Feed aggregator

One small phone attachment can spot cams, leaks, insulation problems and more. I was amazed when I tried it at my house.

AI conversations have moved past the point of curiosity.

Boards and leadership teams are no longer asking what AI might eventually do.

They are asking where it is actually working, what measurable value it is creating - and why so many promising experiments still fail to become durable operating advantages.

Across industries, companies have invested heavily in AI pilots, proofs of concept and impressive demos.

Yet many remain stuck in what I think of as pilot purgatory: the place where a tool works in a controlled environment but never survives contact with the complexity, exceptions and accountability required in production.

The problem usually isn’t the model

In my experience, AI initiatives rarely fail because the underlying technology is not powerful enough. They fail because of how the technology is applied. A model can be impressive in a sandbox and still be irrelevant to the business if it is not embedded into a real workflow, connected to the right data, governed appropriately and measured against outcomes that matter.

That is why access to AI is no longer a differentiator. Anyone can buy access to models or integrate a third-party tool. The real advantage lies in the things that can’t be bought off the shelf: proprietary data, deep domain expertise, and the discipline to continuously improve AI once it is operating at scale.

For us, those principles come together in our Lean AI approach, rooted in a Lean operating model that drives continuous improvement through testing, learning, and acting. Instead of chasing technology for technology’s sake, our Lean AI approach helps us move AI beyond experimentation and into production, where it can improve service, boost productivity and create real business value.

Production AI requires discipline, not experimentation for its own sake

This is where many organizations get stuck. They treat AI as a portfolio of experiments instead of an operating capability. Organizations that successfully operationalize AI tend to do the opposite.

They prioritize AI opportunities based on business value and points of operational friction, identifying manual, repetitive and high-volume work. Then, they build and deploy agents where automation can improve speed, accuracy, scalability or service quality across the entire customer workflow.

There is no hobby AI in this model. Every deployment needs a clear business case, a workflow owner, measurement, feedback loops, and a plan to scale. That discipline is especially important with agentic AI, because agents operate with more autonomy than traditional software.

Progress is not always linear. Systems improve, encounter new edge cases, retrench and improve again. Human oversight isn’t a temporary bridge either; it is part of the architecture.

The application layer is where the moat gets built

The AI ecosystem is often described in layers, from the underlying IT infrastructure and large language models to the applications built on top of them. Those foundational layers are essential, but they are not where most enterprises will build durable, competitive moats.

The real advantage comes at the application layer—where AI is integrated into workflows, systems, exceptions, data and human judgment that define how a business actually runs. If an enterprise does not own or deeply control that layer, it risks turning AI into another generic capability rather than a competitive advantage.

Take supply chain logistics as one example. Moving a single shipment isn’t a linear task. It may require coordinating moves by truck and ship and rail, customs documentation in multiple countries, handoffs at multiple facilities, and weather and market conditions that change by the hour. A generic AI tool does not understand that workflow out of the box.

Context is the hard part

Every industry has its own data and context that powers it. In supply chains, that context lives in historical pricing patterns, warehouse operations, customer-specific policies, shipment characteristics, driver performance, market cycles and the judgment of people who have solved messy freight problems for years.

That context cannot simply be purchased. It has to be collected, structured, governed and applied. AI becomes more effective when it’s built into your technology platform and can learn from those realities rather than relying on generic information alone. Just as important, employees add institutional knowledge through continuous feedback, teaching AI agents the same way they would train a new operations employee.

Take something as seemingly straightforward as scheduling a truck to pick up freight. On the surface, it sounds like a narrow task. In practice, it requires understanding customer requirements, freight characteristics, facility policies, loading dock constraints, appointment systems and exceptions that may vary by location. An AI agent can only automate that work reliably if it has been engineered with the right context and oversight.

The companies that get the most from AI will be the ones that move beyond pilots and treat it as an operating model. That means starting with real business problems, owning the application layer where differentiation happens, feeding agents with proprietary context, keeping humans in the loop and measuring outcomes relentlessly.

AI will not reward the companies with the most demos. It will reward those that can operationalize learning faster than their competitors.

Check out out list of the best cloud backup services.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

AI spend made headlines again recently with the Claude Fable 5 model from Anthropic. Before security concerns led to the model being suspended, there were also cost concerns. Anthropic says Fable costs $10 or approximately €9 per million input tokens and $50 per million output tokens. This is double the price of the company’s previously most expensive model, Claude Opus 4.8.

Posts soon began to pop up on LinkedIn, showing just how quickly teams were going through their tokens and, as a result, their budget. There are some caveats here. Namely, that Fable 5 is an advanced model and, for most businesses, won’t need to run non-stop or be used for every task.

But therein lies a key issue: AI use is accelerating and models are evolving. But the level of control and visibility businesses have over how much is being spent, by who and for what is lagging behind.

How AI consumption became a finance problem

There is a massive shift within the UK software market toward AI and specifically Anthropic’s ecosystem. Proprietary data from Pleo looking at the top tech merchants based on number of spending customers, shows that Anthropic (Claude) surged from 12th place in Q4 2025 to 7th in Q1 2026. Meanwhile, the average spend per customer increased +43.0% in this time.

This rapid climb signals that Anthropic has reached enterprise maturity in the UK market with businesses moving beyond the experimentation phase. But while this reflects growing confidence in AI adoption, it also presents some financial challenges.

