89% of AI Use Hidden from IT, Salesforce's AgentExchange Launch, Microsoft’s New AI Tools

[00:00:02] It's Thursday, March 6, 2025 and I'm Dave Sobel for Things to Know Today. Employees secretly chatting with AI behind IT's back. Salesforce opens an AI marketplace, Microsoft sends digital helpers to work, MSPs, time to get those bots on the payroll. AI hits home as Microsoft and Snowflake empower local models, meanwhile Anthropik bags billions. And, AI voices get so real users catch feelings.

[00:00:30] Apple promises series glow up, but don't hold your breath, until 2027. This is the Business of Tech. I want to set the stage a bit before we talk some products. A report by LayerX reveals that an alarming 89% of enterprise artificial intelligence usage occurs outside the visibility of IT departments, exposing organizations to critical security risks.

[00:00:55] The report finds that while many organizations are adopting or evaluating AI solutions, hard data on their usage remains scarce. Approximately 15% of users access AI tools daily, with 50% using them every other week. Notably, almost 72% of employees access these tools through personal accounts, making the majority of AI usage invisible to their organizations.

[00:01:23] LayerX also found that corporate data is pasted into AI applications nearly four times a day on average among users who submit data, raising concerns over data leakage. A recent report by WalkMe highlights the critical need for better education around artificial intelligence in the channel.

[00:01:43] The State of Digital Adoption 2025 Special AI Edition report reveals a significant disconnect between business leaders and employees regarding AI adoption, with leaders feeling confident in achieving transformation goals while employees report inadequate training.

[00:02:01] Only 25% of employees noticed efficiency improvements due to AI, leading to an estimated loss of over $104 million in 2024 due to underutilized technology. Why do we care? The findings highlight a dual crisis, visibility and adoption.

[00:02:21] Organizations may be investing in AI technology, but the massive disconnect between employee use and IT oversight creates a glaring risk and a significant opportunity. First, consider the visibility gap. With nearly 89% of AI use hidden from IT oversight, organizations face unprecedented risks of data leaks, privacy violations and compliance failures.

[00:02:45] Employees copying sensitive corporate data into personal AI tools nearly four times a day emphasizes how urgently organizations need robust governance models, security policies and monitoring tools to manage AI adoption safely. Providers should recognize that this gap isn't merely a security challenge, it's an opportunity to deliver value. Helping customers regain control through data visibility and protection frameworks positions providers as essential partners.

[00:03:15] Second, there's the adoption disconnect. WalkMe's report makes it clear that investment in AI solutions often falls short of delivering measurable benefits, not because of technology failures, but due to employee readiness. Only a quarter of employees recognize efficiency improvements, resulting in tens of millions of dollars in wasted potential.

[00:03:36] This indicates a critical role providers can play in closing this gap by offering strategic AI training, digital adoption services and proactive change management. Providers who enable customers to translate AI investments into productivity gains will have a clear strategic advantage. Lastly, the discrepancy between organizations believing they manage around 37 applications while actively dealing with over 600 applications highlights how unaware businesses are about their true digital footprint.

[00:04:05] Providers have an opening here. Auditing, rationalizing and consolidating the sprawling shadow AI infrastructure can greatly improve efficiency, reduce risk and streamline costs for clients. In essence, MSPs can thrive by guiding customers through two intertwined challenges, managing AI security risks and bridging the gap between investment and meaningful adoption. Now, remember, visibility doesn't automatically mean control.

[00:04:33] Even with improved visibility and AI tools, organizations might struggle to enforce strict governance without significantly impacting productivity. Simply identifying the problem doesn't inherently resolve complex human behaviors like employees' preference or easier frictionless tools. Today's episode is supported by Huntress. Most cybersecurity solutions are built from massive enterprises with big budgets, not Huntress.

[00:05:02] They're the fully managed cybersecurity platform built for all businesses, not just the 1%. Huntress purposely builds security solutions like EDR, ITDR, SIM and security awareness training to equip their team of elite threat hunters to handle the heavy lifting of security for you. When threat actors strike, Huntress' 24x7 Global Sock shuts them down before they're even on anyone else's radar. But they do more than just chase alerts.

[00:05:29] They lead the charge in industry research and knowledge, bringing expert protection and peace of mind. That's why users on G2 rate their EDR number one for growing businesses. To see how their expert threat hunting team gets the job done, visit Huntress.com slash MSP Radio. It's within that context where we look at the announcements. Let's start with the agentic AI items.

[00:05:56] Salesforce has launched Agent Exchange, a new marketplace for its digital labor platform, Agent Force, aimed at partners and developers in the digital labor market. The initiative follows the success of AppExchange, the first enterprise cloud marketplace, which has achieved over 13 million app installations. With more than 200 partners already engaged, including major players like Google Cloud and DocuSign, the platform offers hundreds of ready-made actions and templates to help organizations quickly develop and implement AI solutions.

