AI This Week: Tech Companies Are Cutting Staff and Blaming AI, While Hackers Are Already Using It Against You

Key Takeaways

• Monday.com is the latest in a list of over 20 tech companies to cite AI as a factor in significant layoffs, signalling a structural shift in how software firms are staffing.
• A reported hack on OpenAI's systems was allegedly carried out at "superhuman speed" by an AI with minimal human guidance, raising the bar on what a credible cyber threat looks like in 2026.
• UK SMEs using cloud-based AI tools need to treat vendor security posture as a procurement question, not an afterthought.
• The AI data centre grid problem in Northern Virginia is a preview of infrastructure instability that could affect UK cloud service reliability.
• Construction industry consolidation (post-ISG collapse) is accelerating, with firms like Wates absorbing government contracts at scale.

Are Tech Companies Actually Replacing People With AI, or Just Saying So?

Monday.com became the latest tech company this week to announce layoffs with AI cited as a contributing factor. TechCrunch has been tracking a running list and it now covers more than 20 significant tech firms. The pattern is consistent: companies reduce headcount in certain functions (support, content, QA, middle-management coordination) and point to AI tooling as the reason efficiency gains make those roles redundant.

The honest answer to the headline question is: both things are true simultaneously. Some of these layoffs are genuine structural responses to AI productivity gains. Others are using AI as a convenient narrative to justify cost-cutting that would have happened anyway. Separating the two matters, because the implication for UK SMEs is different depending on which dynamic is actually at play.

If AI genuinely replaces functional roles at scale inside software companies, the downstream effect is that the tools those companies sell get more capable and cheaper faster. That is good for a small construction firm or a manufacturing business trying to automate admin without hiring a developer. The productivity tools get better, and the price pressure on competitors who are not using them intensifies.

But the more important signal for UK business owners is what this says about where AI actually produces results quickly enough to justify headcount decisions. The roles going first are not strategic ones. They are high-volume, repeatable, process-bound tasks: responding to routine support queries, generating first-draft content, routing information between systems, and checking outputs against rules. That is exactly the same category of work that creates operational drag for a 10-person trades firm or a mid-sized manufacturer. Missed enquiry follow-ups, manual job scheduling updates, quote chasing, invoice reminders. None of it requires strategic thinking. All of it takes time.

What the tech layoff wave confirms is that AI handling these functions is no longer a pilot programme or an experiment. It is a staffing decision. For UK SMEs who cannot afford to hire additional admin resource, that is actually the most useful framing. You are not adopting an experimental technology. You are deploying the same category of tooling that large software companies are now betting their headcount on.

The practical takeaway is this: if your business still has a human manually chasing quotes, manually responding to the same five enquiry types, or manually moving job data between systems, you are absorbing a cost that AI handles reliably. The workflow and admin automation systems we build for trades and construction businesses do exactly this. The reason is not that it is interesting technology. It removes a fixed time cost from a business that cannot easily absorb it.

One caveat worth stating plainly: if your process is not documented, AI cannot automate it. The firms making redundancies on the back of AI have invested in mapping their workflows first. Before any automation is worth deploying, you need to know what the process actually is, step by step. That is not an obstacle unique to AI. It is just good operations. But it is the prerequisite that most SMEs skip.

Should UK Businesses Be Worried About the OpenAI Hack?

The short answer is yes, though probably not for the reason the headline implies.

BBC Technology reported this week on a breach linked to OpenAI's systems, with Hugging Face stating that the attack was conducted at "superhuman speed" and involved an AI operating with little or no human guidance. Whether this is a warning shot or a publicity event is genuinely debatable. But the mechanism described is not.

AI-assisted attacks are not new, but the speed and autonomy described here represents a meaningful escalation. Traditional cyber attacks rely on human operators making decisions at each stage: reconnaissance, credential testing, lateral movement, exfiltration. If that decision-making can be delegated to an AI agent running at machine speed, the window between initial breach and significant damage compresses dramatically. A security team that might catch a human-paced intrusion in hours may have only minutes against an AI-directed one.