On the whole, AI has redefined how the workplace runs, but it is not a free trial. The cost of tokens has gone up, and new models that can achieve what was seemingly unthinkable a few years ago come with a price tag to match. The new challenge for business leaders is to leverage these technologies but also limit rampant spending.

This is why many organizations are turning to their finance teams. Finance has the visibility to dig into the details and map AI use across the organization, whether it quietly shows up as a subscription renewal or a new budget request. But more than that, they can be instrumental in ensuring teams embrace open conversations, not just OpenAI.

AI activity does not translate to AI value

Just about every organization will have developed transformational ways of using AI tools. But, whether they know it or not, there will be wasteful ones too.

When it comes to inefficient use, some of the major culprits include asking AI agents open-ended questions, model mismatch where tokens are burned unnecessarily; and duplicate tools, resulting from shadow AI and overlapping subscriptions. These prevent businesses from seeing the full picture; one that is, in all probability, very expensive.

User literacy can improve this. But for finance teams they must start with the grey area of AI consumption. Two teams might show as active AI users, but one that’s using an LLM to produce more content faster is doing something fundamentally different to one that’s using it for peripheral productivity tasks. In fact, only 29% of European SMEs using Gen AI are doing so in core business activities.

To improve the control they have over AI, organizations must start by elevating their visibility from who is using AI, to who is using it to become smarter, faster and more productive.

How to regain control over AI use

A complete view of AI spend is essential, regardless of whether costs are rising.

Breaking spend down by department, team and budget helps identify both disproportionate usage and areas where adoption may be lagging. These should be combined with performance metrics such as the time-to-first-draft on marketing content; code review cycle times in engineering; support ticket resolution time in customer support; and so on.

This combination of spend and performance can reveal whether AI investment is translating into measurable productivity gains and not just higher software costs.

Visibility should also extend to model-level usage. As mentioned before, the cost difference between frontier reasoning models and lighter alternatives can be tenfold. Monitoring model and vendor usage alongside token consumption helps organizations route routine tasks to lower-cost options, maximize ROI and reduce unnecessary spend.

Finance teams should therefore expand reporting and budgeting frameworks to include AI-specific metrics. A key question at month-end is whether AI-enabled teams are increasing output and capacity without increasing headcount. This provides a clear headline for AI's impact, can justify investment and distinguish between high-value and low-value AI usage.

Ultimately, effective control over AI is not about costs alone. It is about understanding where AI is creating value and ensuring investment is aligned with business outcomes.

AI control is at your fingertips

The good news is that none of these metrics require a sophisticated AI analytics stack. Finance teams should already have the tools for real-time visibility into what’s being spent and where. All that’s needed now is to fold AI into the mix and collaborate with other departments to measure and improve its ROI.

The outcome is that organizations control AI use through oversight, without restricting spend, adoption or innovation through lengthy procurement processes. Spend policies, category controls and clear approval thresholds control what is spent, and everything is measured.

But crucially, teams don’t slow down as a result. The only difference is that AI is optimized for impact.

We've featured the best AI chatbot for business.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Most travel eSIMs ask you to pay before you even know if they work in your destination. Firsty flips that. You get a real, working data connection for free, supported by short ads, before spending a single euro. For a product category that's been moving toward higher prices and longer commitments, that's a genuinely different pitch.

Firsty is a Dutch eSIM startup founded in September 2023, built around one idea: give travelers free connectivity first and let the product sell itself. With a 4.1/5 on Trustpilot and over a million users, it's working. But the free tier has limits, the paid speeds are modest, and the company is still young.

Want more options? Take a look at our best eSims for international travel.

Firsty: Pricing and plans

Plan

Price

Speed

Data

Ads

Hotspot

Firsty Free

€0

256 Kbps - 1 Mbps

Unlimited (ad-gated)

Yes

No

Comfort+

From €1/day

Up to 5 Mbps (tiered)

Unlimited (tiered speed)

No

Yes

First Class

From €2/day

Extra high (up to 5G)

5GB full speed, then lower

No

Yes

Firsty Free

(Image credit: Firsty )

Free is exactly what it says. You pay nothing, watch a short ad every 59 minutes, and get data in return. The speed sits between 256 Kbps and 1 Mbps, enough for WhatsApp, Google Maps, email, and Uber. Not enough for Instagram Reels or a Zoom call, and Firsty isn't pretending otherwise.

On the downside, the free plan isn't available globally. It's currently limited to users in North America, Europe, and APAC. If you're traveling to regions outside those three, you'll need a paid plan from day one.

The free plan also has a referral perk. Invite a friend and you get 7 days of ad-free browsing added to your account. It's a small touch, but it shows Firsty is thinking about retention, not just acquisition.

• We've rounded up the best eSIMs for Europe.

Comfort+ (Daily plan, starts at €1/day)

(Image credit: Firsty)

Comfort+ is the entry-level paid option, and for most travelers it's probably the right call. At €1 per day, you get ad-free browsing, faster speeds, and access to social media and navigation without throttling out of the gate.

The speed structure is tiered:

• First 500 MB: up to 5 Mbps
• 500 MB to 1.5 GB: 1 Mbps
• Beyond 1.5 GB: 512 Kbps

You stay connected throughout. The speed just steps down as you use more data. For a travel day of maps, messaging, and light browsing, 500 MB at 5 Mbps is more than comfortable. Video calls are pushing it. Streaming is not what this plan is for.

