardcoded.

• For full details on all seven patterns, including architecture diagrams, usage guidance, and implementation examples, see the summary or full article.

• The full articles also dives into customer support and coding agents, which have shown particular promise, as well as prompt engineering best practices and the use of frameworks like LangChain (see Anthropic’s cookbook).

Why it’s important: Anthropic's experience in building agents offers best practices to help enterprises leverage gen AI. These patterns can be adapted and combined for various use cases. While more agentic complexity can improve performance, it often increases latency and cost, so such tradeoffs should be considered.

Full report here.

CITI

Agentic AI in the future of finance

Brief: Citi’s 44-page report on Agentic AI, led by Ronit Ghose, Head of Future of Finance at Citi Global Insights, explores how gen AI and agentic AI could surpass the internet era’s impact on the economy and finance.

Breakdown:

• This shift means tasks such as those typically outsourced will increasingly be handled by agentic AI.

• Agentic AI is largely in an experimental phase. This report examines what is being built now with insights from 30+ AI transformation and tech experts.

• These insights cover agentic AI use cases in compliance, fraud prevention, onboarding, KYC, wealth management, treasury, and more.

• For example, in onboarding and KYC, the report addresses common challenges faced by banks and shows how agentic AI can solve them.

• Greg Ulrich, Chief AI & Data Officer at Mastercard, shares how the company uses gen AI-based assistants for onboarding, leveraging RAG and fine-tuning with Databricks, along with many other case studies.

Why it’s important: AI’s true value is realized when it’s implemented and adopted at scale across industry use cases. This report offers comprehensive insights and real-world case studies on how financial services companies are turning their AI ambitions into reality.

Full report here.

PwC

Powering automation with agents

Brief: PwC released a 20-page report on Agentic Powered Automation (APA), providing leaders and AI professionals insights into its benefits, enterprise use cases, executive perspectives, and guidance on aspects of governance.

Breakdown:

• APA combines gen AI agents, the brain, understanding requests and planning actions, with automation (e.g. APIs, RPA), the hands, executing actions.

• The agents validate outputs throughout, and learn from results to improve performance over time.

• The report highlights 20 agent use cases across finance, supply chain, sales, marketing, IT, and HR. For example, expense reconciliation in finance.

• A HR talent search deep dive compares the as-is process vs. agentic solution (image above), highlighting how roles shift from transactional to strategic.

• Data & Analytics Leader Sudipta Ghosh mentions the need to balance agentic innovation and compliance, among other PwC perspectives.

Why it’s important: The market is filled with terms: agentic AI, agentic-powered automation, or autonomous agents. Automating knowledge work isn’t new; but it can be faster and more expansive with gen AI, increasing agency for fuller, end-to-end automation and augmenting distinct human capabilities.

Full report here.

MENLO VENTURES

From RAG to autonomous agents

Brief: Menlo Ventures article AI Agents: A New Architecture for Enterprise Automation explores six gen AI examples from RAG to Autonomous Agents, detailing reference architectures and levels of autonomy. Architecture summary here.

Breakdown:

• The fully autonomous agents of tomorrow might possess all four building blocks of AI agents: reasoning, external memory, execution, planning. But today’s LLM apps and agents do not.

• The popular RAG architecture isn’t agentic but relies on reasoning and external memory. The key distinction is that these apps use the LLMs as a "tool" for search, synthesis, or generation, but their steps remain pre-determined by code.

• By contrast, agents emerge when the LLM controls the application's flow, dynamically deciding actions to take, tools to use, and how to interpret and respond to inputs. Menlo ventures outlines three types of agents.

• Decisioning Agent: At the most constrained end are “decisioning agent” designs, which use LLMs to traverse predefined decision trees.

• Agent on Rails: “Agents on rails” offer more freedom with a higher-level objective, but constrain the solution space with an standard operating procedure (SOP) and a predetermined library of tools to choose from.

