AI This Week: Notion Just Killed Its Email App Because AI Agents Made It Redundant

Key Takeaways

• Notion killed its Skiff-influenced email app because most users shifted to AI agents to manage their inboxes instead.
• UK construction sites are facing real operational risk from extreme heat, with energy supply, productivity, and safety all under pressure.
• AI agents replacing software products entirely is happening now, in tools businesses already use.
• The shift from "AI feature inside a tool" to "agent that replaces the tool" is accelerating faster than most SME owners realise.
• UK trades and construction businesses that have not audited their admin workflows are now measurably behind.

Why Did Notion Kill Its Email App?

Notion has shut down its email client because the vast majority of its users stopped using it in favour of AI agents managing their inboxes directly. This is a significant product decision. Notion's own language is "going all in on using agents to run your inbox."

That is a significant public statement from a major productivity software company. They are pivoting because user behaviour changed underneath them, not because the email app was technically flawed. They are pivoting because user behaviour changed underneath them. People found that handing their inbox to an agent, something that reads, categorises, drafts replies, and surfaces what actually needs attention, was more useful than a cleaner interface for doing those things manually.

For UK SME owners, the relevance here is direct. Think about how much time your business spends processing email. Enquiries that need a quote, suppliers chasing confirmation, subcontractors asking about job details, clients wanting an update on timelines. For a 10-person electrical contractor or a busy roofing firm, the inbox is often the single biggest source of unstructured, unmanaged workload. Someone has to read everything, decide what matters, respond to the urgent stuff, and file or ignore the rest. That process, repeated dozens of times a day, is exactly what agents are now replacing for Notion's user base.

The practical takeaway is this: if a well-funded software company looked at its own email product and concluded agents had made it redundant, your business processes deserve the same honest audit. The question is not whether AI can help with email. It is whether the way you handle enquiries today would survive comparison with a properly configured agent doing it instead. In the systems we build for trades and construction businesses, email enquiry handling is often where the most immediate time savings appear, because it is the most consistent, repeatable part of the day that nobody has ever bothered to systematise.

It is also worth noting what Notion's decision signals about the direction of software development broadly. Products built around manual user interaction are under pressure. Agents do not need a beautiful UI. They need good instructions and access to the right data. That reframes what "good software" even means for a business, and it is a shift UK SMEs are largely unprepared for.

What Does the UK Heatwave Mean for Construction Site Operations?

The UK recorded its highest ever June temperature this week, 36.1°C, and Western Europe is in the grip of a dangerous heat event. For construction sites, that is not background noise. It is an operational problem with immediate financial consequences.

Aggreko, the energy supply specialist, issued a direct warning to construction managers this week: review your site energy supply now. Temporary generators and distribution systems are not all rated for sustained high-temperature operation. Cooling loads increase sharply when temperatures climb, meaning power draw from site cabins, welfare units, and sensitive equipment can spike beyond what the installed supply was sized to handle. If a generator fails on a site running tight against a programme, the cost is not just the repair. It is standing time for a full crew, potential penalties for delays, and the knock-on effect across the schedule.

The Considerate Constructors Scheme raised a separate but related concern: heat affects workers differently depending on role, physical condition, and acclimatisation. A one-size approach to heat management, the same water breaks, the same shift patterns for everyone, does not reflect how risk actually distributes across a site workforce.

The construction industry's exposure to weather is not new, but the intensity and frequency of extreme heat events is changing the calculus. Sites that previously treated heatwave planning as a once-a-decade edge case now need it built into standard operating procedure.

The practical implication for construction businesses is that operational resilience now needs a digital layer. Technology does not fix the heat, but the administrative burden of managing a weather-disrupted site compounds quickly. Rescheduling subcontractors, notifying clients, logging welfare checks, updating programmes, reordering material deliveries that no longer align with the revised timeline. All of that lands on a site manager or a director who is already dealing with the physical disruption. In the systems we build for construction businesses, automated scheduling and communication workflows are often what prevent a one-day weather delay from turning into a week of administrative chaos.

If your business operates on sites and you do not have a clear, repeatable process for communicating disruption to clients and subcontractors automatically, this week is a reasonable prompt to fix that.

Is the White House Quantum Deadline Relevant to UK SMEs?

