Generative AI Enterprise
Posts
Accenture: What 2000+ client projects revealed

Accenture: What 2000+ client projects revealed

Plus, McKinsey state of AI, the once in-a-generation opportunity, and more.

Lewis Walker
March 16, 2025

WELCOME, EXECUTIVES AND PROFESSIONALS.

How can organizations make reinvention truly real in the age of gen AI? Accenture reflects on lessons learned from the past year of project delivery and provides insights for leaders working to implement and scale this epochal technology.

Since the previous 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:

Making reinvention real with gen AI.
McKinsey: The state of AI.
OpenAI reveals AI governance proposal.
The once-in-a-generation opportunity.
Transformation and technology in the news.
Career opportunities & events.

Read time: 4 minutes.

MARKET INSIGHT & CASE STUDIES

Making reinvention real with gen AI

Image source: Accenture

Brief: Accenture’s 19-page reinvention report, informed by 2,000+ gen AI projects and 3,000+ C-level survey responses, explores key areas for organizations to prioritize as they seek to harness gen AI and create material impact.

Breakdown:

Advances in small language models and agentic architectures help explain why 83% of execs now see gen AI’s potential as exceeding initial expectations.
Firms addressing five imperatives are 2.5x more likely to be in the high-impact group of those creating significant enterprise-level value.
The imperatives: leading with value, reinventing talent/ways of working, building a secure AI digital core, responsible AI, and continuous reinvention.
Yet many still struggle, only 36% of surveyed execs say they have scaled gen AI solutions, and just 13% report achieving significant enterprise-level value.
A case study highlights how a bank drove $200M+ in annual efficiency, with over 80% of employees reporting increased confidence in using gen AI tools.

Why it’s important: As this report makes clear, most organizations are still at the starting blocks with gen AI. Moving from isolated experimentation to value at scale requires reinventing business processes across silos, focusing on talent and operating models, and putting people at the center of the transformation.

MARKET & BEST PRACTICE INSIGHT

McKinsey: The state of AI

Image source: McKinsey & Company

Brief: McKinsey published its 2025 state of AI report, surveying 1,491 survey participants across 101 nations, exploring AI adoption, how organizations are capturing value, and perspectives from McKinsey leaders.

Breakdown:

Fifty-three percent of surveyed executives say they are regularly using gen AI at work, compared with 44 percent of midlevel managers.
The survey highlights AI deployment structures: 57% centralize risk and compliance, while tech talent and AI adoption often follow a hybrid model.
27% of orgs using gen AI review all AI-generated outputs, while a similar share says 20% or less is checked before use.
Of gen AI best practices, tracking well-defined KPIs drives the most impact. At larger orgs, a clear adoption roadmap is also among the most effective.
AI hiring remains tough, though fewer respondents than in past years say filling roles is “difficult” or “very difficult.”
Inaccuracy, cybersecurity, and IP infringement are the top gen-AI risks causing negative impacts, according to survey responses.

Why it’s important: Gen AI use continues to surge, but it’s still early days. While enterprises are increasingly realizing value, more than 80 percent of respondents say their organizations aren’t currently seeing a tangible impact on enterprise earnings from their use of gen AI.

AI GOVERNANCE

OpenAI reveals AI governance proposal

Image source: OpenAI

Brief: OpenAI shared its policy proposal with the White House for the US AI Action Plan. The submission covers national security, infrastructure, energy, and the freedom to innovate, building on the January Economic Blueprint.

OpenAI warns that the 781 state-level AI bills introduced this year could stifle U.S. innovation against China's AI ambitions.
The proposal calls for additional infrastructure investment, copyright reform, and expanded access to government datasets for AI training.
It argues that if fair use copyright laws are not applied for AI training, China’s unrestricted data access could make the AI race “effectively over.”
OpenAI also called for a ban on models like DeepSeek, labeling them as security risks from a “state-controlled” lab.
Proposed infrastructure policies aim to catalyze U.S. reindustrialization, including modernizing the nation's energy grid.

Why it’s important: Policy proposals can help enterprises anticipate future regulatory and competitive dynamics. While perspectives on state law exemptions, open vs. closed source, copyright law, lab control, and other key issues vary, OpenAI's proposals are likely to carry weight due to their influence.

MARKET INSIGHT

The once-in-a-generation opportunity

Image source: Deloitte

Brief: Deloitte released a 20-slide report on seizing the "once-in-a-generation" opportunity in data centers and AI infrastructure. While aspects focus on Southeast Asia, the fundamentals apply globally to nations, enterprises, and investors.

Breakdown:

Gen AI applications rely on foundational models that demand vast amounts of power and compute for both training and inference.
By 2028, AI inference is expected to drive 85% of global GPU demand, while AI training will account for the remaining 15%.
Consequently, global data center electricity consumption is projected to nearly double from 536 TWh in 2025 to 1065 TWh in 2030.
Governments, enterprises, and investors are urged to treat AI infrastructure as critical assets and act fast to build them on their shores.
For investors, data centers offer stable cash flows, generating $1.5-2M per MW yearly, while GPU cloud models can yield $15M per MW.

Why it’s important: The risk of underinvesting in AI infrastructure far outweighs overinvestment, and continues to grow by the day. AI is a once-in-a-generation opportunity, with data centers as critical enablers. Sovereignty, security, and sustained prosperity require control over AI infrastructure.

Infosys published its 30-page AI Business Value Radar 2025, revealing that 19% of AI use cases achieve all business objectives. The report explores change management, leading industries, top use cases, and strategies to improve success.

Bain examines lessons from early adopters of wearable AI and how Ray-Ban Meta’s success signals gen AI’s arrival in wearables. Its research shows that 10%–20% of US adults are likely to adopt AI wearables by the end of 2025.

Deloitte explores how organizations equip C-suite leaders for gen AI adoption, from technical to transformation upskilling. One global firm enrolled 85% of senior leaders in core AI and applied AI leadership courses in 2024.

Booz Allen shared an 18-page paper on DeepSeek’s impact on the AI market. It includes an assessment of DeepSeek-R1’s architecture, exploring training optimization and explains how these innovations can be applied elsewhere.

AWS published a case study with Cognizant on Agent AI Ally, which supports call center agents with real-time info, sentiment analysis, post-call analytics, and more. It also shared an article on creating asynchronous AI agents.

Google launched Gemma 3, offering performance that rivals larger models on a single GPU/TPU. It also unveiled new image-generation features for Gemini 2.0 Flash and that it owns 14% of Anthropic with $3B in investments.

Microsoft is reportedly developing MAI, a new AI model family that rivals OpenAI and Anthropic in performance. If it also delivers on cost, its availability via Azure AI and Copilot could shift the foundation model power balance in the enterprise market.

Manus partnered with Alibaba’s Qwen to develop a Chinese version of its autonomous agent platform, following its viral success. Meanwhile, IBM assessed what Manus amounts to and MIT also put it to the test.

Cohere launched Command A, an AI model that rivals top enterprise competitors in performance while running on just two GPUs. It delivers high-speed, multilingual support and is integrated into Cohere’s North platform.

Harvey unveiled new AI agents with planning and adaptation capabilities that match or exceed lawyer performance on key legal tasks. These will be generally available in Harvey at the end of March.