Edition in partnership with

Welcome executives and professionals. AI infrastructure is becoming one of the defining capital allocation challenges of the next generation.

Since the previous edition, we have reviewed hundreds of the latest insights in agentic and generative AI, spanning best practices, case studies, market dynamics, and innovations.

This briefing outlines what is driving material value — and why it’s important.

In today’s briefing:

• Global data centre outlook 2026-50.

• Agent security in the enterprise.

• OpenAI releases GPT-6 Astra.

• The incumbents are coming.

• Transformation and technology in the news.

• Insights for Executive+ members.

• Career opportunities & events.

Read time: 4 minutes.

MARKET INSIGHT

Global data centre outlook 2026-50

Brief: PwC projects US$31.6 trillion of capital expenditure through 2050 to build the compute capacity that AI demands. Modelled with Oxford Economics across 46 countries, it dwarfs the railways, electrification and the internet.

Breakdown:

• Global data centre capex could reach $31.6tn through 2050, with a plausible upside near $50tn if AI adoption accelerates further.

• Investment keeps rising to 2050. Servers, GPUs and other ICT equipment need replacing every four to six years, not just once.

• Power availability, sovereignty rules and chip trade flows will decide where capacity gets built, and where your workloads can run.

• AI inference stays close to users for latency, privacy and sovereignty reasons, while training chases cheap power, chips and talent.

Why it’s important: For enterprises consuming data centre compute, this signals a market shaped by scarcity rather than abundance. Where a workload runs will be set by sovereignty as much as by price, and the refresh cycle keeps pressure on what you pay.

IN PARTNERSHIP WITH THE HACKETT GROUP®

The Hackett Group® establishes enterprise AI World Class benchmarks

Brief: The Hackett Group’s AI World Class benchmarks quantify what AI-enabled performance can look like across the enterprise, showing advantages of up to 75% over Hackett industry peer groups for organizations redesigning work around AI.

The benchmarks are part of The Hackett Group’s proprietary intelligence layer, which combines performance data, process intelligence and best practices, and powers the company’s AI-enabled delivery platforms: XT™, AIXelerator™, XDA™, Hackett AI XPLR™ and ZBrain™.

Together, they help organizations identify the highest-value AI opportunities and determine what changes are needed to capture them:

• AI and Digital World Class® benchmarks quantify the performance potential

• Best-practice process intelligence identifies where and how work needs to be reimagined

• AI-enabled platforms turn transformation opportunities into measurable, breakthrough outcomes, leveraging existing, extended and agentic automation

The result: performance targets connected directly to transformation and execution, turning AI investment into measurable business value. See how it works.

BEST PRACTICE INSIGHT

Agent security in the enterprise

Brief: OpenAI published a 30-page practical guide to rolling out AI agents safely. Written for security leaders introducing agents across the enterprise, it covers assessing risk, setting expectations and a CISO action checklist.

Breakdown:

• Five principles run through it: know which agent is acting, keep it to the task, trust no input, enforce limits, plan the rollout.

• Section one covers agentic risk: what an agent can reach, how much harm it could cause, and which duties sit with your AI vendor.

• Section two sets the boundaries for execution, with eight ways agents fail or get attacked, from prompt injection to runaway execution.

• Section three covers day-to-day operations: what to log, and a plan to stop, contain, recover and learn when something goes wrong.

Why it's important: For executives, it turns agent risk into questions a board can ask. Who owns this deployment, what can it do without anyone else approving it, and can we stop it and prove what happened? It also sets an expansion gate to clear before widening agent autonomy.

INNOVATION INSIGHT

OpenAI releases GPT-6 Astra

Brief: OpenAI introduced the long-awaited GPT-6 Astra, calling it the "most intelligent and aligned model in the world," with benchmarks hitting new highs across areas like science, math, computer use, coding, and cybersecurity.

Breakdown:

• Overall, the model lands at 61 on the Artificial Analysis Intelligence Index, behind Fable 5.1, Fable 5, Opus 5, and Muse Spark 1.3.

• Astra is priced at $10/$50 per million tokens in the API, around 2.5x GPT-5.6 Sol, though it uses tokens more efficiently per task.

• It is rolling out over the coming days to all paid ChatGPT tiers including Enterprise, plus the OpenAI API, Azure, and AWS Bedrock.

• Pro, Business, and Enterprise users also get Astra Pro for multi-step work. Enterprise admins must enable it; it is off by default.

