Table of Contents

• Strategic Deep Dive: The 14-Month Production Timel …

• This Week in AI

• Future Signals & Tooling

Hey there,

Welcome to the first edition of the AI for Business Leaders newsletter! We're here to deliver actionable AI insights that help you stay ahead of the curve and execute successfully in this transformative era of technology.

This week we're tackling the elephant in the room: why so many AI pilots never make it to production. We've been analysing what separates the companies that successfully scale their AI investments from those stuck in endless testing phases.

The Short Version: 92% of companies that successfully scale their AI pilots see positive ROI within 12 months (1). Here's what surprised us: according to Forbes only 10% of pliots actually make it into production, with the other 90% not moving beyond their pilot phase anytime soon (2). The difference between success and spinning your wheels? A structured approach to the 14 month pilot-to-production journey. We've been digging deep into what actually works, and we're excited to share what we've learned.

Strategic Deep Dive: The 14-Month Production Timeline

Getting to Production: Unlocking the Benefits

Here's what we're seeing: 35% of companies get trapped in what we call "pilot purgatory."(3) You know the feeling, your proof of concept works beautifully, but somehow moving it to real operations feels impossible. With 71% of companies now using AI in some form (4), but so few actually getting to production scale within six months, you're definitely not alone in this struggle.

We've been tracking successful companies, and there's a pattern. The best performers follow a 14-month timeline from initial pilot to meaningful ROI. This isn't just arbitrary. It gives you enough time for proper testing (usually 3-6 months for complex stuff) while keeping momentum going. And here's the kicker: if you're running a smaller, more agile company, this timeline can actually become your secret weapon against slower-moving competitors.

Before we dive into the how, Let's talk numbers on why this is important:

The Money Side. That $3.50 return for every dollar invested in AI? (5) It only happens when you actually get to production. Companies stuck in pilot mode are seeing significants less business impact and lower growth, whilst successful adopters are seeing outsized returns (6).

Competitive Reality. The companies getting this right are crushing it. For the AI achievers we are talking a whopping 50% greater revenue growth on average, compared with their peers! (7). And get this: 75% of business leaders say AI is key to their competitive advantage (8).

From the Frontline. Those using AI day to day in their business processes are overwhelming in their support for the benefits. 84% of salespeople says it helps increase sales by speeding up customer interactions and 90% of service professionals report it helps them server customers faster (9).

The Real Barriers to Production

Before diving into the framework, let’s review the main challenges. Getting AI to production fails on three main fronts: organisational dysfunction, operational gaps and trust barriers. Most companies focus only on the tech while ignoring the human and process elements that actually determine success.

Organisational and Leadership

• Lack of Dedicated Teams: AI systems require dedicated operational teams, not just borrowed developers or project-based resources.

• Leadership Misalignment & Unclear Ownership: AI projects often sit at the intersection of IT, data science, and business units, creating confusion over who drives deployment, budgets, and ongoing maintenance. Clear executive sponsorship and cross-functional governance are essential.

• Change Management & Human Factors: Employee resistance, fear of job displacement, and lack of trust in AI outputs can undermine adoption if not proactively addressed.

Process and Operations

• Data Readiness & Quality: Production AI must handle messy, incomplete, and constantly changing real-world data. Implementing schema validation, incremental data ingestion, and continuous quality monitoring is critical to avoid costly failures.

• MLOps & Infrastructure Gaps: Unlike traditional IT projects, AI requires specialised operational frameworks, often called MLOps or LLMOps which automate deployment, monitoring, retraining, and rollback.

• Continuous Improvement: AI models require ongoing retraining and monitoring; neglecting this leads to rapid performance decay.

Risk, Compliance and Trust

• Security, Privacy, & Compliance: Production AI must comply with regulations like the EU AI Act. Lacking audit trails, explainability, or privacy safeguards can result in fines and reputational harm.

• Building Trust: Both employees and customers need confidence in AI systems, which requires transparency, education, and clear communication.

The Pilot to Production Roadmap

We’ve distilled the successes of industry leaders into 5 actionable core pillars: Leadership & Governance, People & Change Management, Process Integration, Data, Technology & Compliance, and Continuous Improvement. Each pillar is paired with proven methodologies to ensure your AI projects are not only innovative but scalable, secure, and aligned with business goals.

Case Study: Unilever's Supply Chain Transformation

Unilever's ice cream forecasting pilot failed when scaling because production data lacked pilot-level cleanliness. Their fix demonstrates the "crawl, walk, run" approach:

Months 1-2: Targeted forecasting pilot in high-variability regions

Months 3-5: Built real-time data pipelines integrating sales, weather, and freezer sensor data

Months 6-7: Implemented weekly retraining cycles with anomaly detection triggers

Months 8-10: Trained retail partners and supply chain planners on new dashboards

Months 11-14: Global rollout with continuous improvement processes

Key Insight: Started with challenging markets first (proving value quickly), then scaled to stable regions. Implemented automated data quality checks and hired dedicated "data product managers." This result in forecast errors dropped from 22% to 9% within 9 months.