• General AI Agents: At the far end of the spectrum are “general AI agents”, essentially for-loops with minimal data scaffolding, relying entirely on the LLMs reasoning for planning, reflection, and course correction.

Why it’s important: Gen AI is entering its agents era, marking a shift toward greater autonomy and sophistication in AI systems. To fully leverage their potential, it’s key to understand agentic designs and the nuances of varying definitions emerging across the market.

Full report here.

DELOITTE

Deloitte AI Institute on AI agents reshaping work

Brief: Deloitte AI Institute’s new 18-page report Prompting for Action explores how AI agents expand capabilities, use cases, and enterprise impact from gen AI. It also highlights key use cases for businesses today.

Breakdown:

• The report compares AI agents to typical language models, highlighting their differentiated ability to not only interact but also reason and act more effectively. It examines scope, planning, memory, tools, data, and accuracy.

• It discusses how multi-agent systems of role-specific AI agents coordinate and collaborate with humans to orchestrate complex work, highlighting their key benefits for enterprises, such as accuracy enabled by “validator” agents.

• A real-world example visually breaks down how multi-agent systems enhance speed, efficiency, and scalability compared to traditional research projects.

• The report also explores four key use cases for enterprises today: individualized financial advisory, dynamic pricing and personalized promotions, talent acquisition, and personalized customer support.

• It outlines implications for strategy, risk, talent, technology, and data, including emerging risks like “agent autonomy” and potential unintended consequences from minimal human oversight.

Why it’s important: Understanding what sets AI agents apart, how they enable new use cases, and their implications and risks is key to enhancing automation and accelerating enterprise transformation.

Full report here.

MICROSOFT

Agents are not enough

Brief: Microsoft research explores how the latest agents, differentiated by gen AI, can achieve greater success by collaborating with Sims and Assistants. This approach addresses challenges like generalization, scalability, and coordination.

Breakdown:

• Microsoft defines an agent, in the context of AI, as “an autonomous entity or program that takes preferences, instructions, or other forms of inputs from a user to accomplish specific tasks on their behalf.”

• Agents are nothing new, evolving from early agents (1950s) to expert systems (1980s), reactive agents (1990s), and more recently multi-agent systems and cognitive architectures.

• Frameworks like AutoGen help modern agents with complex tasks in narrow domains, but challenges persist. Research suggests an ecosystem of Agents, Sims, and Assistants to enhance standardization, personalization, and trust.

• Sims are representations of the user, built from their profile, preferences, and behaviors, capturing key aspects of who the user is. Sims can act on the user’s behalf, interacting with agents to accomplish tasks, guided by the user’s Assistant.

• Assistants are programs that deeply understand users and can proactively or reactively call on Sims or Agents to complete tasks. Assistants act as private agents, accessing personal information to perform tasks on the user's behalf.

Why it’s important: The synergy between Agents, Sims, and Assistants enables more precise, personalized interactions in completing complex tasks. This research aims to help advance standards for how AI systems work with each other and humans in the enterprise.

Full report here.

UNIVERSITY OF CHICAGO

Transforming industries with vertical AI agents

Brief: The University of Chicago’s 31-page paper authored by Fouad Bousetouane attempts to define a level of standardization for vertical AI agent design patterns and examines practical use cases across industry.

Breakdown:

• The paper explores how vertical AI agents are emerging as a critical innovation to address the limitations of traditional SaaS platforms.

• Agents are differentiated by their potential for advanced domain-specific intelligence and flexibility to adapt to evolving scenarios.

• Task-specific, multi-agent, and human-augmented agentic system patterns are explored, leveraging agent(s) to automate complex work.

• It details multi-agent systems' architecture patterns such as RAG Orchestrated Multi-Agent Systems (see image above).

• Practical use cases for multi-agent systems investigated include enterprise reporting, legal case analysis, financial portfolio management, and more.

Why it’s important: The rise of agentic systems with LLMs as the cognitive backbone signals a paradigm shift in enterprise software and automation. This paper is a valuable resource for understanding foundational concepts, architecture patterns, and practical use cases in industry.