Why it’s important: Astra’s benchmarks are impressive, but the real test comes when more enterprises get access, and with Fable 5.1 here at the same price, the frontier now has a pretty clear head-to-head. With OpenAI’s share of the enterprise market growing faster than Anthropic’s, the race is on.

AI-NATIVE PROFESSIONAL

How to review budget vs. actuals with ChatGPT

Brief: In this guide, you'll learn how to give ChatGPT a budget, an actuals export and close notes, then have it map actuals to plan, calculate variances, flag reconciliation issues and separate explanations from open questions.

Step-by-step:

1. Attach the budget plan, the actuals export and your close notes, or give exact file references along with the source for each one.

2. Run the starter prompt, asking for an editable .xlsx workbook, then open the workbook it produces for you in ChatGPT.

3. Expand it into the full-screen view to inspect the raw inputs, the category mappings, variance formulas and summary tab.

4. Before you share it, use the follow-up prompt to have ChatGPT audit the categories, the formulas and the variance explanations too.

Best practice: If the source files are in a connected plugin, mention the exact files or folder instead of asking it to search a broad workspace.

For the full guide, including prompts, upgrade to Executive+ or The Boardroom.

MARKET INSIGHT

The incumbents are coming

Brief: Andreessen Horowitz set out the bullish case for incumbent systems of record: AI makes them more important, not less, because the data and actions they control become the foundation that agents depend on to do work.

Breakdown:

• As agents do more work through these systems, the data and actions incumbents control, and charge for, become even more valuable.

• An incumbent system of record paired with a general agent, or one the incumbent builds, may then be enough to get the work done.

• Salesforce and Anthropic just announced this: Claudeforce lets teams use the Salesforce CRM from inside Claude, without opening it.

• Incumbents are pushing their own agents too: Docusign's Iris reviews contracts, Atlassian's Rovo resolves requests (image above).

Why it’s important: There are jobs where a system of record plus a general-purpose agent will be enough, but there's still real opportunity for vertical AI-native companies: the strongest cases are where work happens often, judgment matters, and the learning loop is strong.

Deloitte rethought the CISO role for an AI-saturated enterprise, and explored how the agentic era risks eroding human judgment at work.

OpenAI published a 24-page defense factory playbook on running a continuous cyber defense operation and starting the surge today.

KPMG published a 19-page, five-move guide for CDAOs on making enterprise data searchable, contextual and trusted for AI agents.

Booz Allen released its Cyber Weapon Index, testing 18 US and Chinese LLMs as autonomous attackers on a production-grade network.

Deloitte launched an Open Model Engineering practice to help clients build and scale agentic AI using a mix of open and proprietary models.

Sequoia argued the cognitive revolution mirrors the industrial one, with machines set to do 99.9% of cognitive work as prices fall.

Meta shipped Muse Spark 1.3, with Max scoring a 62 on Artificial Analysis' Intelligence Index, behind only Fable 5.1 and Opus 5.

Google released Gemini 3.8 Flash, holding 3.7's $0.75/$3.75 pricing while scoring 59 on the Intelligence Index with agentic gains.

Nvidia announced a $12.9B acquisition of Hugging Face, with Jensen Huang pledging the platform stays open to every cloud and chip.

The US and China are reportedly planning mid-September AI safety talks, their first bilateral session devoted solely to AI.

Sen. Bernie Sanders published a one-page summary of a bill with Rep. Greg Casar to outlaw superintelligent AI, with harsh penalties.

The U.S. government filed a brief backing OpenAI in the NYT copyright case, arguing narrower fair use would cost the US its AI lead.

CAREER OPPORTUNITIES

Microsoft - AI Transformation Senior Director

Anthropic - Head of Partner Ecosystem BD

Accenture - AI-Native Strategy Director

EVENTS

Salesforce - Agentic AI Contact Center - October 6, 2026

Microsoft Ignite - November 17-20, 2026

AWS re:Invent - November 30 - December 4, 2026

Reach enterprise AI decision-makers:

• 66% of readers are C-level executives or VP and Director-level leaders.

• 63.2% of the audience is based in the U.S., EU, UK, ANZ, and Singapore.

• Read by leaders at Microsoft, Deloitte, the Fortune 500, and more.

Guaranteed impression and custom sponsorship packages available, with post-send performance reporting.

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

We welcome your feedback.