Generative AI Enterprise
Posts
Winning in 2025: Next-level AI

Winning in 2025: Next-level AI

Plus, Anthropic AI Agents, Deloitte framework, and more.

Lewis Walker
December 23, 2024

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:

Capitalise on IBM’s trends to lead in 2025.
Anthropic on building effective AI agents in enterprises.
Next-level AI poised to impact enterprises.
Deloitte India framework for driving GenAI value.
Transformation and technology in the news.
Career opportunities & events.

Read time: 4 minutes.

MARKET INSIGHT

IBM unveils enterprise trends for 2025

Image source: IBM

Brief: IBM's 26-slide report outlines five key trends for 2025, offering strategies for attaining competitive advantage. The insights are backed by data from 400 leaders across 17 industries and six regions.

Breakdown:

Trend 1: Agentic AI will transform your business-but first you must reskill your people. In 2024, CEOs estimated 35% of their workforce needed reskilling—up from 5% in typical years.
Executives predict a 21% rise in decision-making by digital assistants within two years, driven by GenAI, which will require new skills in oversight of autonomous decision-making.
Trend 2: Despite efforts to curb technical debt, only 25% strongly believe their IT infrastructure can scale AI enterprise-wide. To address this, leaders can leverage strategies like linking modernization efforts to long-term productivity gains.
Trend 3: In the age of AI, location is everything. In 2024, 67% of executives reported AI influenced their location strategies, and expect that to increase to 93% by 2026, prioritizing comparative advantages like talent and infrastructure.
Trend 4: The rapid pivot to AI has upended IT budgets, but self-funding is imminent. 75% of leaders view GenAI as an innovation investment, reallocating traditional IT funds. By 2026, 95% expect GenAI to be at least partially self-funded.
Trend 5: AI product and service innovation is the #1 CEO goal, yet business models aren't keeping up. 62% of CEOs believe they must rewrite their playbooks for future success; 89% of executives say AI will drive product and service innovation.

Why it’s important: IBM’s findings highlight key strategies for enterprises to tackle workforce readiness, technical infrastructure, operational agility, and business model innovation to win in an AI-driven future.

BEST PRACTICE INSIGHT

Anthropic on building effective AI agents in enterprises

Image source: Anthropic

Brief: Anthropic published an article sharing best practices from building effective agents with teams across industries, identifying 7 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. They're 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 for example).

Why it’s important: Anthropic's experience in building agents offers best practices to help enterprises leverage GenAI. 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.

INNOVATION INSIGHT

Next-level AI poised to impact enterprises

Image source: ARC

Brief: OpenAI introduced o3, an AI model that significantly outperforms all others, signaling the potential arrival of more intelligent AI in the enterprise. This could unlock greater automation, augmentation, and an increased need for enterprise reinvention to remain competitive.

Breakdown:

o3 significantly outperforms all prior models on the ARC-AGI benchmark, which measures an AI's ability to learn and generalize from very few examples, akin to human intelligence.
o3 achieved 87.5% on ARC-AGI in high compute mode, tripling o1's performance. For perspective, human performance is 85%, and it took four years for ARC-AGI to move from 0% with GPT-3 in 2020 to 5% with GPT-4o in 2024.
Despite some general skepticism around benchmarks, o3 was tested against some of the toughest and most reliable benchmarks available, which are unlikely to have been influenced by training data leaks.
On GPQA, o3 scored 87%, outperforming PhDs who scored 34% outside their specialty and 81% within it. o3 is currently slow and costly, but much of the work in enterprises isn't PhD-level and can be automated or augmented by faster, more affordable models.
o3 opens opportunity for more complex tasks where paying high costs for accurate answers won’t be prohibitive if o3 proves reliable. Besides, its use cases will expand as costs decrease driven by extensive resources backing AI.
While o3 is as capable as the 175th best competitive coder in the world on specific tasks, it's not yet universally proficient across all areas.
Currently in preview, o3 is available only to safety and security researchers. OpenAI skipped o2 due to copyright concerns.

Why it’s important: For enterprises, this leap in intelligence signals that now is the time to reinvent, if not already underway, to handle increased automation and business model disruption, ensuring competitiveness in the coming years.

BEST PRACTICE INSIGHT

Deloitte publishes framework for driving GenAI value

Image source: Deloitte

Brief: Deloitte India's 42-page guide outlines strategies for leaders to drive GenAI adoption, with a focus on Global Capability Centres (GCCs), though the fundamental insights apply broadly across business areas.

Breakdown:

The report opens with a quote from NVIDIA CEO Jensen Huang, who states that Generative AI will be bigger than the PC, mobile, and internet.
The readiness framework (pages 8-12) scores parameters across two dimensions: ecosystem enablers (strategic factors like alignment with organizational goals and leadership buy-in) and capabilities (such as infrastructure, data, talent, and governance needed to deliver GenAI solutions).
The prioritization approach (pages 13-21) identifies opportunities across process taxonomies at level 3, assesses feasibility (data, tech, etc.), and prioritizes investments (high-impact, low-hanging fruit) based on benefits and effort. A Finance example illustrates the approach.
The implementation approach spans three key phases: building a proof of concept, solution deployment, and value capture (see page 25 for an overview). It also outlines key metrics and strategies for scaling GenAI solutions successfully.
The report includes production case studies and touches on AI agents and multi-agent systems (pages 32-34)

Why it’s important: Considering it's publicly available, this guide provides a relatively comprehensive framework for enterprises to start assessing GenAI readiness, prioritizing opportunities, and more, along with real-world examples.

World Economic Forum published a 28-page white paper on AI agents' evolution and impact. Capgemini and NTT Data also released reports on AI agents, while AWS shared a blog on multi-agent systems leveraging open-source.

IBM released a 9-minute video for leaders budgeting for GenAI, while AWS’s 35-minute video offers a six-step framework for leaders to unlock value. Gartner’s 21-minute video explores when not to use GenAI.

Accenture secured a record $1.2 billion in Gen AI bookings for Q4 2024, bringing their total to $4.2 billion since September 2023. Deloitte introduced Agentforce and Zora AI specialized agents.

Google highlighted five ways AI will shape business by 2025. Bain published a case study on online education with AI. A 6-page paper attempts to simplify the GenAI landscape, and a 27-slide report examines privacy-first AI governance.

Booz Allen’s AI workshop executive summary explores AI promises, perils, trends, legal considerations, and more. SoftBank pledged $100B investment in U.S. AI to create 100,000 jobs in the next four years.

OpenAI launched its new o3 model, API access to the advanced o1 reasoning model, real-time API upgrades, a new preference fine-tuning method, expanded ChatGPT Search access, and more.

Google launched Gemini 2.0 Flash Thinking Experimental, for complex problem-solving, akin to OpenAI’s o1 model but faster and free. Introduced its video generation model Veo 2, and launched Gemini Code Assist.

Microsoft AI announced the rollout of Copilot Vision, its AI companion that interacts with your browser in real-time, for U.S. Copilot Pro users on Windows. Databricks secured $10B in funding, bringing its valuation to $62B.

Anthropic’s co-founders released a video on the company’s past, present, and future. Perplexity AI secured $500M, tripling its valuation to $9B, and acquired Carbon to enhance its AI search platform with connected apps.

Lockheed Martin launched Astris AI for defense and commercial AI. Ukraine collected 2 million hours (228 years) of battlefield video to train AI. A 21-page report explores how AI can transform the military.