You only pay for the days you actually use it. Pause it when you're back home, reactivate when you land somewhere new. Your unused days don't expire.

First Class (Daily plan, starts at €2/day)

(Image credit: Firsty)

First Class is for people who need their phone to work like it does at home. At €2 per day, you get 5GB of high-speed data, connecting to the fastest available local network, up to 5G where available. Streaming works, hotspot works, and if you need to get your laptop online through your phone, this is the plan for it.

After 5GB, speed steps down rather than cutting you off completely, which is the right call compared to providers who just shut you down.

Firsty: Features

Firsty keeps the experience deliberately simple. You are not choosing between regional bundles, juggling add-ons, or shuffling through settings to find hidden upgrades. The core product is straightforward, and a few features stand out more than the rest.

• Freemium model: Most "free" tiers in tech are either so restricted they're useless or so good they're unsustainable. Firsty's free tier threads the needle. You get enough connectivity for the moments that matter most after landing: booking a ride, pulling up directions, checking your hotel reservation, or messaging the person you're meeting. It is a simple idea, but a useful one, and the ad-supported model is what makes it possible.
• Single global eSIM: You install Firsty once and keep using the same eSIM wherever you go. There is no buying a new plan for every country, switching profiles, or sorting through a marketplace of regional options. The app handles the network side automatically, connecting to a local provider in the background. On the Daily plan, that coverage extends to more than 180 countries.
• Flexible daily billing: You pay only for the days you use a paid plan. If you're on Comfort+ and you spend a day on a beach with no need for more than basic speed, you drop back to Free and don't get charged. That flexibility is genuinely useful and separates Firsty from subscription-first competitors.
• Pause and save: Unused paid days do not disappear when your plans change. You can pause your plan and pick it back up later, with the remaining credit still intact. There is no pressure to burn through data just because you paid for it, which is especially useful if your travel schedule changes unexpectedly or you end up needing less connectivity than planned.

Firsty: Installation and activation

(Image credit: Firsty)

Getting Firsty running takes about five minutes, and most of that time is waiting for the eSIM profile to download.

Download the app from the App Store or Google Play, create an account using your phone number or email, answer a few quick questions about which apps you use most, and pick your plan. The app then walks you through the eSIM activation steps: labeling the new line, setting it as your data line, and toggling the right settings on your device.

Two things to get right before you start. First, do the setup on Wi-Fi. Installing an eSIM profile over mobile data is asking for trouble, and Firsty's own instructions flag this clearly. Second, once the eSIM is active, turn off automatic data switching. If your phone defaults back to your home carrier for data, you'll get hit with roaming charges and wonder why Firsty isn't doing anything.

After that, it's set and forget. You don't re-install for each trip or each country. The app handles network selection automatically wherever you land.

Firsty: Compatible devices

Firsty covers a solid range of devices across Apple, Samsung, Google, and a handful of other Android brands, which means most travelers won't have a compatibility issue before they even check.

That said, the official supported devices list hasn't been updated since December 2024, so newer releases like the iPhone 16 series, iPhone 17, Pixel 9, and Galaxy S25 are missing from it entirely. If you're on a recent flagship, chances are it works fine, but it's worth checking with Firsty's support team to confirm before you travel.

Brand

Compatible Models

Notes

Apple iPhone

iPhone XR, XS, XS Max, 11, 11 Pro, 11 Pro Max, SE 2 (2020), 12 / Mini / Pro / Pro Max, 13 / Mini / Pro / Pro Max, SE 3 (2022), 14 / Plus / Pro / Pro Max, 15 / Plus / Pro / Pro Max

iPhones purchased in China and Hong Kong are not eSIM compatible

Apple iPad

iPad Pro 11" (A2068, 2020+), iPad Pro 12.9" (A2069, 2020+), iPad Air (A2123, 2019+), iPad (A2198, 2019+), iPad Mini (A2124, 2019+), iPad 10th gen (2022)

LTE-capable iPads only

Samsung Galaxy S

S20 / S20+ / S20+ 5G / S20 Ultra, S21 / S21+ 5G / S21 Ultra 5G, S22 / S22+ / S22 Ultra, S23 / S23+ / S23 Ultra / S23 FE, S24 / S24+ / S24 Ultra

Some US, South Korea, China, and Hong Kong models are not compatible

Samsung Galaxy Note

Note 20, Note 20 Ultra 5G

Some US, South Korea, China, and Hong Kong models are not compatible

Samsung Galaxy Z

Z Fold, Z Fold2 5G, Fold3 5G, Fold4, Fold5 5G, Z Flip, Flip3 5G, Flip4, Flip5 5G

Some US, South Korea, China, and Hong Kong models are not compatible

Samsung Galaxy A

A54 (select model numbers only)

Some US, South Korea, China, and Hong Kong models are not compatible

Google Pixel

Pixel 2 XL, 3 / 3 XL / 3a / 3a XL, 4 / 4a / 4 XL, 5 / 5a, 6 / 6a / 6 Pro, 7 / 7a / 7 Pro, 8 / 8 Pro

Some Pixel 3 and 3a models from Australia, Japan, Taiwan, and those on US or Canadian carriers are not compatible