For UK businesses using cloud-based AI tools (including OpenAI's own products via API or through platforms like Microsoft Copilot, Zapier, or Make), the practical implication is straightforward. Your exposure is not just your own systems. It is every vendor in your stack whose breach could expose your data or disrupt your access: a supply chain risk that most SMEs do not assess.

The specific questions worth asking of any AI vendor you use: Where is your data stored? Is it processed in the UK or EU? What is the vendor's breach notification policy and how quickly would you know? Does the platform hold any of your customer data, and under what terms? For businesses in construction or manufacturing handling client contracts, pricing data, or subcontractor details, these are not abstract compliance questions. They are operational ones. A breach that exposes a tender price or a client list has immediate commercial consequences.

GDPR still applies here in full. If a vendor breach results in personal data exposure, the obligation to notify the ICO within 72 hours sits with you as the data controller, not the vendor. Most SMEs are not prepared for that. Building a short data-flow document that maps which AI tools touch which categories of data takes an afternoon and is genuinely worth doing.

The broader point is this: AI is becoming a component of attack infrastructure at the same pace it is becoming a component of business operations. The answer is not to avoid AI tools. It is to treat security posture as a selection criterion when choosing them, not an afterthought once you have already integrated them into your workflow.

What Does the AI Data Centre Grid Problem Mean for UK Cloud Users?

TechCrunch reported this week on a close call in Northern Virginia, where a single fallen power line exposed serious vulnerabilities in how AI data centres respond to grid disruptions. Northern Virginia is not incidental geography. It hosts a disproportionate share of global cloud infrastructure, including data centres that underpin AWS, Microsoft Azure, and Google Cloud, all of which serve UK businesses directly.

The core problem is straightforward. AI workloads consume vastly more power than standard compute. A cluster of GPU servers running inference or training jobs draws electricity at a scale that traditional data centre power management was not designed to handle. When the grid hiccups, the failover systems designed for conventional loads are not always adequate for AI-dense facilities. The TechCrunch piece describes a situation where the response to a grid event was slower and less coordinated than it should have been, with cascading risk across multiple tenants.

For UK SMEs, this is not an immediate crisis. But it is a relevant trend. If your business relies on cloud-hosted AI tools, whether that is an OpenAI integration, a CRM with AI features, or a custom workflow automation built on platforms like Make or n8n: you are downstream of this infrastructure. An outage at a major data centre cluster does not just mean a slow website. It can mean your automated enquiry handling goes dark, your AI-assisted quoting system is unavailable, or your job management integrations fail mid-process.

The practical response is not to avoid cloud AI tools. The infrastructure risk is real but manageable with the right architecture. The key design principle is graceful degradation: building systems so that if an AI component fails, the process falls back to a human-manageable queue rather than simply stopping. In the enquiry handling systems we build for trades businesses, this matters practically. If an AI agent is unavailable, an enquiry should still land somewhere visible and actionable, not disappear into a failed webhook.

Redundancy across cloud providers also helps. A business running everything through a single cloud vendor has a single point of failure. Spreading critical functions across AWS and Azure, or using platforms with multi-region failover, reduces that exposure meaningfully. It adds modest complexity but significantly improves resilience for businesses where system downtime directly translates to missed revenue.

The infrastructure conversation is also relevant to the UK specifically because domestic data centre capacity is growing but still constrained. Government investment in AI infrastructure has been announced, but the building timeline is years, not months. Until domestic capacity matures, UK businesses will remain dependent on US-hosted infrastructure for the majority of AI compute. Understanding that dependency is the first step to managing it sensibly.

If you want to understand where your own operations are exposed to AI-related risk or where automation could close genuine gaps, the AI automation checklist is a practical starting point. It takes less than ten minutes and gives you a clear map of where your business is losing time or revenue to manual processes.