Edition sponsored by

WELCOME, EXECUTIVES AND PROFESSIONALS.

AI agents are transforming customer interactions and driving operational efficiency with tailored, near-human generative AI.

Since the previous edition, we’ve reviewed hundreds of the latest agentic and generative AI best practices, case studies, market dynamics and innovation insights. Here’s the top 1%...

Note: Effective August 17, 2025, the name of this briefing (appearing above/next to the subject line in your inbox) will change to Enterprise AI Executive. The content and sender email address will remain unchanged.

In today’s briefing:

• Inside McKinsey’s AI agent: Ollie.

• Anthropic’s trustworthy agents framework.

• State of AI Technology Q2 2025.

• OpenAI releases GPT-5.

• Transformation and technology in the news.

• Career opportunities & events.

Read time: 4 minutes.

CASE STUDY

Inside McKinsey’s AI agent: Ollie

Brief: McKinsey unveiled Ollie, a FrontlineAI smart agent that uses pre-built conversational agents within its agentic framework to automate customer interactions with hyper-personalization, handing off to humans when needed.

Breakdown: .

• Ollie is multilingual, multi-channel, scalable, always available, always learning, with persistent memory and connected system updates.

• It runs analytics on customers and services, delivering hyper-personalized interactions that adapt to each user’s needs in real time across channels.

• McKinsey’s insights introduce Ollie “in her own words,” showing her first day, handling 5-10x the calls at once compared to human colleagues.

• A European utility scaled Ollie to 3M customers in three months, focusing first on simple queries handled by interactive voice response (IVR).

• Results: 30% of calls automated end-to-end, 10% faster handling by humans, and a 6 percentage-point CSAT uplift compared to legacy IVR.

Why it’s important: Ollie demonstrates that AI agents can be deployed and scaled rapidly, delivering swift results. Customers preferred Ollie over the prior system; with satisfaction increased, and handling times decreased, showing how AI enhances both efficiency and customer experience.

PRESENTED BY STACKAI

The most powerful way to bring AI to your enterprise

Brief: Global financial institutions, defense leaders, and Fortune 500s trust StackAI to run mission-critical workflows without compromising security. From operations and finance to legal and beyond, StackAI delivers the platform to deploy AI agents at scale with full control.

With StackAI, you benefit from:

• No-code logic builders to easily design multi-step AI agent processes

• Role-based access, on-prem/VPC deployment, and built-in PII protection

• 100+ native integrations (e.g. SAP, Salesforce) and leading LLMs

• Enterprise-grade compliance with SOC 2 Type II, GDPR, and HIPAA

Deploy secure, scalable AI agents today with StackAI

BEST PRACTICE INSIGHT

Anthropic shares trustworthy agents framework

Brief: Anthropic released an early framework for responsible agent development to help shape emerging standards. It emphasizes principles that balance autonomy with safety, and align agents with human values.

Breakdown:

• Agent design should balance autonomy with oversight. Independence drives value, but humans retain control over how goals are pursued.

• Transparency is key. Agents should explain reasoning, allowing humans to verify facts/sources, and steer them to better decisions.

• Agents don't always act as humans intend. Ensuring alignment is complex, making transparency and human control vital until solutions mature.

• Agents retain data raising privacy risks. Tools and processes that agents utilize should haves strong controls, like those in MCP.

• Design should safeguard sensitive data and prevent misuse, including where malicious prompts trick agents into unintended actions.

Why it’s important: AI agents hold tremendous potential to transform work. But this potential is only realized if agents are built with safety and reliability in mind. A strong framework ensures alignment with human values, safeguards privacy, and prevents harmful or unintended consequences.

MARKET INSIGHT

State of AI Technology Q2 2025

Brief: Artificial Analysis, a benchmarking and insights firm, published its 37-slide State of AI Q2 2025 report, offering data to help guide investment, platform strategy, and policy in an increasingly AI-native global landscape.

Breakdown:

• Competition is intensifying in agent domains like coding, research, and computer use, with multiple players vying for leadership.

• GitHub Copilot and Cursor lead AI coding tool use, staying ahead of Claude Code and Google Gemini Code Assist in developer rankings.

• US labs (OpenAI, Google, Anthropic, xAI) lead proprietary reasoning benchmarks, while China leads open weight models (e.g. via Alibaba)

• The analysis predates OpenAI’s Aug 7 GPT-5 frontier release, and Aug 5 gpt-oss open release (competitive with China’s open models).