Executive Summary

The journey from AI pilot to production success isn't just about technology, it's about orchestrating people, processes, and governance alongside your technical implementation.

Each of the five pillars we've outlined represents what actually works in the field, not just in theory.

Let us know which of these areas resonates most with your current challenges, we're happy to deep dive into any of these pillars in future newsletters, complete with additional case studies, implementation templates, and actionable frameworks tailored to your specific industry context.

This Week in AI

This week the landscape was shaped by headline-grabbing partnerships, product breakthroughs, and renewed scrutiny of AI’s societal impact. Amazon’s landmark licensing deal with The New York Times underscored a new era of collaboration between tech and media, while Anthropic, Microsoft, and Google each unveiled next-generation AI capabilities. Meanwhile, the White House faced criticism over AI-generated errors in a major health report, and leading voices like Gary Marcus reignited debate about the limitations of large language models. As the pace of innovation accelerates, so too does the conversation about trust, transparency, and the future role of AI in business and society.

Enterprise AI Breakthroughs

Anthropic launches voice mode for Claude:
Anthropic rolled out a beta voice mode for Claude mobile apps this week, enabling users to have fully spoken conversations with the AI assistant.

Microsoft Build 2025 delivers autonomous AI agents:
Microsoft announced major updates at Build 2025, transforming GitHub Copilot into an autonomous coding agent that can handle GitHub issues, generate pull requests, and revise code independently.

Google I/O 2025 focuses on AI-first search:
Google introduced significant AI capabilities at I/O 2025, including AI Mode for Search with deep research features and live camera integration. The company also announced Gemini 2.5 Pro's new Deep Think mode, which can pause and evaluate multiple possibilities in parallel, similar to strategic game-playing systems.

Amazon licenses New York Times content for AI training:
Amazon and The New York Times announced a major licensing deal on May 29, 2025, allowing Amazon to use NYT content for training Alexa and other AI systems. This is a landmark agreement in the ongoing debate over AI and copyright, signalling a shift toward formal partnerships between media and tech giants.

Policy & Governance

White House MAHA Report Scrutinised for AI-Generated Errors
Experts flagged inaccuracies in the White House’s MAHA Report on declining U.S. life expectancy, with 37/522 citations linking to nonexistent studies. The Washington Post traced these errors to unreviewed AI-generated content.

Ethics & Sustainability

Gary Marcus Critiques LLM Limitations
The AI skeptic doubled down on criticisms at Web Summit Vancouver, calling LLMs “inherently broken” and citing ChatGPT’s 29% error rate in legal research.

Security Alert

Critical GitHub AI Agent Vulnerability Exposed:
Invariant Labs has disclosed a major flaw in GitHub’s MCP integration, which powers many AI coding agents. Attackers can plant malicious GitHub Issues to trick connected AI agents into leaking sensitive data from private repos.

Future Signals & Tooling

In this section we dive into the trends driving the future of Generative AI and the tools that are leading the way.

Executive AI Advisors

AI is rapidly evolving from back-office automation to becoming a strategic partner in the C-suite(10). A new class of tools, "executive AI advisors" are now capable of synthesising vast datasets, surfacing actionable insights, and even recommending next steps for business leaders.

Tool Spotlight

Fyxer AI has achieved remarkable growth in the executive AI space, going from launch to $9M ARR in under 12 months while maintaining 60% month-over-month growth (11).

The platform integrates directly into Gmail and Outlook, automatically drafting email replies in the user's tone, organising inboxes with smart categorisation, and providing AI-powered meeting notes 1.

Glean has emerged as a leader in AI-powered enterprise search and knowledge management, reaching $100M ARR in 2024 (triple the previous year) and achieving a $4.6B valuation (12). The platform was named #6 on Fast Company's World's Most Innovative Companies of 2025 and #1 in the Applied AI category.

Granola has captured attention in the executive productivity space, raising $43M at a $250M valuation while growing 10% weekly since launch (13). The AI-powered note taking app goes beyond meeting transcription, with users increasingly utilising it for personal note-taking and comprehensive information management.

Executive Takeaway

As AI matures, it’s quietly moving from the back office to the boardroom, with a new generation of executive AI advisors now offering business leaders real-time insights and practical recommendations.

Tools like Fyxer, Glean, and Granola are setting the pace, streamlining everything from inbox management to enterprise knowledge search and note-taking, all while posting impressive growth numbers.

These platforms represent a practical step toward streamlining routine tasks and surfacing relevant information, helping executives stay organised and informed.

That’s all for this week. Let us know your biggest AI deployment challenge via the poll below or replying to this email. We love talking about this stuff.

Luke & Marco

P.S. - Next week we're diving into the Billion Dollar AI Arms Race and who’s already lost. Tech giants are planning to pour over $300 billion into AI this year, but as the race heats up, some surprising players have already figured out how to pull ahead without breaking the bank. We break down who’s winning, who’s burning cash, and what it means for your next move.