Bluevine · Jul 15
Bluevine study: 74% of small businesses use AI, but 78% don't trust it with basic tasks
A Bluevine study of 942 small businesses found 74% now use or test AI and 48% save four or more hours a week, yet 78% do not trust AI with basic tasks and 82% hit barriers to deeper use. Data security and privacy jumped to the top barrier at 33%, up from 23% a year ago.
▸ The MSP Angle
Why don't small businesses trust the AI they already use?
Adoption is not the gap anymore; trust is, and the top reason is security, now cited by a third of SMBs and rising fast. That is your opening stated in the client's own words. The MSP that shows up with governed, auditable AI where the data stays controlled sells the missing ingredient, and the four hours a week AI already saves them funds the conversation.
Read at Bluevine ↗
TechCrunch · Jul 21
Google ships Gemini 3.6 Flash and a cheaper Flash-Lite, and starts training Gemini 4
Google DeepMind released its workhorse Gemini 3.6 Flash and a cheaper 3.5 Flash-Lite, with 3.6 Flash claiming up to 17% lower token usage and stronger coding and multimodal scores, available day one across AI Studio, the Gemini API, and Vertex AI. Google said 3.5 Pro remains delayed and it has begun its most ambitious pre-training run yet for Gemini 4.
▸ The MSP Angle
Do frequent model updates matter if I resell AI to clients?
A 17% token cut is margin if your delivery can adopt it without a rebuild, and dead weight if a client is pinned to last quarter's model. The lesson repeats every few weeks now: model-agnostic routing lets you move to the cheaper, better option the day it ships. Price and architect for a world where the underlying model changes constantly and the client never notices.
Read at TechCrunch ↗
ReliaQuest · Jul 16
Ransomware victims fell 15% for the quarter but are up 51% on the year
ReliaQuest counted 2,252 named ransomware victims in Q2 2026 across 90 groups and 99 countries, down 15% from Q1 but up about 51% year over year, with a newcomer called The Gentlemen topping the list at 300 victims. New tradecraft includes kernel-level EDR bypass and LLM-accelerated affiliate tooling, and the report flags rising third-party and vendor supply-chain exposure.
▸ The MSP Angle
Is ransomware slowing down or speeding up?
The quarterly dip is noise; the year-over-year climb and the churn of new crews is the signal, and attackers are now using LLMs to move faster. Two takeaways to sell: EDR alone is being bypassed, so layered detection and tested recovery matter more than any single tool, and the supply-chain warning means your own vendor stack is part of every client's risk surface.
Read at ReliaQuest ↗
The Hacker News · Jul 20
Researchers expose an LLM-built phishing kit behind a global infostealer campaign
Rapid7 found an exposed delivery server for a phishing operation whose materials showed LLM-generated output, abusing a Windows shortcut flaw and attacker-controlled WebDAV shares to bypass SmartScreen. It logged 77,098 requests from 3,892 IPs across 101 countries in a single week, delivering a .NET infostealer that grabbed wallets, browser credentials, and Telegram sessions.
▸ The MSP Angle
Is AI making phishing harder to catch?
The tells your users were trained to spot, broken English, clumsy formatting, are gone; the lures are now machine-written and clean. Awareness training still helps, but the durable defense is technical: block the delivery techniques, watch for the WebDAV and shortcut-file behavior in this campaign, and assume any credential a user can be tricked into typing will be. Phishing-resistant MFA is the answer that survives better bait.
Read at The Hacker News ↗
Channel Insider · Jul 16
Delinea launches an AI-enabled partner program with margin protections
Identity-security vendor Delinea rolled out a three-tier partner program adding AI-driven just-in-time training, an intelligent partner portal, codified deal registration, teaming protections, and renewal discounts for the incumbent partner. The launch cites Delinea's finding that 92% of organizations expect AI to amplify identity-related threats.
▸ The MSP Angle
What should I look for in a vendor partner program now?
Two things worth reading past the AI-enablement headline: deal-registration and incumbency protections are how a program keeps another partner from underbidding you on your own renewal. As identity threats grow, expect more security vendors to court MSPs this way. Judge these programs on margin defense and lead protection, not on how much AI training they bundle.
Read at Channel Insider ↗
Tech Startups · Jul 22
Candid Health raises $120M as its AI billing agents post 190% growth
Candid Health closed a $120 million Series D, taking total funding to $219.5 million, on 190% year-over-year growth and 180% net dollar retention for AI agents that automate medical billing and claims. It is a concrete professional-services case: a bounded, single-function agent producing measurable financial results.
▸ The MSP Angle
Where are AI agents actually making money right now?
Not in do-everything assistants, but in narrow, boring, high-volume back-office work like billing and claims, where results are measurable and the process is repeatable. That is the template for a sellable AI service: pick one painful workflow a client runs constantly, automate it end to end, and bill against the hours or errors it removes. Depth in one process beats breadth across ten.
Read at Tech Startups ↗
Tech Startups · Jul 22
Defense-tech startup Cathedral raises $160M for AI cyber operations
Cathedral raised $160 million led by Andreessen Horowitz and Sequoia to build AI for military cyber operations, one of several nine-figure AI-security rounds in the July 22 tally. It signals continued heavy capital flowing into both offensive and defensive AI-cyber capability.
▸ The MSP Angle
Why does military AI-cyber funding matter to my MSP?
Because offensive capability built for nation-states does not stay there; the techniques and tooling trickle down to criminal crews within a couple of years, as this year's AI-run ransomware already shows. Treat every nine-figure round into AI attack capability as a preview of the threats your clients face next. The defensive gap it opens is the demand your services fill.
Read at Tech Startups ↗
Google Workspace Updates · Jul 20
Google Sheets adds combo charts and cleaner Excel imports
Google Sheets added native combo charts, such as clustered column plus line, and now preserves secondary-axis settings when importing Excel files that previously dropped them, rolling out from July 20. It is a small compatibility gain for client reporting workflows.
▸ The MSP Angle
Does better Excel compatibility in Sheets matter for migrations?
For any client you are moving between office suites, chart fidelity is exactly the kind of small breakage that generates support tickets and second-guesses the migration. Fewer of those is smoother projects and fewer callbacks. Keep a running list of these compatibility gains; they are the difference between a migration that sticks and one a client quietly reverses.
Read at Google Workspace Updates ↗
9to5Mac · Jul 21
OpenAI courts small businesses directly as ChatGPT Work and Codex pass 10M users
OpenAI announced a small-business program bundling virtual training, in-person US AI academies, getting-started guides, and small-business-tuned agents, disclosed alongside a milestone of 10 million combined ChatGPT Work and Codex users. The push aims to move SMBs from casual chatbot use to repeatable workflows.
▸ The MSP Angle
Is OpenAI going to sell AI to my clients directly?
It is trying, and that is worth watching without panic. A model vendor can hand a small business a chatbot and a training webinar; it cannot configure their tenant, govern their data, integrate their tools, or answer the phone when something breaks. The direct-to-SMB push validates the demand and defines your lane: the delivery, security, and support a self-serve program leaves on the table.
Read at 9to5Mac ↗