Huawei

P40, P40 Pro, Mate 40 Pro, P50, P50 Pro

Oppo

Find X3 Pro, Find N2 Flip, Reno 5A, Reno 6 Pro 5G, Reno 9A, Find X5, Find X5 Pro, A55s 5G

Sony Xperia

10 III Lite, 10 IV, 10V, 1 IV, 1 V, 5 IV, 5 V, Ace III

Xiaomi

12T Pro, 13, 13 Lite, 13 Pro, 13T Pro

Motorola

Razr 2019, Razr 5G, Razr 40, Razr 40 Ultra, Razr+, Edge+, Edge 40, Edge 40 Pro, Edge 40 Neo, G52J 5G, G52J 5G II, G53J 5G, G54 5G, G55 5G, G84

Firsty also lists compatible models from Sharp AQUOS, Rakuten, Honor, Vivo, and Nokia. These cover a smaller share of travelers but the full model breakdown is available on Firsty's supported devices page.

Firsty: Support

(Image credit: Firsty)

Firsty's support options are straightforward. You can reach the team via email at support@firsty.app, through the in-app help section, or through the online help center which covers the most common setup and troubleshooting scenarios.

The response quality holds up. App Store reviewers consistently describe the support team as responsive and easy to reach. On Trustpilot, where Firsty currently holds a 4.1/5 score with 283 reviews, customer support gets regular mentions as a positive.

Reddit tells a more mixed story. Most complaints are about device compatibility on unsupported Android models and connection issues during setup, nearly always caused by not installing on Wi-Fi.

For straightforward travelers, iPhone users and anyone on a mainstream Samsung, the setup is clean and support rarely needs to get involved after the first install.

Firsty: International calling

(Image credit: Firsty)

Most eSIM providers stop at data. Firsty added calling in 2025, and it is one of the more practical additions to the service. If you need to ring a hotel, contact a host, or make a quick local call, you can do it without relying on a separate app. The feature is currently available in the US, Canada, Singapore, Japan, and Europe, with additional regions expected over time.

The calling works through a credit top-up system that's separate from your data plan. You buy credits in flexible amounts from €1 to €50, they don't expire, and you pay per minute based on the destination. Calls go to 200+ countries, and you keep your own caller ID throughout, so whoever you're calling sees your regular number, not some unknown international line.

Rates are transparent and published openly on the website. Europe and North America average around €0.03 per minute. South America runs higher at around €0.07. Asia averages €0.12 per minute. Africa and Oceania are the most expensive at €0.56 and €1.08 per minute respectively. The app shows you the real-time rate before every call, so there's no bill shock at the end of a trip.

Firsty: Final verdict

For budget travellers, Firsty’s free plan alone is worth installing before any international trip. You land, you have Maps and Uber working, and you haven't spent anything. For people who need more, Comfort+ at €1/day covers a normal travel day without breaking the bank. First Class at €2/day handles everything, including streaming and hotspot.

The limitations are real but narrow. The free plan only works in North America, Europe, and APAC. Comfort+ isn't built for heavy data use. Plus, Firsty is still a relatively young company, founded in September 2023, which means its network depth in some regions won't match providers who've been at it for a decade.

But for what most travellers actually need, it delivers. And no other eSIM provider at this price point is offering a permanent free tier with this kind of coverage.

Firsty: FAQsWhat is Firsty?

Firsty is a Dutch eSIM provider founded in September 2023. It offers a permanent free data tier supported by short ads, alongside paid Daily plans for travelers who need faster speeds. One eSIM covers 180+ countries with no SIM swaps or regional plan purchases required.

How does the free plan work?

You watch a short ad every 59 minutes in exchange for data at speeds between 256 Kbps and 1 Mbps. That's enough for WhatsApp, Google Maps, email, and Uber. The free plan is available to users in North America, Europe, and APAC only.

What is the difference between Comfort+ and First Class?

Both are Daily paid plans. Comfort+ costs €1/day and gives you tiered speeds starting at 5 Mbps for the first 500 MB, stepping down to 1 Mbps and then 512 Kbps as you use more data. First Class costs €2/day and gives you 5GB of high-speed data up to 5G, plus video streaming and hotspot. After 5GB, you stay connected at a lower speed rather than being cut off.

Does Firsty work on my phone?

Firsty works on most modern eSIM-capable smartphones including iPhone XR and later, Samsung Galaxy S20 series and later, Google Pixel 2 XL and later, and select models from Huawei, Oppo, Sony, Xiaomi, Motorola, and others. iPhones purchased in China and Hong Kong and some Samsung models bought in the US, South Korea, China, and Hong Kong are not compatible. Check Firsty's supported devices page for the full list.

Can I make calls with Firsty?

Yes. Firsty supports international calling to 200+ countries through a pay-as-you-go credit system. Credits range from €1 to €50 and never expire. The feature works on all plans including Free, and your own caller ID shows on every call. It's currently available in the US, Canada, Singapore, Japan, and Europe.

Do I need to buy a new plan for each country I visit?

No. You install the eSIM once and it works across all supported countries automatically. The app connects you to the best available local network without any manual switching.

How do I install Firsty?

Download the app, create an account, pick a plan, and follow the in-app steps to activate the eSIM. Do the setup on Wi-Fi, not mobile data, and turn off automatic data switching on your device after activation to avoid accidentally falling back to your home carrier's roaming rates.

