Quick data security tip: Many LLMs will use your data to train their models unless you specifically request them not to. If you're giving it data that is even in the grey area of confidentiality, you'll want to make sure you adjust your settings. To do this, you'll log in to your LLM of choice (in this case, I'll use Perplexity) and go to Account & Settings, then All Settings, then Preferences. Then I'll go to: AI data retention AI Data Retention allows Perplexity to use your searches to improve AI models. Turn this setting off if you wish to exclude your data from this process. And just toggle the button off. Allowing AI Data Retention on means Perplexity uses your searches to improve its AI models; turning it off excludes your data from this training process. The key implications are: Allowing Data Retention (ON) Model Improvement: Your searches help train and improve Perplexity's AI models for all users. Privacy: Your data is encrypted and not shared with third parties; agreements with OpenAI/Anthropic prohibit them from training on your data. Personalization: Enables better personalization through Memory features that remember your preferences. Default Setting: Free and standard Pro plans default to ON—you must manually opt out. Turning Data Retention Off Privacy First: Your searches won't contribute to AI model training. No Service Impact: Core Perplexity functionality continues normally. You still get answers. Limited Personalization: Memory may retain less information about your preferences over time. Future-Only Opt-Out: Only stops future data collection; previously gathered data remains. Key Takeaway - Choose ON if you want to contribute to AI improvement and don't mind your data being used for model training (with encryption and third-party protections) - Choose OFF if privacy is your priority and you want to exclude your data from AI training entirely For maximum privacy (guaranteed no training on your data), Perplexity's Enterprise Pro plan provides enterprise-grade privacy with SOC2 compliance, though turning off the setting achieves similar protection for Free/Pro users.

It seems like everyone pushes AI agents these days, but few explain which one, for what, or why they matter. Without really knowing where to start, people stick to ChatGPT's basic mode or bail on "powerful" tools entirely. Of course there are more out there, but we'll focus on a select few, mostly well-known names. The goal here is clarity and usefulness, not a comprehensive list of all the tools. The space splits into three clear tiers. Match yours to your needs to save hours of trial-and-error. Tier 1: Personal Assistant Tools Start here if you're non-technical. These handle complex tasks like pros. - Manus: Orbits multiple AI models for research. Feed it a prompt, and it plans, browses, generates visuals/charts, and delivers an interactive report (not just text). Save winning workflows as reusable "skills" for instant leverage on new topics. - Claude's Computer Use (via Claude desktop/app): Accesses your files. Point it at a folder of 300 screenshots—it renames, categorizes, organizes. No follow-ups needed. - Open Interpreter: Links to email/calendar/browser. Runs 24/7 in your messaging app, learns from feedback. Futuristic but setup-heavy—skip unless you're committed. Tier 2: Workflow Automation Tools Embed AI into repeatable pipelines beyond simple if/then logic. - Zapier: Easiest entry. Describe your flow; AI builds it (1-2 tweaks max). Perfect for SaaS integrations. Example: Sponsor email arrives → agent researches company → drops summary in Google Drive. - n8n: Technical powerhouse. Exposes full logic for multi-agent pipelines, branching, and human review. Ideal for scaled content ops, but there is a steep learning curve. - MindStudio: n8n-like but way easier for most users. Drag-and-drop builder with AI nodes lets you create custom agents/workflows without deep coding. Connect apps, add logic/branching, and deploy fast. Great middle ground: n8n power minus the overwhelm. (Pro tip: Start with their templates for lead gen or content automation.)

I used to use AI like most people: like a fancier Google. Then I got hit with bright-shiny-object syndrome. I tried every new tool the moment it dropped. Then I snapped back to reality. Back to my process-improvement and Six Sigma days. Back to the systems I’ve built with my marketing agency, which turns 20 this July. Now I am, together with George, teaching people how to use AI systematically, building workflows that actually save time and compound results. That’s what Pivot With AI is about. When you think in systems, work gets easier. You stop staring at individual tasks and start seeing the flow: inputs → processes → outputs → feedback loops. Map that flow, and bottlenecks jump out. See them, and you can: - Break them down - Optimize the processes feeding them - Automate the repetitive parts That’s when one-time improvements turn into compounding time savings. Most professionals waste hours on the same drains: - Email triage - Reading & analyzing reports - Admin (expenses, approvals, scheduling) - Writing the same messages over (emails, proposals, marketing copy) Ad-hoc fixes help briefly. Treating them as system failures creates lasting change. Start here: Map: Spend an hour documenting how work arrives, who touches it, where decisions happen, and how outputs are delivered. Note frequencies and average times. A simple swimlane or checklist exposes duplicated handoffs, waiting periods, and unnecessary approvals. Measure: Replace guesses with data; response times, meetings per project, % of emails needing action, time editing docs. Measurement prioritizes where you’ll get the most hours back. Optimize: Email: triage rules, check-in windows, reply templates Reports: standardized dashboards so people scan, not decode Admin: batch tasks, delegate with clear SOPs, use expense integrations Messaging: reusable frameworks and modular templates Automate: filters, scheduling links, calendar rules, no-code tools for routine data moves. Automate report generation and alerting so people only open reports when something’s out of bounds.

OpenAI’s rumored AI‑first phone Analyst Ming‑Chi Kuo says OpenAI is collaborating with MediaTek and Qualcomm on custom smartphone chips, with Luxshare as the exclusive co‑design and manufacturing partner for a new AI‑focused phone expected around 2028. Rather than relying on traditional apps, this device is rumored to lean heavily on AI agents, and it would sit alongside other hardware concepts reportedly in the works, such as ChatGPT‑powered home devices like smart speakers, glasses, and lamps. Sources: - Kuo’s post on X: https://x.com/mingchikuo/status/2048587389791269182 - CNET overview: https://www.cnet.com/tech/services-and-software/openais-rumored-phone-would-replace-apps-with-ai-agents/ Copilot Agent Mode comes to Outlook Microsoft is rolling out Agent Mode in Outlook, giving Copilot more autonomy to manage your inbox and calendar based on high‑level instructions instead of one‑off prompts. You can delegate ongoing tasks like triaging messages, sending routine follow‑ups, or keeping certain projects on your radar, while still reviewing and adjusting what the agent does along the way. The feature is currently available to organizations in the Microsoft 365 Copilot Frontier program, with examples already emerging of people scheduling recurring “prompted” workflows for their agents. Source: - Frontier/agents update example: https://www.linkedin.com/posts/nickdc_we-can-now-schedule-prompts-for-declarative-activity-7400525282555719680-eGl3 (You can also go to Microsoft’s main Copilot page if you want a more general explainer) https://www.microsoft.com/microsoft-365/copilot China halts Meta’s $2B Manus deal Chinese regulators have ordered Meta and AI startup Manus to unwind their roughly $2 billion acquisition agreement, ending a months‑long review with little public explanation. The decision fits into a broader pattern in which Beijing has reportedly been instructing domestic AI companies, including labs like Moonshot and Stepfun, to avoid taking US capital as it seeks tighter control over its local AI ecosystem.