¶ Introduction
Hi everyone. I'm Andy, and this is the AI Breakdown.
Welcome to your weekly news edition where I'll cover what happened in AI last week, why it matters all in less than 10 minutes. 4 00:00:13,921.228724356 --> 00:00:20,161.228724356 Let's jump straight in with a story that I think signals a proper turning point for workplace ai and honestly. 5 00:00:20,506.228724356 --> 00:00:22,546.228724356 Might send a bit of a shudder down the spine. 6 00:00:22,726.228724356 --> 00:00:31,216.228724356 If you are in customer support, Salesforce, the giant of cloud software just confirmed it's replaced nearly half of its support workforce. 7 00:00:31,306.228724356 --> 00:00:35,176.228724356 4,000 jobs gone by rolling out AI agent technology. 8 00:00:35,836.228724356 --> 00:00:38,86.228724356 Mark Benioff, Salesforce, CEO. 9 00:00:38,266.228724356 --> 00:00:39,916.228724356 Put it about as bluntly as you can. 10 00:00:40,336.228724356 --> 00:00:44,716.228724356 I've reduced it from 9,000 heads to about 5,000 because I need less heads. 11 00:00:45,526.228724356 --> 00:00:48,451.228724356 AI agents now handle around half of all support cases. 12 00:00:49,141.228724356 --> 00:00:52,651.228724356 And even chase up millions of sales leads that real people never got to. 13 00:00:53,251.228724356 --> 00:01:02,851.228724356 Some of the reassigned staff have landed in other parts of the business, but make no mistake, this is one of the broadest AI driven job shifts at a big tech firm so far. 14 00:01:03,616.228724356 --> 00:01:10,186.228724356 Now, depending on your perspective, this is either a productivity master stroke or a worrying taste of what's to come. 15 00:01:10,816.228724356 --> 00:01:21,676.228724356 Either way, it's a watershed moment and has forced the awkward question about whether all those cheery reassurances from tech bosses that AI won't mean mass layoffs are out of date or real. 16 00:01:22,276.228724356 --> 00:01:32,686.228724356 According to the latest Capgemini Research Institute report, 60% of organizations expect AI to be an active team member or supervisor within the next 12 months. 17 00:01:33,46.228724356 --> 00:01:39,676.228724356 That's a significant jump from the current 44% who already have AI functioning in these collaborative roles. 18 00:01:40,486.22872436 --> 00:01:51,16.22872436 This shift represents AI evolving from simple tools to genuine team participants with companies restructuring their operations to enable better human AI collaboration. 19 00:01:52,355.49443594 --> 00:01:59,195.49443594 Next up following on from a topic I touched on a few weeks back, big Tech is making major inroads with the US government. 20 00:01:59,615.49443594 --> 00:02:05,75.49443594 Microsoft has joined the party and has now gone all in on government adoption with its one gov deal. 21 00:02:05,465.49443594 --> 00:02:20,525.49443594 They're offering millions of US federal employees, free access to Microsoft 365 co-pilot for the next year, plus substantial Azure Cloud discounts that could extend for three years and get this, the total estimated saving for US taxpayers. 22 00:02:20,915.49443594 --> 00:02:22,430.49443594 Is a whopping $3.1 23 00:02:22,430.49443594 --> 00:02:25,145.49443594 billion in just the first year alone. 24 00:02:26,261.18017134 --> 00:02:28,931.18017134 Now let's talk about the shift to agentic ai. 25 00:02:29,441.18017134 --> 00:02:35,351.18017134 Gartner's making a bold prediction by 2026, a whopping 40% of enterprise applications. 26 00:02:36,86.18017134 --> 00:02:41,216.18017134 We'll have task specific AI agents built right in that took from just 5% today. 27 00:02:41,606.18017134 --> 00:02:51,386.18017134 And by 2035, AgTech AI could represent $450 billion in enterprise software revenue, capturing nearly a third of the entire market. 28 00:02:52,16.18017134 --> 00:02:56,6.18017134 And whilst they always take these predictions with a pinch of salt, the direction is clear. 29 00:02:56,426.18017134 --> 00:03:03,146.18017134 We are quickly moving into the AI powered era, but today's biggest challenge isn't just dreaming up AI applications. 30 00:03:03,536.18017134 --> 00:03:07,616.18017134 It's turning those slick PowerPoint promises into genuine real world value. 31 00:03:08,276.18017134 --> 00:03:15,326.18017134 Salesforce is tackling this head on with their new CRM Arena Pro, essentially, a flight simulator for AI agents. 