The legal sector’s use of AI is maturing, and its use is going beyond simply being an external chatbot or standalone tool. The focus is shifting to embedding AI into everyday legal work, giving it access to the documents, matters, knowledge and systems lawyers use every day.

As firms increase their investment in AI technology, they need to look beyond what AI models can generate and focus on a more practical question: how do those models connect to the information lawyers rely on, where does that information sit, and how is access governed?

That connection challenge is why Model Context Protocol (MCP) has started to attract so much attention.

Separating the standard from the assumptions

MCP is an open standard that gives AI tools a more consistent way to connect to external systems, data sources and applications. For all the excitement around MCP, firms need to be clear about what it does, and just as importantly, what it does not do. Otherwise, there is a risk that firms either overestimate what MCP can solve on its own or dismiss it as just another technical acronym.

The reality sits somewhere in the middle. MCP can help AI tools connect to the systems and data sources firms already use, but it does not automatically solve challenges around governance, integration, permissions or legal context. With misconceptions starting to spread across the legal sector, here are five common myths to clear up.

Myth 1: MCP is only for Claude

Although MCP was created by Anthropic, it is not just a Claude feature. It is an open-source framework and is now becoming part of the broader conversation around how AI agents connect to external tools, data sources and systems.

That distinction matters for law firms and legal organizations generally. If MCP is treated as a single-vendor feature, it can be dismissed as something tied to one model or product roadmap. But as a broader connectivity standard, firms need to think about how it fits into their wider AI strategy, integration architecture and governance model.

Myth 2: MCP replaces APIs

APIs still matter. They remain the backbone of platform-to-platform connections, allowing software systems to exchange data and trigger actions. MCP does something different: it acts as a protocol layer on top of APIs, giving AI agents a more standard way to discover and interact with approved legal systems, tools and data sources.

Put simply, MCP is closer to a universal adapter for AI, than an alternative to APIs. Just as USB gives different devices a common way to connect, MCP gives AI tools a more consistent way to understand what systems and functions are available to them, and how they can interact with those systems.

But it does not remove the need for APIs, authentication, system owners or clear rules about what AI tools can and cannot access.

Myth 3: MCP means moving documents into AI tools

There is a common misconception that connecting AI to legal systems means copying a large volume of documents into external AI platforms. In practice, AI should only be given controlled access to governed systems. Documents, precedents and matter files can remain within the firm’s trusted environment, with AI tools using MCP to retrieve only the information they are authorized to access.

Rather than creating another uncontrolled copy of sensitive material, firms can let AI work with the right information while existing permissions, security policies and governance controls remain intact. To hit the right balance, legal IT leaders should be asking “Where does the data stay, what is exposed, and how is access controlled?

Myth 4: All MCP integrations are the same

As MCP becomes more common, there will be a temptation to treat any MCP-compatible integration as broadly equivalent. That would be a mistake. MCP standardizes the connection, rather than the value of what comes through it. One integration may provide a basic route to retrieve files.

Another may provide richer information about permissions, matter relationships, document history, metadata and audit trails. Both may be MCP-compatible, but they will not deliver the same outcome.

Law firms need to focus on what the AI receives, whether access rights are enforced and whether activity can be audited.

Myth 5: MCP automatically gives AI legal context

MCP creates the route into systems, but it does not decide what the AI receives or understands. Connection does not mean context. In legal, this is more than a simple retrieval problem. An AI tool does not just need access to a document in a DMS. It needs to understand the matter, client, permissions, version history, related work and institutional knowledge around it.

For example, an AI tool may be able to find a precedent agreement, but does it know whether that precedent is current, whether it belongs to a similar matter, whether it reflects the firm’s preferred position, or whether the lawyer has permission to access the related material? Without that context, AI may generate an answer, but the answer may not be reliable enough for legal work.

That is why MCP should be seen as the access layer, not the intelligence layer. It can help AI tools connect to legal systems, but the value comes from what those systems expose through MCP: governed, matter-aware and permission-sensitive context.

Connection is only half the story

MCP gives law firms a more standard way to connect AI tools to the systems they already use. But legal AI depends on more than connection. The firms that benefit most will be those that treat MCP as the starting point, not the destination. The quality of the information, the controls around it and the context that gives it meaning will determine whether connected AI makes a genuine impact in legal work.

We've featured the best AI website builder.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

run by Society of the Sacred Heart, is closing after a decade.'/>

For 10 years, a Catholic-run home offered young women an affordable place to live in the nation's capital as they studied and launched careers. Now, that community is saying goodbye.

(Image credit: Charlotte Kesl for NPR)

Unexpected absences can cost employers tens of thousands of dollars a month. An app that helps workers with last-minute transportation, food and childcare is making a difference.

(Image credit: Claire Harbage)

New bans on using student IDs for voting in Indiana and New Hampshire have sparked lawsuits after creating the latest hurdles to the ballot box for the U.S. voting-age group least likely to vote.

(Image credit: Joseph Prezioso)

Four states are suing the parent company of Facebook and Instagram, alleging that the platforms were designed to hook kids and that Meta hid the mental health risks.

(Image credit: Chris Delmas)

In some wellness circles, tadalafil is promoted as a longevity drug. There is evidence showing a link between it and a lower risk of death in men, but there are no well-controlled trials in humans.

There's a lot homeowners can do to prepare for wildfires, such as managing flammable plants and using fire-resistant building materials. But some states struggle to pass rules requiring those steps.

