Generative AI Enterprise
Posts
Microsoft: Agents are not enough

Microsoft: Agents are not enough

Plus, enterprise threats, value creation deep dives, and more.

Lewis Walker
January 06, 2025

WELCOME, EXECUTIVES AND PROFESSIONALS.

Since the last edition, we've reviewed hundreds of the latest insights on best practices, case studies, and innovation. Here’s the top 1%...

In today’s edition:

Microsoft research: Agents are not enough.
BCG’s reveals AI value creation in leading enterprises.
Booz Allen on enterprise AI threats and countermeasures.
AWS's cost optimization strategies for GenAI applications.
Transformation and technology in the news.
Career opportunities & events.

Read time: 4 minutes.

BEST PRACTICE INSIGHT

Microsoft research: Agents are not enough

Image source: Microsoft Research

Brief: Microsoft research explores how the latest agents, differentiated by GenAI, 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. The research suggests an ecosystem of Agents, Sims, and Assistants to enhance standardization, privacy, 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 could help advance standards in how AI systems work together in the enterprise.

CASE STUDIES

BCG reveals AI value creation in leading enterprises

Image source: Boston Consulting Group

Brief: A 27-slide publication from BCG offers 42 examples and nine deep drives on how leading enterprises (those scaling AI, per BCG's Build for the Future 2024 Global Study) are generating value with AI, including GenAI.

Breakdown:

The publication showcases 42 examples with measurable outcomes across nine functions: sales, customer service, pricing & revenue management, marketing, manufacturing, field forces, R&D, technology, and business operations.
For instance, a biopharma company used AI in R&D to accelerate drug discovery, achieving a 25% cycle time reduction, $25M in cost savings, and $50M–$150M in revenue uplift.
Nine detailed case studies showcase AI (including GenAI) transformation and impact across functions, such as a GenAI Co-pilot for relationship managers at a universal bank (sales).
Additional deep dives highlight AI for BPO call agents (customer service), GenAI for data governance at a payments provider (technology), AI transforming credit processing at a European bank (business operations), and more.
It explores how AI leaders (26% of enterprises successfully scaling AI value) reshape functions rather than merely deploying or inventing, along with other success factors.

Why it’s important: This publication articulates the transformative value of AI across a breadth of functions, offering deep dive examples and measurable outcomes, all presented in a clear, easily digestible format.

BEST PRACTICE INSIGHT

Booz Allen on how enterprises can counter AI threats

Image source: Booz Allen

Brief: Booz Allen’s 16-page paper explores the key risks, threats, and countermeasures necessary to ensure enterprise resilience with AI. It highlights how AI security differs from traditional cybersecurity and introduces strategies for protection.

Breakdown:

AI security is critical due to factors like the "black box" nature of AI, hidden risks in third-party and open-source models, and vulnerabilities amplified by AI's distributed usage.
The paper identifies key attack types across the AI lifecycle, including Data Poisoning (manipulation of training data to compromise models), and Malware (embedding malicious code in model files).
Further threats include Model Evasion (perturbing inputs to control outputs), Data Leakage (theft of sensitive training data, IP, or model behavior) and LLM Misuse (overriding large language models to bypass safety and alignment).
The paper introduces a five-step strategy to AI security spanning Planning (risk modelling), Measurement (red teaming etc.), Security Engineering (model scanning etc.), Operations (monitoring etc.) and Control (governance).
It also provides a representative MoSCoW method to identify security requirements across a range of GenAI model deployments from third-party to “homegrown” models.

Why it’s important: AI security is non-negotiable. With the increasing exposure to AI, enterprises must adopt proactive security strategies to mitigate risks, protect sensitive data, and ensure the integrity of their AI systems.

BEST PRACTICE INSIGHT

AWS shares how to optimize gen AI application costs

Image source: Amazon Web Services

Brief: An AWS blog outlines cost optimization strategies for deploying GenAI applications in the context of a RAG virtual assistant example, offering largely vendor-agnostic insights to help capture ROI from GenAI.

Breakdown:

Cost optimization levers outlined include model selection, model choice, model customization, token count, token limits, token caching, and pricing models (on-demand, provisioned throughput).
Further optimization levers focus on security guardrail costs, vector database costs, and relative costs of chunking strategies such as fixed size chunking, hierarchical chunking, or semantic chunking.
It explores volume usage scenarios and respective costs for an organization seeking to provide customers with a RAG virtual assistant that can answer their questions with a high degree of accuracy, consistency, and safety.
It also breaks down various factors that can influence costs with cost optimization tips for input/output tokens, vector embeddings, database costs, guardrails and more.

Why it’s important: Cost optimization strategies are key for capturing ROI from GenAI applications. Understanding and applying these levers effectively can materially reduce operational costs while maintaining performance and scalability in AI deployments.

Everest’s Peak Matrix recognized Accenture, Capgemini, Cognizant, Deloitte, IBM, and TCS as top GenAI service providers. ZoomInfo reported AI job listings hit record highs in 2024, with C-suite positions increasing 428% since 2022.

McKinsey explored software-defined hardware in the age of AI and GenAI applications in automotive software development. Deloitte and Nvidia introduced "Factory as a Service," a fully managed and optimized infrastructure.

Google launched an accelerator program for GenAI social impact solutions. AWS announced new GenAI Competency Partners and updated resources in its Generative AI Center of Excellence.

IBM highlighted essential skills for 2025 and its advancements in optics technology for data centers. KPMG shared transformative GenAI use cases in the tax function. PwC introduced "prompting parties."

Perplexity AI founder Aravind Srinivas envisions a future in digital advertising where AI agents, rather than humans, could become the primary targets for ads. Microsoft released its new future of work report.

Microsoft shared its vision for U.S. leadership in AI and $80B AI data center investment plan for 2025, released AIOpsLab for autonomous cloud AI agents, and Hugging Face launched Smolagents for low-code AI agents.

OpenAI’s Sam Altman shared blog reflecting on the past two years. The company also published its plans to become a public benefit corporation, and defined AGI contractually with Microsoft as generating $100B in annual profits.

DeepSeek unveiled DeepSeek-V3, a cost-efficient language model rivaling top players. ByteDance launched 1.58-bit FLUX, reducing image generation costs. Alibaba cut Qwen-VL prices by 85% to drive adoption.

Nvidia acquired Run:ai for $700M and will open-source its hardware optimization software. Meta removed its AI character profiles. Samsung’s 2025 Bespoke refrigerators will feature ‘AI Vision Inside’ tech to order groceries.

Carnegie Mellon's benchmark testing shows AI agents can autonomously complete 24% of real-world workplace tasks in software environments. Defense contractors prepare for acquisition spree in AI, drone, and space technologies.