How to use ChatGPT's new Build-Your-Own Agent workflow

AI is often talked about as if it’s only good for low-level, repetitive tasks. But here’s the truth: it's also a game-changer for high-level execs: Preparing for meetings Summarizing contracts Fine-tuning your board updates Writing more persuasive comms Brainstorming ideas for growth strategies Analyzing pros/cons of business decisions Summarizing reports & teasing out relevant insights Getting alternate perspectives on high stakes decisions That's why I was shocked to hear one PE person I interviewed for the book say that he estimated 10 - 20% of portco CEOs were using a foundational tool (like ChatGPT) regularly.

I stopped checking my portfolio 17 times a day. Now ChatGPT does it for me - automatically. (download the prompt to follow along: https://lnkd.in/eB-Hrd8R) Here's what changed: With ChatGPT's automation tool, you can set up custom automations that help you save time. So I set up one that gives me a personalized market brief end of day. Every trading day at 4:15 PM ET, ChatGPT briefs me on my holdings. It pulls: → News that moves the needle → Filings that matter → Price action + volume spikes → Catalyst classification (positive/negative/neutral) Then it delivers one table. Only what impacts P&L. I went from reactive scrolling to proactive positioning. And I never miss what matters. P.S. Prompt to set this automation up for yourself here: https://lnkd.in/eB-Hrd8R P.P.S I write a weekly newsletter showing you how to invest better using AI www.davewang.ai

I had a weird experience this week. I was trying to fix a problem inside a software I use. Their customer support was “AI-powered.” They make it almost impossible to speak to a human. Just an AI chat box to ask my question. Nowhere in the software or on the site could I find how to reach a real person. The bot spit out a bunch of technical stuff that I didn't understand. So I copied its responses and fed them into ChatGPT. Then I copied ChatGPT’s response back into the company’s AI chat. And somehow, the two bots started solving the problem together. I just sat there watching them go back and forth. It reminded me of the old game Rock’em Sock’em Robots. The issue eventually got fixed. But I can’t say I was completely satisfied with the experience or how long it took. Have we reached the point where one AI is going to speak to another AI? Companies are using this technology to replace access to real people. AI is incredible, but IMO it’s not a substitute for an actual conversation with a human (certainly in some cases). Are we going to be overly reliant on AI? It seems we are so obsessed with automation and AI, maybe human access will be differentiator for customer experience. Has anyone else tried having two AIs talk to each other yet? And when do you still want to speak to an actual human?

Google made a small coding update last week that exposed a big truth about the AI tools market. Credit to Jack Porter's post where I first saw this! Google removed a line of code that allowed systems to pull 100 Google results at a time. It’s now limited to just 10. This change caused problems for a lot of companies whose products relied on scraping search results and feeding them into ChatGPT. Because they could pull 10x fewer results at a time - the costs of the service shot up. It exposed how much of what’s branded as AI or automation isn’t actually doing the work itself Many tools are just built on top of ChatGPT and fed through it to get the answers This ties into an objection I get from prospects all the time: “we’ve seen this AI stuff before. You're just selling us a ChatGPT bot" Which is why I make the difference clear. DWIGHT is automation, not AI. Without getting too techy - he's not actually artificially intelligent. One thing our customers take comfort in is knowing that we own the tech and build DWIGHT specifically for their processes. Our clients can trust that DWIGHT will handle the repetitive, time consuming back-office work so their teams can focus on what drives revenue. I'll drop a link to an article that helped me understand this below

🤌Executives who always default to one ChatGPT model waste time and money!Whereas, those who know WHEN ChatGPT-3.5 is enough & WHEN ChatGPT-4 is essential or WHEN ChatGPT-5 is a must → run smarter workflows! Why It Matters Model choice ≠ “just” technical Model choice = strategic decision ——— • GPT-3.5 → fast, cheap, good for lightweight tasks (lists, short drafts, quick recaps). • GPT-4 → slower but precise, ideal for critical documents, nuanced reasoning, or board-level communication. • GPT-5 → the cutting edge: multi-modal, advanced reasoning, must for high-stakes strategy and analysis. ——— ☹️The mistake? → Treating them as interchangeable. Over-reliance on one model either slows you down or risks errors at scale! EXAMPLES: ❌ Bad Prompt (No Model Awareness) “ChatGPT, write me a detailed business plan.” —— → Risk: if run on GPT-3.5, you’ll get a superficial draft; If always on GPT-4, you’ll waste time and tokens on tasks that don’t need it… ✅ Good Prompt (Model Choice Framing) “ChatGPT, draft a 1-page executive summary of this business plan. Task:speed > depth. Which model is best suited - GPT-3.5, GPT-4, or GPT-5?” —— → Clear expectation: the assistant tells you which model to use, before producing the output. 🦾 Advanced Prompt “ChatGPT, here’s my task: analyze customer churn data (50k rows), identify top 3 drivers, and generate recommendations for the board. Step 1: Tell me which model - 3.5, 4, or 5 is best suited and why. Step 2: Run the analysis using that model.” BOTTOMLINE⁉️ Leaders don’t just pick random tools, instead they pick the RIGHT tool for the RIGHT job P.S. 🤫🎥 If this caught your attention, you’ll love the full Episode →"ChatGPT as Your Personal Al Assistant I Automate Tasks & Boost Productivity” Check it out whenever you feel like: https://lnkd.in/dbHy5v7w I think you are going to love it! :)