[00:06:27] Salesforce executives emphasize that Agent Exchange not only accelerates execution for companies, but also opens opportunities for partners and startups to thrive in the emerging digital labor market. Microsoft has announced the upcoming release of two software agents designed to automate business functions for its Dynamics 365 and Microsoft 365 platforms. These agents, the Sales Agent and Sales Chat Agent, will enter public preview in May of 2025.

[00:06:54] The company is also launching the Microsoft AI Accelerator for Sales on April 1, 2025, aimed at helping businesses transition from legacy customer relationship management systems. Microsoft has launched a new AI tool named Dragon Copilot, designed to help doctors and nurses automate tedious administrative tasks in healthcare.

[00:07:15] The tool combines features from Microsoft's Dragon Medical One and DAX Copilot, allowing doctors to efficiently draft clinical notes, assist in preparing patient summaries and referral letters while retrieving information from medical sources. Set to be available in the U.S. and Canada starting in May, this launch follows Microsoft's acquisition of Nuance Communications in 2022, which specializes in conversational AI for healthcare.

[00:07:41] According to Microsoft, Dragon Copilot has the potential to save healthcare professionals an average of five minutes per patient encounter, highlighting the tool's efficiency in improving clinical workflows. Microsoft also reports that nearly 70% of Fortune 500 companies currently utilize Microsoft 365 Copilot, and over 160,000 organizations have used Copilot Studio to create more than 400,000 custom agents in the last three months. Why do we care?

[00:08:10] Marketplaces and packaged AI solutions sound promising, but providers should carefully vet actual functionality against customer expectations. Initial solutions often oversell automation capabilities, requiring more manual intervention than initially promised. That's a key value. A marketplace-driven approach offers significant opportunities for MSPs, including rapid deployment, scalable solutions, and a new revenue channel from delivering and customizing digital labor.

[00:08:39] Providers can benefit by specializing in industry-specific or niche workflows, building reusable templates, and accelerating customer implementation timelines. Those new chat agents and tools, particularly the sales agent and sales chat agent, point to a strategic move towards integrating AI directly into core business processes. Serving MSBs and mid-market organizations can leverage these standardized agents to facilitate smoother CRM transitions.

[00:09:04] The launch of Dragon Copilot by Microsoft highlights a critical vertical opportunities for providers operating within healthcare, emphasizing tangible improvements in efficiency, saving healthcare providers up to five minutes. For MSPs, this represents a high-value, measurable benefit that aligns directly with customer priorities, reducing administrative burdens, improving patient care experiences, and enhancing compliance and data management. I also have a few more to broadly cover.

[00:09:32] OpenAI has made its latest model, GPT-4.5, available to ChatGPT Plus members, expanding access after an initial release for pro subscribers and developers. The highly anticipated model is touted as the largest and most advanced yet, designed to tackle complex problems with improved understanding and reduced hallucination rates.

[00:09:51] OpenAI CEO Sam Altman described GPT-4.5 as the first model that feels like conversing with a thoughtful person, although he cautioned that it does not significantly outperform existing models and artificial intelligence benchmarks. The rollout is expected to take up to three days, with usage limits subject to change based on demand. OpenAI is preparing to enhance its infrastructure, requiring tens of thousands of graphics processing units to support this giant and costly model.

[00:10:18] Microsoft has announced the release of two new local artificial intelligence models, DeepSeq R1 Distilquen 7b and DeepSeq R1 Distilquen 14b for its co-pilot plus PCs. This follows the earlier introduction of a 1.5 billion parameter model, which has already demonstrated impressive capabilities. The new models represent significant advancements in the democratization and accessibility of AI, allowing developers and researchers to utilize powerful machine learning directly from their devices.

[00:10:48] Snowflake has announced an expanded partnership with Microsoft, integrating the Microsoft Azure OpenAI service into its Cortex AI platform. The collaboration will allow enterprises to build AI-powered applications using OpenAI's advanced models directly within Snowflake, enhancing capabilities in audio, video, and text processing in real time. According to a report by MIT Technology Review Insights, 59% of enterprises identify data governance, security, or privacy as significant challenges in deploying generative AI.

[00:11:19] The integration aims to address those concerns by offering built-in data governance and security measures. Snowflake will make these models available across select Microsoft Azure regions in the U.S. with plans for global expansion. Additionally, Snowflake Cortex agents will be introduced to Microsoft 365 co-pilot Microsoft Teams, enabling users to interact with their data using natural language, improving productivity and decision-making within organizations.