32 00:03:15,656.18017134 --> 00:03:29,941.18017134 This clever testing ground lets teams put their AI through the ringer before unleashing it on actual customers, their goal to dramatically speed up the journey from pilot to production and boost the chances these systems will move beyond mere experimentation. 33 00:03:31,511.18017134 --> 00:03:34,871.18017134 Sios, Salesforce's head of AI puts it brilliantly. 34 00:03:35,321.18017134 --> 00:03:42,761.18017134 Pilots don't learn to navigate in a storm during an actual storm, and neither should AI agents face their first crisis without proper training. 35 00:03:43,793.18255356 --> 00:03:46,313.18255356 Now let's shift gears and talk infrastructure. 36 00:03:46,613.18255356 --> 00:03:55,853.18255356 Because the AI hardware race is far from over AWS has just made a massive power play on veiling their Traum two powered Ultra servers. 37 00:03:56,393.18255356 --> 00:04:02,333.18255356 Enormous clusters with hundreds of thousands of chips designed specifically for heavy duty AI workloads. 38 00:04:02,753.18255356 --> 00:04:04,13.18255356 And this isn't just Talk. 39 00:04:04,223.18255356 --> 00:04:10,973.18255356 Apple has already signed on as an early customer while Anthropic plans to run its models on AWS's new supercomputer. 40 00:04:11,243.18255356 --> 00:04:19,463.18255356 What's really interesting is Amazon's bold claim that a can train certain AI models at up to 40% lower cost compared to NVIDIA's hardware. 41 00:04:20,63.18255356 --> 00:04:21,743.18255356 Now, why does this matter so much? Because. 42 00:04:22,703.18255356 --> 00:04:30,413.18255356 The notorious NVIDIA tax on GPUs has been a major headache for anyone wanting to deploy large models without bankrupting themselves. 43 00:04:30,893.18255356 --> 00:04:43,613.18255356 With Amazon now vertically integrating its own chips and cloud infrastructure, we're likely to see both faster innovation and crucially some relief from the GPU shortages that have been strangling the AI industry. 44 00:04:44,228.18255356 --> 00:04:53,888.18255356 While it's early days when industry giants like Apple are jumping on board, it's a clear signal that the rest of us should be paying attention and Nvidia might wanna start looking over a shoulder. 45 00:04:55,120.39168979 --> 00:04:59,50.39168979 Let's talk about AI and copyright in a bombshell development. 46 00:04:59,110.39168979 --> 00:05:01,885.39168979 And Thropic has just agreed to a massive $1.5 47 00:05:01,885.39168979 --> 00:05:08,20.39168979 billion settlement with authors after allegedly using about half a million books from Shadow libraries to train. 48 00:05:08,20.39168979 --> 00:05:12,100.39168979 Its Claudia AI that breaks down to roughly $3,000 per book. 49 00:05:12,700.39168979 --> 00:05:16,780.39168979 Critically, they've agreed to purge all that pirated content from their systems. 50 00:05:17,260.39168979 --> 00:05:18,700.39168979 The judge made it crystal clear. 51 00:05:19,60.39168979 --> 00:05:29,110.39168979 While training on copyrighted text might sometimes qualify as fair use, wholesale copying and storage definitely crosses the line, hence anthropics eye watering payout. 52 00:05:29,740.39168979 --> 00:05:33,880.39168979 So what does this mean for the industry? It's potentially a game changing precedent. 53 00:05:34,465.39168979 --> 00:05:42,445.39168979 Rather than fighting lengthy legal battles, other AI companies might now rush to the negotiating table to avoid similar financial blood baths. 54 00:05:42,925.39168979 --> 00:05:45,565.39168979 For authors, though, reactions are decidedly mixed. 55 00:05:45,895.39168979 --> 00:05:50,755.39168979 Some are celebrating, but tech crunches, Amanda Siling offered a more sobering take. 56 00:05:51,235.39168979 --> 00:05:53,35.39168979 This isn't a victory for authors. 57 00:05:53,215.39168979 --> 00:05:55,285.39168979 It's yet another win for tech companies. 58 00:05:56,245.39168979 --> 00:05:59,95.39168979 The stark reality for anyone building large language models. 59 00:05:59,440.39168979 --> 00:06:04,0.39168979 Is that licensing costs and legal paper trails are no longer optional extras. 60 00:06:04,300.39168979 --> 00:06:05,920.39168979 They're absolute necessities. 61 00:06:07,78.0741673 --> 00:06:11,458.0741673 Now let's shift to the genuinely hair raising side of AI security. 62 00:06:11,968.0741673 --> 00:06:20,998.0741673 The cybersecurity community is absolutely stunned by Hex strike and AI tool that can identify and exploit zero day vulnerabilities in mere minutes. 63 00:06:21,538.0741673 --> 00:06:27,263.0741673 What once required teams of experts working for months can now happen in the time it takes to make a cup of coffee. 