“What AI skill should my team and I actually learn right now?” I will scream this from the rooftops of NYC. ➡️ Learn agent delegation Target a dedicated workflow or task. Assign an AI agent said role, define the outcome, set constraints, and schedule review gates. Treat it like a junior teammate and give it work, while monitoring so you can review for accuracy. Here’s my do-this-now stack, and how I’d run it with a team ⏬ If you’re a beginner: Start with ChatGPT Agent Mode. Open a new ChatGPT chat and change the dropdown to ‘Agent Mode’. It can plan tasks, execute steps, and return cited outputs for market scans, vendor comparisons, executive briefs, and decision memos. Kick off the job, let it run, WATCH IT RUN, and then review the completion. If you’re more technical or ops-heavy: Use Claude Code when the work requires operating UIs or your computer - clicking through portals, filling forms, wrangling spreadsheets, saving down documents. Expect more upfront setup and ownership, so keep a step-by-step prompt checklist, add automatic reruns for failing steps, and update the checklist only when the site’s labels or paths change. If you’re living in Google Workspace: Turn on Google connectors (Drive, Gmail, Calendar) inside ChatGPT or Claude. Ask the model to find your team’s file, summarize threads, compare document versions, prepare for and schedule meetings, or draft from past emails. This lets your agent pull context and act on it without manual hunting. How to turn this into outcomes in 30 days ⏬ → Twice a week, use Agent Mode to produce a one-page brief with citations and a recommendation on a real business question. Track cycle time and data/citation quality, and, where relevant, use Claude Code to automate in parallel. At the end of the month, you should know where a few agents can tackle real work and have the data to support what to scale. #AIinWork

AI isn’t just about saving a few minutes here and there—it’s about scaling how we work. I love Allie K. Miller’s take on treating AI agents like junior teammates: delegate, set expectations, review outputs, and track results. Her “twice-a-week brief” approach makes adoption practical and measurable—you can actually see ROI in 30 days. I’m especially interested in testing Agent Mode for executive briefs and Claude Code for workflow automation. 🚀 👉 Curious—if you could delegate one recurring task to an AI agent right now, what would it be?

I like the push toward “agent delegation,” but here’s the hard part I see in real teams: most people struggle to delegate to humans, so they also struggle to delegate to AI. If you hover and rewrite everything, you will call the agent “useless.” The fix is to set one clear outcome, a few rules, and a review time. Then let it work. Pick a daily, low-risk task. For example, “make my priority list from calendar and email, draft a 5-bullet summary, and flag anything with a deadline this week.” Turn on the Google connectors, run it, and check the result in two minutes. Approve or correct. Save the good prompt. After a week, move to a real workflow: vendor research, meeting prep, or weekly status reports. Keep an “escape hatch” rule: if the agent is unsure, it must ask or stop. That’s how you build trust and real output fast.

Have you noticed ChatGPT feels weaker lately? You’re not imagining it. It is absolutely been nerfed. It makes sense as a business decision why, but that's a different post. The headline here is how you deal with it... Commercial AI tools are changing quickly. Daily. Quality shifts. Features disappear. Pricing doubles. You don’t control it. That’s the risk. If you depend too heavily on one external tool for critical functions, you’re exposed. What to do? 👉 Own your data. Don’t let it live only inside someone else’s platform. 👉 Build systems where tools are modular and swappable. 👉 Use commercial AI for leverage, not as your foundation. 👉 Add custom layers... prompts, processes, and integrations you own... so the core doesn’t break when a vendor changes direction. Tools will keep evolving. That’s not the problem. The problem is depending on them so heavily that you can't operate without them or your business breaks when they do. There is a way to design resilient systems. There is a way to design bulletproof workflows where the tools are replaceable. 🧠 What’s one tool your business couldn’t live without today? And how would you handle it disappearing tomorrow? -- I’m Michael Trezza, AI automation expert & CEO at Lithyem. I help CEOs and founders eliminate their biggest bottlenecks with AI. If you're a founder or CEO who wants to: 🔥 Eliminate bottlenecks 🔥 Automate busywork 🔥 Streamline workflows 🚀 Follow me to see what’s possible when AI meets real business problems.

Lately, I’ve been thinking a lot about ChatGPT’s new agentic OS and what it means for the future of agentic commerce, meaning where AI agents buy, manage, and act on our behalf (its not as far as you think it is...) It’s exciting but at the same time, a little troubling. Because every time a big shift like this happens, there’s a wave of cheap wrappers that flood the space, shiny “AI agents” that overpromise, underdeliver, and slowly erode user trust. That had me questioning my own direction for QuietSignals. A few days ago, I was in a moment of indecision, trying to figure out how to position the product in a market that’s evolving faster than you blink. During that time, I came across a few ideas that helped me find some clarity. They started as notes to myself, but maybe they’ll help other founders navigating this same wave 👇 1️⃣ Own the problem, not the interface. Don’t build another chat wrapper. Build something people can’t replace, a workflow or decision layer that’s painful to live without. 2️⃣ Build feedback loops. Agents that don’t learn get dumb fast. The best moat is learning faster than everyone else. 3️⃣ Earn trust like a human. In a flood of AI tools, most will break trust before they build it. The ones that win will make people feel confident their product actually listens and does what it says. In no way am I an expert on this, and it remains to be seen how effectively I’ll execute these ideas myself. But that reflection helped me see where QuietSignals fits: in helping businesses listen at scale before they act. The next few months will be rollercoaster But for now, im getting some fresh air and exercise!