The White House issued an order this week dramatically shortening the deadline for US federal agencies to move away from quantum-vulnerable cryptography, citing national security risks if post-quantum cryptography is not adopted in time. At first glance, that sounds like a concern for governments and defence contractors, not a roofing firm in Maidstone or a manufacturer in the Midlands.

But the supply chain implication deserves a moment of attention. If you supply to large enterprises, US-connected businesses, or government-adjacent contracts, those clients will increasingly require their suppliers to meet minimum security standards. That pressure cascades down. GDPR already obligates UK businesses to protect personal data using appropriate technical measures, and "appropriate" is a word that moves over time as threats evolve. Post-quantum cryptography is not yet a compliance requirement for UK SMEs, but the direction of travel is clear, and the timeline is compressing faster than most people expected.

The more immediate relevance for smaller businesses is simpler: the underlying principle of the quantum cryptography story is that security standards businesses assumed were adequate are now being reclassified as inadequate. The same logic applies to data handling processes that SMEs built years ago and have not reviewed since. If your business stores customer data, sends quotes by email, or keeps financial records in systems that have not had a security review, the quantum story is a useful prod to do that audit. Not because your quotes are a national security concern, but because the expectation of what "secure enough" means is rising, and ignoring it until a client or regulator asks is the expensive way to find out.

For trades and construction businesses specifically, Companies House records, CIS scheme data, and any information held on subcontractors or employees all carry obligations. GDPR compliance is not a one-time checkbox. It requires ongoing attention to how data moves through your systems, and the tools you use to handle it.

This is one of those stories where the direct operational takeaway is not "act immediately" but "add it to the audit." If you have not looked at your data handling setup in the last 18 months, this week gave you two good reasons to: a US federal deadline that will shape enterprise supplier requirements, and the broader trend of security standards tightening faster than small businesses typically expect.

Why Are AI and Tariffs Driving Up the Price of Hardware?

Tech firms including Microsoft, Nintendo, and Valve have publicly cited a combination of AI infrastructure investment and tariff pressure as reasons for significant price increases on consumer and enterprise hardware. Xbox consoles, the Nintendo Switch 2, and Valve's Steam Deck have all seen notable price rises in recent months.

For UK businesses, the consumer gaming angle is not the point. The point is what is happening to the underlying economics of hardware procurement more broadly. Data centre buildout for AI model training and inference is consuming an enormous share of global chip production. That creates upward pressure on component costs across the board, and it does not stay contained to consumer electronics. Servers, networking equipment, and business laptops are all drawing from the same constrained supply chains.

For an SME considering new hardware investment, whether that is upgrading site office equipment, adding compute capacity for design or estimating work, or refreshing an aging server setup, the practical message is that waiting for prices to normalise is probably not a sound strategy this year. The investment in AI infrastructure from major cloud providers is not slowing down. If anything, the competition between Microsoft Azure, AWS, and Google Cloud to build capacity for AI workloads is accelerating, and that competition is not reducing component costs at the SME level.

There is a secondary implication that is more specific to businesses thinking about AI adoption. Much of what UK SMEs actually need to implement AI automation does not require owning hardware at all. The workflow automation systems we build for trades and manufacturing businesses run on cloud infrastructure, meaning the hardware cost problem is largely abstracted away. You are not buying servers. You are buying outcomes: faster quoting, fewer missed enquiries, less manual admin. The hardware price story is a good reminder that trying to build AI capability in-house, on owned infrastructure, is a path that makes sense for very few businesses at the SME level. Renting compute from established cloud providers and building logic on top of it is almost always the right approach.

The broader economic context here is worth holding onto. AI is expensive to build at the frontier. That cost does not disappear. It gets absorbed into hardware prices, cloud subscription costs, and the pricing of any software product that has AI baked into it. Businesses that treat AI as a free or near-free addition to their existing toolkit will find that assumption tested over the next 12 to 24 months. The businesses that get ahead are the ones that tie AI investment to specific, measurable operational problems, not broad experimentation. That is, frankly, the only way to justify the cost when hardware and cloud pricing are both moving upward.

If you want to work out where AI automation would actually move the needle in your business, the AI automation checklist is a reasonable place to start.