[00:11:45] And Anthropik has successfully raised $3.5 billion in a Series E funding round, bringing its valuation to $61.5 billion. This funding round, led by Lightspeed Venture Partners, saw participation from notable investors such as Bessemer Venture Partners and Fidelity Management. Notably, Anthropik has received over $8 billion in total investments from Amazon, which includes a significant initial investment made in 2023.

[00:12:13] The company plans to use the capital to enhance its next-generation artificial intelligence systems and expand its global reach. Recently, Anthropik launched its latest generative AI model, Cloud 3.7 Sonnet, which is reported to have improved coding capabilities. Why do we care? Microsoft envisions Windows as a leading platform for AI innovation, stating that the future of AI is happening now. I'm much more interested in the broader idea of AI brokerages, like Microsoft's AI Foundry.

[00:12:43] Let's also highlight the power of local models. These localized, accessible AI models represent a democratization of AI capabilities, enabling providers to deliver advanced AI services on-premise or directly within client infrastructure. It supports scenarios where data privacy, security, or latency concerns prevent reliance on cloud-based models, giving MSPs tangible ways to differentiate in a crowded market.

[00:13:08] Finally, I'm dismissive of Anthropik's soaring valuation, which raises concerns of an inflated AI bubble. Providers should remain cautious about committing heavily to a single vendor or ecosystem, which makes the brokerage strategy the logical one.

[00:13:26] A new conversational voice model from AI startup, Sesame, has generated attention for its astonishing realism, with some users reporting emotional connection to the AI voices named Miles and Maya. Released last month, the model has been described as crossing the uncanny valley of AI speech, with one user noting the experience felt genuinely human.

[00:13:48] The model was developed using a combination of two AI systems that both process text and audio, achieving near-human quality with 8.3 billion parameters trained on about 1 million hours of audio. I wanted to highlight that they'd seen this in multiple sources. Ars Technica is the primary link, and The Verge covered it too. Apple's long-awaited upgrade for Siri may not arrive until 2027, according to a report by Mark Gurman from Bloomberg.

[00:14:15] While an immediate update powered by a large language model is expected in iOS 18.5, it will operate separately from the existing Siri system. The ultimate goal is to merge these into a more efficient architecture that can handle both simple commands and advanced artificial intelligence tasks. Apple had initially planned to introduce this combined system in iOS 19.4, but delays mean it would not be unveiled during the upcoming Worldwide Developers Conference in June.

[00:14:42] Internally, issues such as ineffective leadership and competition for AI talent are hindering progress, exacerbating challenges as rival companies advance their own systems. Why do we care? I tried Sesame, and it's compelling, distinctly the most natural voice interface I have worked with. Making it convincing and useful is the trick. Sesame is compelling, but lacks capabilities. Siri isn't even capable. Alexa Plus joins the party in a month or so.

[00:15:09] Voice as a UI has the pieces to be something it seems really far away. Apple might have time. Are you and your clients tired of the time-consuming ticket tennis of coordinating meetings and help desk calls? Wouldn't it be better to automate this process with a tool that connects directly to ConnectWise Manage or Autotask?

[00:15:35] TimeZest offers scheduling automation that gives you complete control of your schedule and eliminates the hassle of calendar ping pong. As the only service designed specifically for MSPs, it integrates into your workflow and makes scheduling appointments easy on you and your clients. Plus, you can try TimeZest for free.

[00:15:58] Visit TimeZest.com slash MSPRadio and use the code MSPRadio to get 10% off your first year of TimeZest. Thanks for listening. Today is National Dentist Day, National Dress Day, National Frozen Food Day, National Oreo Day, and apparently, Day of the Dude. Nerdy Ocon will be held in Palm Springs, California from April 7th through 9th. Visit NerdyOcon.com to learn all about it.

[00:16:25] The Business of Tech is written and produced by me, Dave Sobel, under ethics guidelines posted at businessof.tech. If you've enjoyed the show, make sure you've subscribed or followed on your favorite platform. It's free and helps directly. Give us a review, too. If you want to support the show, visit Patreon.com slash MSPRadio and you'll get access to content early. Or buy our Why Do We Care merch at businessof.tech.

[00:16:55] Have a question you want answered? We take listener questions, send them in, ideally as a voice memo or video to question at MSPRadio.com. I answer listener questions live on our Wednesday live show on YouTube and LinkedIn. If you've got a comment or a thought on a story, put it in the comments if you're on YouTube or reach out on LinkedIn if you're listening to the podcast. And if you want to advertise on the show, visit MSPRadio.com slash engage.

[00:17:22] Once again, thanks for listening and I will talk to you again on our next episode. Part of the MSP Radio Network. We're going to talk to you. We'll talk to you again on YouTube and see more and hope to list them. Thank you.