(Image credit: Josh Edelson)

Landowners and activists worry about impact to the environment around SpaceX's complex.

(Image credit: Reynaldo Leaños Jr.)

USS Washington to relieve USS Lincoln after concerns about conditions, Ukraine launches one of its largest aerial attacks on Russia, a look at where talks stand regarding the future of Gaza.

Enterprise AI strategy spent two years chasing a single objective: reach the frontier before competitors do.

The default path was a public cloud account, an API key from OpenAI or Anthropic, and a willingness to absorb cost in exchange for speed.

That reality is now running out of road.

The numbers tell the story. Gartner forecasts worldwide AI spending reaching $2.52 trillion in 2026, up 44% year on year, with $1.37 trillion of that flowing into AI infrastructure alone.

In fact, in mid-2025, they claimed that procurement of AI had entered a “Trough of Disillusionment,” where scaling depends on predictable ROI, rather than visionary pilots.

The pressure has now shifted from how fast enterprises can pilot AI to whether they can sustain, govern, and defend it in production.

The Race to the Front is Over – Now Comes the Bill

We are now past AI 1.0, where simple access to cutting-edge AI was the differentiator. Now it’s AI 2.0’s turn, where inference economics, data gravity, latency and control decide the outcomes. Token prices have fallen almost tenfold annually since 2021, but AI spend overall by organizations has increased. That’s because more capable models have enabled greater ambition.

Anthropic, OpenAI, and Mistral are now stratifying offerings between flagship reasoners and lower-cost workhorses precisely because customers refuse to pay flagship prices for every task. McKinsey’s 2025 State of AI survey confirms the pattern - adoption is increasing, but impact at scale remains elusive for most organizations.

Now CIOs have stopped asking which model, but where each workload needs to run and how much it’s going to cost.

Inference Cost Inflation

Banks delivering the next best action are a good example: the in-app, in-branch, or call-center recommendation served in milliseconds against a customer’s live context. The best banks prove that personalization at this layer can lift revenue by 5-15%. To give a firsthand example, a global bank we work with launched an AI assistant that has already resolved more than 1.5 million customer inquiries in its first year, driving huge efficiencies.

But the inference economics are unforgiving at this scale. A single agentic decision can chain five to twenty model calls, each carrying its own context window. The cost gap between £0.50 and £3 per million input tokens seems trivial in a single-turn demo. Spread across hundreds of millions of customer events, it becomes the difference between a money-making feature and a money-burning one.

This isn’t a hypothetical either. Uber’s 5,000-strong engineering team’s use of Claude Code burned through the company’s entire annual AI budget in the first four months of this year. And AI companies are responding to this market shift. Decagon, after re-architecting onto an open-source multi-model stack on NVIDIA Blackwell, dropped cost per voice query by sixfold. Next best action isn’t a marketing decision anymore; it’s an economic decision.

Organizations making the structural shift now will outcompete those treating model selection as an afterthought.

Complexity Doesn’t Disappear, It Just Moves

The hardest lesson of the past 18 months is that model commoditization does not reduce enterprise complexity but relocates it. Open weights from Mistral or DeepSeek cut experimentation cost, but orchestration, governance, evaluation, and integration burdens move up the stack and sit with the buyer.

Enterprise leaders should be measuring unit economics per useful task, operational burden per deployed agent, and the ratio of inference spent on the governance scaffolding around it. That ratio is typically 1:5 or worse.

A second architectural shift is arriving: sub-quadratic attention.

Approaches from DeepSeek, Google, and Cartesia are collapsing the cost of long-context reasoning by orders of magnitude, with recent benchmarks showing 100x to 300x cost reductions at comparable accuracy.

Large banks will now be able to run whole-portfolio risk modelling, multi-decade fraud detection and cross-jurisdiction Know-Your-Customer (KYC) as single-pass operations - no more chunked retrieval workarounds.

Telcos can make network operations, predictive maintenance and multi-year customer journey reasoning more economically viable at scale. And manufacturers can move full-plant simulation and supply-chain disruption forecasting from periodic batch jobs to continuous reasoning.

The architecture that wins will not be the one with the cheapest token. It will be the one that places compute closest to the data, under the right jurisdiction, with governance that holds.

Sustainable, sovereign, controlled - that’s the new triad. The enterprises that build for it now will define the next decade.

We've reviewed, rated, and ranked the best business plan software.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Artificial intelligence (AI) is rapidly becoming embedded in the modern workplace, with employees are increasingly turning to AI tools to work more efficiently and boost productivity.

This growing demand for faster, more effective ways of working is driving the rise of shadow AI - the use of AI tools outside approved organizational controls and governance frameworks – which results in organizations quickly losing visibility into data usage and potential risks.

The scale of this challenge is significant. While 90% of executives are confident in their organizations' visibility into AI tools, just 52% of employees admit to using AI tools without approval, often through personal accounts.

As a result, organizations are left grappling with a widening gap between AI adoption and AI governance.

The next frontier of AI risk

When AI is used without formal oversight, it can bypass governance controls, increasing the risk of errors, regulatory breaches and sensitive data leakages. Organizations are most exposed when AI is already influencing business-critical activities, from customer service and operational decision-making to software development and content creation.

The challenge will intensify as businesses move beyond large language models, which generate information, to large action models and agentic systems that can take action. These systems can diagnose issues, recommend actions and execute workflows with minimum human input, increasing both the speed and scale at which mistakes occur.