Full report here.

AWS

Best practices for building gen AI agents

Brief: AWS released a two-part series (part 1, part 2, summary of both, code) on best practices for creating gen AI apps using Amazon Bedrock Agents. It covers aspects of architecture, evaluation, reusability, and more, with many aspects being platform agnostic.

Breakdown:

• Gather accurate ground truth data (including expected API usage, knowledge base access and guardrails) and clearly define the scope with sample interactions to align agent capabilities with specific business needs.

• Architect collaborative AI agents as small, focused units that interact with one another, maximizing modularity, scalability, maintainability, and ease of testing.

• Craft the user experience with clarity, focusing on clear instructions and integrate knowledge bases through indexed documents and citation configuration.

• Establish evaluation criteria such as response accuracy, task completion rate, and latency. Employ human evaluators and diverse perspectives for continuous refinement. Implement logging and observability.

• Use infrastructure as code for consistency and reusability, optimize models for both cost and performance, and implement robust testing frameworks, including test case generation with large language models (LLMs).

• Build in security measures such as flexible authorization, encryption, and broader responsible AI practices. Develop a reusable actions catalog and follow a 'crawl-walk-run' approach to scale agent usage gradually.

Why it’s important: AWS’s best practices provide valuable insights on developing and deploying efficient, scalable, and secure gen AI agentic applications for production. AWS’s repository includes detailed examples and use cases to help you get started quickly and effectively.

Full report: part 1, part 2.

SEQUOIA

The agentic reasoning era begins

Brief: Sequoia, a prominent Silicon Valley venture capital firm, discusses in Generative AI’s Act o1 the new era of agentic reasoning and how startups beating corporate IT and global systems integrators by focusing on GenAI at the application layer is "pretty doable".

Breakdown:

• Competition in the foundational layer (infrastructure and models) is stabilizing, led by Microsoft/OpenAI, AWS/Anthropic, Meta, and Google/DeepMind, due to high capital requirements.

• Gen AI is evolving from "thinking fast" (pre-trained responses) to "thinking slow" (reasoning at inference), unlocking the potential for disruptive "killer apps" in consumer and enterprise markets.

• The cloud transition brought software-as-a-service. The AI transition is about building applications that automate and amplify knowledge work, 'service-as-software', a much larger market worth trillions.

• However, transforming gen AI capabilities into reliable end-to-end enterprise agentic apps still requires significant expertise and engineering effort.

• Established enterprises must adapt or risk being outpaced by new businesses that master the intricacies of knowledge work and manage to reach customers.

• Similar to the emergence of 20 $1 billion+ companies in cloud and mobile applications, Sequoia expects a comparable opportunity with AI.

Why it’s important: There is vast potential for enterprise disruption and reinvention in a rapidly evolving market. Innovation at the application layer will play a key role in defining the next generation of industry leaders. The choice is clear: disrupt or be disrupted.

Full report here.

ADDITIONAL REPORTS

PWC - Agentic AI

McKinsey - Why agents are the next frontier of generative AI

Google - Agents

WEF - The rise of AI agents

OWASP - Agentic AI threats and mitigations

UiPath - Preparing for the agentic era

Hugging Face - Fully autonomous AI agents should not be developed

ADDITIONAL MUST-READS

• Enterprise AI case studies (20 reports)

• Playbooks for AI leaders (16 reports)

• Enterprise AI market (10 reports)

LEVEL UP WITH GENERATIVE AI ENTERPRISE

Generative AI is evolving rapidly in the enterprise, driving a new era of transformation through agentic applications.

Twice a week, we review hundreds of the latest insights on best practices, case studies, and innovations to bring you the top 1%...

Explore sample editions:

• McKinsey: Now leaders must step up.

• What board directors need to know.

• Goldman cuts time-to-value by 95%.

• What 2000+ client projects revealed.

All the best,

Found this valuable? Share with a colleague.
Received this email from someone else? Sign up here.
Let's connect on LinkedIn.