Daily AI brief · September 2, 2026
DoD Opens GenAI Portal as Agentic Coding Reshapes Enterprise
The Pentagon officially launched a secure GenAI platform for defense personnel; McKinsey finds 32% of enterprises are now building software with agentic tools instead of buying it; and SMBs are adopting AI but lagging on governance, creating a critical service opportunity for MSPs.
Department of Defense Launches GenAI.mil Portal for 3 Million Personnel (Reuters/Defense.gov)
The DoD opened GenAI.mil, a secure platform bundling ChatGPT Mil, xAI's Grok for Government, and Google Gemini, with 1.7M users already onboarded.
What it means for SMBs / MSPs: Government-grade AI security templates are now public reference; you can audit and pitch equivalent controls to compliance-sensitive customers in healthcare, finance, and manufacturing.
32% of Enterprise Workforces Now Skip Software Licenses to Build With Agentic Tools (McKinsey)
A new McKinsey survey finds that 32% of organizations have skipped buying software or features because they built them internally using agentic coding tools, while large enterprises scaling agents rose from 27% to 40%.
What it means for SMBs / MSPs: This signals the death of one-off API integrations; if you're not building managed agent workflows for your clients, smaller consulting shops will. Start with internal R&D.
Anthropic's Revenue Run Rate Hits $30B; Customer Logos Double to 1,000+ in Eight Weeks (Anthropic)
Claude customer demand accelerated in 2026, with run-rate revenue surpassing $30 billion—up from $9 billion in late 2025.
The company more than doubled customers spending over $1M annualized from "over 500" in February to exceeding 1,000 by early September.
What it means for SMBs / MSPs: Claude adoption is mainstream for enterprises; positioning it as a managed service through reseller agreements now has concrete ROI benchmarks you can show clients.
SMB AI Governance Lags Behind 90% Adoption Rate—A Service Gap for MSPs (SecureWorld / GTIA)
90% of SMBs are now somewhere on the AI adoption curve, with 61% actively using AI tools in daily operations and 29% experimenting.
Yet
founders driving AI decisions in 42% of firms are explicitly calling for help, acutely aware of risks like exposed customer data and employee misuse of unapproved tools.
What it means for SMBs / MSPs: Governance is no longer a compliance checkbox—it's a $50K–$500K annual service line; shadow AI discovery + acceptable-use policies are now table-stakes differentiators.
Meta-Backed AI Agent Company Manus Formally Resumes Independence Amid China Regulation (TechCrunch)
Manus, which China's regulators forced Meta to divest from in April 2026, has formally resumed independent operations as a general-purpose agent lab for research, coding, and report writing.
What it means for SMBs / MSPs: Open-source agent tooling is consolidating around independent labs; don't bet exclusively on vendor-locked AI agent ecosystems.
LangChain / CrewAI for Multi-Agent Workflows
What it is:
LangChain, LlamaIndex, and CrewAI are orchestration frameworks that make building multi-agent systems straightforward—you can have a basic agent running in an afternoon.
Who it's for: SMB tech leads and AI champions evaluating whether to build vs. buy agent solutions; MSPs looking to productize internal automation.
How long a first test takes: 2–4 hours (scaffold a single workflow on your own data using CrewAI's template).
First concrete step: Pick one recurring SMB admin task (e.g., lead qualification, daily report compilation, invoice processing). Spin up a CrewAI agent on your local machine using a public LLM (e.g., Qwen 3-Coder via Ollama) and measure time-to-first-output. Log token consumption and supervision cost—you'll need those numbers to pitch to clients.
Ask your team which AI tools they use for work, and write down the answers. 61% of SMBs now use AI tools in daily operations, and founders say they worry about exposed customer data and staff using unapproved tools. A five-minute, no-blame conversation gets you the list. Then add one line to your AI rules naming the customer data that never goes into an unapproved tool.
*For MSPs:* Shadow AI discovery, an acceptable-use policy with staff training, and ongoing governance make a natural three-step service.
Overpromising fully autonomous agents.
Real agent deployments in 2026 require human approval at each step; agents that propose actions and wait for sign-off finish more tasks than fully unattended systems.
Supervision cost kills margins if you're not transparent about it upfront.
The "Agentic Labor Arbitrage" is shrinking fast.
Large enterprises are scaling agents in one or more functions (40%), but smaller firms remain flat at 22%.
By Q1 2027, SMBs that haven't productized at least one agent workflow will start losing candidates to competitors who have. The advantage window for MSPs is now: build one vertical-specific agent template (e.g., for HVAC scheduling, legal document review, or accounting reconciliation), then resell it as a managed service. Differentiation will be gone in six months.