A shadow agent operating outside approved governance frameworks could trigger harmful actions before organizations have the visibility and governance capabilities needed to intervene.

There is also a longer-term risk that future AI systems will be trained on synthetic or lower-quality data, weakening performance and decision-making over time. Transparency and traceability will be critical to maintaining accountability, protecting ethical standards and preserving the effectiveness of AI systems as adoption continues to accelerate.

AI governance as an enabler

What works is AI governance that enables innovation while putting clear guardrails in place that are integrated, transparent, auditable, and aligned with existing risk and compliance frameworks. If AI is to be used safely, firms must be able to successfully identify exactly what went wrong and why when issues arise.

In practice, mature governance starts with an approved AI tool stack that provides safe and trusted options for common use cases. This should be supported by risk-based policies that make clear the data being handled, what can and cannot be shared, which tools are permitted, and where human approval is required.

Low-risk tasks such as drafting or summarizing content should not be governed in the same way as high-risk uses involving customer data, regulated information or business-critical decisions.

Training is equally important. The challenge, beyond only enforcing controls, involves helping employees understand why those controls exist and how to use AI responsibly. As agents increasingly diagnose issues, recommend actions, and execute workflows with minimum input, human oversight and approval processes must scale alongside them.

Interoperability will be critical to making this workable at scale, allowing organizations to operate across jurisdictions and multiple AI models without repeatedly rebuilding governance processes and systems from scratch.

Making responsible adoption the easy choice

For security and compliance leaders, the goal should be to make responsible AI adoption the path of least resistance. Employees turn to shadow AI when approved tools are unavailable, difficult to access or fail to meet their needs. Companies that focus solely on restricting usage risk driving activity further underground and losing out on the efficiency and innovation gains that AI can deliver.

Organizations that successfully balance AI productivity and control over their systems recognize that shadow AI use is often a symptom of unmet demand. Employees typically turn to unauthorized tools because they are easier to access, faster to use or better suited to the task at hand. Rather than focusing on restrictions alone, leaders should understand where AI is already being used across the business and ensure approved alternatives are available for the most common use cases.

With three-quarters of office professionals saying they would be likely to look for a new job that offered better AI skills development, firms that combine governance with opportunities to build AI skills are likely to see stronger adoption of approved tools and, as a result, less reliance on shadow AI.

Building an AI-enabled culture means giving employees the tools, knowledge and confidence to innovate within clear boundaries. By doing so, shadow AI can be reduced while the speed and agility that workers increasingly expect is maintained.

The organizations best positioned to succeed

The businesses that strike the right balance for AI success will be those that view governance as a foundation for AI adoption and not a barrier to it. By making the secure, approved path the easiest path, shadow AI risk is reduced without sacrificing productivity.

Embedding strong governance, supported by trusted and well-managed data foundations, avoids costly mistakes and allows AI to be deployed and scaled with greater safety and confidence.

We've reviewed, rated, and ranked the best small business software.

This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.

The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit

Lanterns has finally premiered on HBO Max — and, in a move reminiscent of Invincible's first-ever episode on Prime Video, the new DC Universe (DCU) TV show wastes no time in delivering an almighty shock in its first chapter.

Indeed, the sci-fi show's premiere, titled 'Pilot', ends with a moment that'll completely stun viewers and likely upset DC comic book fanatics.

I'm about to dig into the biggest talking point from Lanterns' debut episode, so consider this your one and only warning: full spoilers immediately follow after the poll below. Turn back now if you haven't seen 'Pilot' yet, or else.

Is Kyle Chandler's Hal Jordan really dead in Lanterns?

Green Lantern fans and DC purists are going to have a lot to say about this... (Image credit: John Johnson/HBO Max)

It certainly seems that way. Of course, this could be a massive fake out — after all, there are all manner of shapeshifting alien species that can impersonate humans in DC Comics, so it's possible that this is the case here. Nonetheless, I believe that Chandler's Jordan has taken his last breath in the DCU.

Okay, but how did we get here? For starters, this season's premiere wasn't shy about teasing Jordan's demise. Indeed, minutes before his death, which takes place during the show's 2026 storyline, is shown, Jordan, Aaron Pierre's John Stewart, and Kelly Macdonald's Sheriff Kerry Kane survive an attack from a suicidal bomber at Rushville's police station.

This incident, which happens as part of Lanterns' 2016 storyline, is caused by an alien masquerading as a human truck driver called Waylon Sanders. As the unidentified extraterrestrial reveals, their skeleton is retrofitted with a biometric neutron device that they can activate with a single thought.

Hal Jordan interrogates Waylon Sanders, who the former correctly identifies as an alien (Image credit: John Johnson/HBO Max)

Prior to detonating the explosive gadget, Sanders goads Jordan over the latter's fearless nature, which Sanders interprets as Jordan wanting someone or something to kill him. In that respect, then, the DCU Chapter One series telegraphs his death before it happens.

Rather than bump off Jordan in 2016, though, Lanterns withholds his passing until a decade later. Indeed, Jordan uses his power ring to form a protective bubble around Sanders just before the assailant detonates his skeletal device, thereby restricting the blast radius and saving the lives of everyone nearby.