64 00:06:28,3.0741673 --> 00:06:34,333.0741673 The proof of concept demonstration, left security professionals speechless, but there's a silver lining. 65 00:06:34,663.0741673 --> 00:06:39,823.0741673 Organizations can use this for defensive testing to spot their own weaknesses before attackers do. 66 00:06:40,978.0741673 --> 00:06:42,718.0741673 But here's the sobering reality. 67 00:06:42,988.0741673 --> 00:06:46,288.0741673 The barrier to entry for cyber attacks is plummeting dramatically. 68 00:06:46,678.0741673 --> 00:06:58,648.0741673 Security experts are frantically urging businesses to ramp up patching and monitoring protocols, because when AI can weaponize a vulnerability, the moment it surfaces, even a slight delay could be devastating. 69 00:06:59,710.07654952 --> 00:07:04,120.07654952 In Europe, there's big money and even bigger ambitions for ai chip making. 70 00:07:04,120.07654952 --> 00:07:10,660.07654952 Titan ESML has become the largest shareholder, and Misra ai, France's AI champion. 71 00:07:11,50.07654952 --> 00:07:12,235.07654952 Invest in 1.3 72 00:07:12,235.07654952 --> 00:07:15,55.07654952 billion Euros as part of a 1.7 73 00:07:15,55.07654952 --> 00:07:18,220.07654952 billion Euros series C funding round. 74 00:07:18,880.07654952 --> 00:07:28,795.07654952 Reuters Apley said The Round will make Mistral the most valuable AI company in Europe with a pre-money valuation of 10 billion euros, or about $11.7 75 00:07:28,795.07654952 --> 00:07:29,800.07654952 billion. 76 00:07:30,610.07654952 --> 00:07:39,670.07654952 A SML isn't just writing a check, they're securing a board seat and plotting hardware, software synergies, all in the name of European technological sovereignty. 77 00:07:40,75.07654952 --> 00:07:46,495.07654952 Tying together two European technology leaders in a move to make Europe less reliant on US and Chinese AI models. 78 00:07:47,568.12899129 --> 00:07:49,848.12899129 Last up, open AI's next big move. 79 00:07:50,118.12899129 --> 00:07:51,468.12899129 They're not just building ai. 80 00:07:51,708.12899129 --> 00:07:53,448.12899129 They want to change the way you get a job. 81 00:07:53,808.12899129 --> 00:08:06,498.12899129 The Open AI jobs platform is set to launch in mid 2026 with a focus on matching employers and candidates, but also helping smaller businesses and local governments connect with AI ready talent. 82 00:08:07,173.12899129 --> 00:08:16,563.12899129 What's new here is that they'll run an open AI certification program, so you can actually prove your AI fluency in a way that presumably the platform can trust. 83 00:08:17,253.12899129 --> 00:08:19,413.12899129 Fiji Sim O Open AI's Applications. 84 00:08:19,413.12899129 --> 00:08:26,73.12899129 Chief said we'll use AI to help find the perfect matches between what companies need and what workers can offer. 85 00:08:26,823.12899129 --> 00:08:36,273.12899129 This is clearly a direct challenge to LinkedIn, which might raise some eyebrows considering Microsoft LinkedIn's parent company is open, AI's largest investor. 86 00:08:36,768.12899129 --> 00:08:39,348.12899129 But their vision goes beyond just job matching. 87 00:08:39,708.12899129 --> 00:08:49,68.12899129 They're aiming for a complete talent ecosystem, smarter hiring, faster matching, fairer assessment, and giving candidates genuine ways to showcase their abilities. 88 00:08:49,758.12899129 --> 00:08:54,888.12899129 While many HR professionals are excited about the possibilities, there's healthy skepticism too. 89 00:08:55,368.12899129 --> 00:09:02,778.12899129 While an AI powered recruitment platform simply entrench existing biases, but at lightning speed, we'll have to wait and see. 90 00:09:03,138.12899129 --> 00:09:04,523.12899129 But one thing is crystal clear. 91 00:09:05,178.12899129 --> 00:09:12,768.12899129 With Anthropic, CEO recently, predicting AI could eliminate half of all entry level white collar jobs by 2030. 92 00:09:13,188.12899129 --> 00:09:16,38.12899129 Tomorrow's job hunt won't be about perfecting your cv. 93 00:09:16,548.12899129 --> 00:09:22,218.12899129 Instead, it'll be about how effectively you can communicate with or through artificial intelligence. 94 00:09:23,303.40821758 --> 00:09:25,373.40821758 That's all for this week's AI roundup. 95 00:09:25,763.40821758 --> 00:09:29,693.40821758 If you found value in this breakdown, please leave a rating and hit subscribe. 96 00:09:29,693.40821758 --> 00:09:30,623.40821758 See you next week.