So, I came across an article earlier in the week from Computing, with the rather bold title of “Large firms sour on AI adoption”, and this got me to thinking, is this really the case and is it really that unexpected.
Over the past few years, AI has been the at the forefront of most organisations digital transformation strategies. From boardroom pitches to keynote stages, the narrative has been clear: AI is the future, and the future is now. However, according to the US Census Bureau, the percentage of large firms (250+ employees) using AI dropped from 14% to 12% in the past year. Now that may seem, and probably is a small shift, and to be perfectly honest I was surprised (perhaps I shouldn’t have been) at the 14% figure, as I thought it would be larger. Anyway, the computing and other outlets (tech.co, Forbes etc.) then highlight several reasons why this “drop” may have occurred, including:
- Overhyped expectations: Many firms expected AI to deliver immediate cost savings or productivity gains. When that didn’t materialise, enthusiasm waned.
- Workforce disruption: AI’s impact on jobs—especially among younger workers—has created internal resistance.
- Lack of measurable ROI: Despite heavy investment, many organisations are struggling to quantify the value AI brings.
What I found interesting about the above is that data security/complexity is not mentioned, which is the main concern being expressed to me, both internally, and from clients (backed up by Microsoft’s focus on security for their new fiscal).
What I am not surprised about from this list is the ROI statement, this was also highlighted in a recent MIT study (https://mlq.ai/media/quarterly_decks/v0.1_State_of_AI_in_Business_2025_Report.pdf), where it stated that only 95% of AI projects are delivering measurable increase in ROI. Anecdotally, at least, ROI is incredibly difficult to measure for Generative AI projects. There are several reasons for this, but the ones most visible to me, is that most tools now seem to focus on individual productivity, and to date I have not come across an organisation that measures worker productivity in a way that easily translates to calculating Gen AI based ROI, another difficulty is that apps tend to span multiple department, business units and applications, making ROI calculations more complex, and finally that benefits related to Gen AI tools will probably have a much longer time to realisation value, i.e. the impact is not immediate, but will be felt over time.
As a result, I have seen incredible success stories: AI systems that boost developer productivity, intelligent automation that transforms customer service, and predictive analytics that reshape supply chains. But I have seen the flip side, pilot projects that never scale, data pipelines that struggle to support real-time inference, and teams that struggle to align on what “AI success” even looks like.
There are a few common attributes that succesfgul AI projects all seem to have exhibited:
- Shift from pilots to platforms: Stop treating AI as a side project. Embed it into your core architecture, with clear ownership and KPIs.
- Invest in data readiness: AI is only as good as the data it learns from. That means better data governance, lineage, and quality.
- Focus on trust and transparency: Explainability, fairness, and accountability aren’t just ethical concerns—they’re business imperatives.
- Build for humans, not just machines: The best AI augments human capability. It doesn’t replace it. Design with empathy and usability in mind.
Final Thoughts
AI isn’t failing. It’s maturing. And like any maturing technology, it’s entering a phase where the hype fades and the hard work begins.
For large enterprises, this is a moment to pause, reflect, and retool. The next wave of AI success won’t come from chasing trends—it’ll come from solving real problems, with real data, in real workflows.
And that’s a challenge worth rising to.
Microsoft Cloud Updates: What’s New
As always, I’ve rounded up the most relevant Microsoft Cloud updates across Azure, Microsoft 365, Intune, and Windows 365 to keep you informed and ahead of the curve.
Azure Cost Management: Partner Access Gets Smarter
Azure now supports assigning the Partner Admin Reader role to service principals, enabling CSPs to securely and programmatically access cost data across customer tenants. This is a big win for partners managing multiple EA customers, allowing for real-time billing insights and automated cost optimisation.
https://azure.microsoft.com/en-us/blog/microsoft-cost-management-updates-july-august-2025/
Microsoft Intune: Enterprise-Ready App Control & Smarter Device Setup
Intune’s August 2025 update introduces:
- Granular App Control: Assign policies to specific groups, not just tenant-wide, enabling Zero Trust application control with precision.
- Autopilot Patching: Devices now auto-install critical updates during setup, so users receive fully patched machines on day one.
Microsoft Intune: Network Endpoint Changes Coming
On 10 September 2025, Microsoft announced upcoming changes to Intune network service endpoints:
- Starting 2 December 2025, Intune will use Azure Front Door IP addresses
- Organisations must update firewall rules to include the new AzureFrontDoor.MicrosoftSecurity service tag
- Failure to update may result in login issues, device disconnection, and app access disruptions
This change is part of the Secure Future Initiative and aims to simplify firewall management and improve security alignment across Microsoft services.
.NET & .NET Framework: September 2025 Servicing Updates
Microsoft released new servicing updates for .NET 8.0, .NET 9.0, and .NET Framework on 5 September 2025. These updates include:
- Non-security fixes for runtime, SDK, and ASP.NET Core
- Updated installers, binaries, and container images
- Known issues and changelogs for each version
This release is part of Microsoft’s regular monthly cadence and is aimed at improving stability and performance across supported platforms.
Cloudy with a Chance of Insights | EP19
In last week’s episode of Cloudy with a Chance of Insights, we explore Microsoft’s evolving Zero Trust strategy, including updates to its assessment tool and workshop, and dive into August’s Intune security enhancements like Spark app control and Apple DDM support. We also spotlight Logic Apps running on Rancher K3s for edge deployments, new Azure Service Groups for tenant-level governance, and Excel’s integration with Fabric OneLake for secure data access. The episode wraps with insights into the US Government’s OneGov cloud deal and developer productivity boosts in VS Code and Azure DevOps.
YouTube: https://youtu.be/yRdTHYC2oHY
Spotify: https://open.spotify.com/episode/3tUfcPBZ9hM5thuJccocWS?si=709d2bd838f745cb
Apple: https://podcasts.apple.com/us/podcast/sovereign-cloud-zero-trust-real-world-microsoft-cloud/id1783369178?i=1000725477232