But, that's not the end of the story. Five minutes before 'Pilot' ends, we skip ahead to 2026, which reteams us with Stewart as he heads to the same college football field that he and Jordan first met Sheriff Kane, and where their 2016 investigation into the deaths of four football fans began.

There, Stewart reunites with Kane and, after a bit of small talk about Kane's now-teenage son Noah, they head to a section of the bleacher seating where Jordan's snow-covered corpse is eventually revealed to be sitting.

Who killed Hal Jordan in Lanterns episode 1? Assessing the most likely candidates

Anybody else react like this when they saw Hal Jordan's dead body? (Image credit: John Johnson/HBO Max)

Alright, so who murdered Earth's first-ever Green Lantern? We don't know, but we can speculate on who pulled the trigger.

The first two — and arguably most likely candidates — are Sheriff Kane and the stadium's groundskeeper. Kane tells Stewart that the latter is the only other individual who currently knows that Jordan is dead. However, it's incredibly unlikely that the groundskeeper will be a prominent character moving forward, so we can rule them out.

We can do likewise with Kane. Sure, she's got a firearm to hand at all times, but I just don't see her wanting to draw heat over Jordan's death, and the potential ire of Stewart, the Green Lantern Corps, and/or the Guardians of the Universe.

Did Jordan's murderer steal his ring, too? (Image credit: HBO Max)

So, who else could it be? Of the other characters we've met, I wouldn't be shocked if Garret Dillahunt's Will Macon is behind it, regardless of whether he actually committed the deed or not. Right now, though, he's formed something of an uneasy alliance with Jordan and Stewart, so a massive, relationship-breaking event would need to happen for Macon to be involved in Jordan's demise.

There are bound to be other would-be murderers who could've taken Jordan's life who we've yet to encounter, so we might be adding more names to our shortlist in the weeks ahead.

That said — and hear me out on this before you pass judgement — what if Jordan killed himself? He's a former special forces pilot, so he knows how to handle a gun. For all we know, Waylon was telling the truth about Jordan wanting is life to be over, too — especially if, as an alien, the former has some form of perceptive superpower that allows his species to read someone else.

Who do you think killed Hal Jordan? Let me know in the comment section below. And, for more on the HBO Max show, read my Lanterns review.

Samsung has had a busy few weeks with the launch of its 2026 foldable and wearable ranges, with pre-orders just wrapping up last week. If you missed out on the pre-order deals or you have your eyes set on something else from the brand, then fret not — Samsung has just unveiled a slew of deals across its product portfolio.

The latest Samsung Secret Sale event features lots of top-rated tech that we’ve reviewed here at TechRadar, with savings of up AU$1,200 or 50% with the code SECRETAUG — but it’s only running for 48 hours, with deals ending at 10am AEST on Wednesday, August 19.

One standout deal is the excellent Samsung Galaxy Watch 8 with a 50% discount, bringing it down to just AU$374.50 for the 40mm LTE model. Given that the new Watch 9 has relatively modest upgrades for AU$699 (with a current deal), this offer is hard to pass up.

Galaxy Watch 8 (40mm, LTE): was $749 now $374.50
Our tester gave the Galaxy Watch 8 a glowing review for its slim design, built-in running coach and advanced health, sleep and AI features. The new Watch 9 has a similar look with relatively modest upgrades (new processor and larger battery), so this deal is a winner if you’re after an LTE-enabled Android smartwatch.View Deal

Music Studio 7 LS70H: was $749 now $449.40
Our friends at What Hi-Fi called the Music Studio 7 “a near-perfect product in a perhaps perfect form”, thanks to its Dolby Atmos support, big stereo sound and wealth of streaming options. This deal also makes it especially tempting for those looking to buy two for a stereo setup, which the review recommends for a full music and movie system.View Deal

Galaxy Tab S10 FE (128GB, Wi-Fi): was $1099 now $649
We’re fans of the mid-range FE-series Galaxy Tab slates, with the S10 FE+ getting praise from our reviewer for its excellent display, premium build and IP68 rating — we’ve previously considered it to be one of the best Android tablets. If you’re keen on a smaller size and want 5G connectivity, Samsung has also discounted the S10 FE to AU$749.40 or 40% off.View Deal

Galaxy S26 Ultra (256GB): was $2199 now $1539
The Samsung Galaxy S26 Ultra was named the best Android phone ever by our reviewer, praising its design, powerful hardware, AI features, excellent cameras and the new Privacy Display. This deal is only available to the online exclusive colourways Silver Shadow and Pink Gold. Also discounted are the 512GB (AU$1,749) and 1TB (AU$2,064) models.View Deal

S95H OLED TV: was $3999 now $2799
Samsung’s flagship OLED TV received a perfect score in our review, with praise for its brightness, colour, gaming features and performance. At this price, it’s still an investment — and note it’s ‘only’ a 55-inch model — but given it’s rare to see savings on Samsung TVs, this is one to consider.View Deal

• Check out our range of dedicated Samsung Australia promo codes for more savings on phones, TVs, wearables and audio

President Trump explained the decision by saying South Korea had declined to help with the "denuclearization" of Iran and he cited his good relationship with North Korea's leader.

(Image credit: Julia Demaree Nikhinson)

Here are hints and the answers for today’s NYT Connections: Sports Edition puzzle for Aug. 17, 2026.

We heard about a Maleficent coaster, a Tomorrowland revamp, a Monsters Inc octopus sushi chef animatronic, the new Cars rides and more.