• Across the broader AI stack, vertical integration; Google leads from TPU accelerators to Gemini, maintaining the most complete AI stack.

Why it’s important: Models continue to advance in intelligence while becoming faster and more cost-efficient. Agentic workflows are shifting from experiments to production use, with coding agents proliferating across development teams, amid intense competition between the US and China.

INNOVATION INSIGHT

OpenAI releases GPT-5

Brief: OpenAI launched GPT-5, delivering modest performance gains over frontier models but important upgrades for simpler enterprise integration, improved cost control, and more predictable, scalable deployment.

Breakdown:

• GPT-5 is a unified system with a real-time router that selects between efficient and reasoning models based on task complexity and intent.

• GPT-5 leads marginally on benchmarks, e.g., scoring 74.9% on SWE-Bench Verified vs. Opus 4.1’s 74.5%, as Anthropic grows AI coding share.

• The performance leap is modest compared to past OpenAI releases, but core product improvements improve the enterprise experience.

• Native routing and multimodality reduce the need for complex pipelines, while new controls lower moderation overhead and error rates.

• Automatic model selection optimizes compute without constant human intervention, enabling a more predictable cost-performance balance.

Why it’s important: GPT-5 offers modest performance gains but also delivers core upgrades that improve reliability and simplify deployment. Beyond enterprise, it brings frontier intelligence to 700 million users, including free users, for the first time. Explore OpenAI’s GPT-5 prompting guide and optimizer.

Capgemini released an 84-slide CMO playbook on how gen AI is reshaping marketing and how humans, humanoids, and agentic AI are set to unite.

Stanford published a 14-page playbook for building enterprise AI, covering use case selection, models, architecture, risk, and more.

Salesforce detailed how executives can leverage AI to hone their strategic vision and signal to employees that AI use isn’t optional, but expected.

Bain outlined how customers are using AI search, and explored how soon agentic AI will redefine Enterprise Resource Planning (ERP).

BCG shared insights on redesigning work to empower employees, accelerate AI adoption, and announced its new AI Talent Promise initiative.

McKinsey explored how agentic AI fights financial crime in banking and unveiled its new generative AI tool that helps transform client insights.

MIT released a taxonomy of 831 AI risk mitigations, covering governance, security, and other domains to support organizational risk management.

Reuters reports how investors increasingly claim that AI hype is securities fraud, citing firms overstating AI capabilities to boost share prices.

OpenAI released open-weight models, announced ChatGPT Enterprise at $1 for U.S. federal agencies, and removed ChatGPT’s opt-in for search indexing.

Anthropic released Claude Opus 4.1. They also revoked OpenAI’s API access for term violations and heavy Claude Code use before GPT-5’s release.

Microsoft incorporated GPT-5 into products like Microsoft 365 Copilot and Azure AI Foundry, bringing frontier intelligence to millions of enterprise users.

Google released Gemini 2.5 Deep Think and exited its asynchronous coding agent, Jules, from beta, while also introducing new tiered plans.

Cloudflare revealed Perplexity hides its AI web crawlers’ identity from sites that block scraping to avoid detection.

xAI plans to open-source its Grok 2 AI model next week, following OpenAI’s first open model releases since GPT-2 in 2019.

Apple CEO Tim Cook told analysts the company is “open to M&A” that speeds up its AI roadmap and helps it catch up with competitors.

Rillet, an AI accounting startup, raised $70M in a round led by Andreessen Horowitz and ICONIQ in a bid to disrupt incumbents like Oracle and Microsoft.

CAREER OPPORTUNITIES

The Washington Post - AI Director

Capgemini - AI Leader

Stability AI - Head of AI Solutions

EVENTS

ALIGN AI - Executive Summit - August 20, 2025

Anthropic - Claude Code in FS - August 21, 2025

Thomson Reuters - Gen AI Forum - September 25, 2025

Originally conceived as a practical communication for executives the editor, Lewis Walker, has worked with, this briefing now serves as a trusted resource for thousands of senior decision-makers shaping the future of enterprise AI.

If your AI product or service adds value to this audience, contact us for information on a limited number of sponsorship opportunities.

We also welcome feedback as we continue to refine the briefing.