AI workflows

Every answer 41 answers

• Can AI do tasks for me in software that has no API?

  Yes. Browser agents now click through software for you. With Codex's replay feature you show it a multi-step workflow once in its built-in browser, name it, then call it by name with new details. Claude's Chrome extension can dig through settings and fix things. My rule: if a digital task isn't natural to you, give it to the AI.

• Can AI fix software bugs automatically?

  Partly, yes. My error tracker sends alerts to Slack, and an agent on a separate computer picks them up and patches the bug, often without me knowing. Teammates can drop a screen-recording link in a bugs channel too. It isn't perfect, it only works if errors are visible, and complex software keeps producing new bugs.

• Can AI pick the best stories for my newsletter?

  Yes, and selection is a better use of AI than writing. Export your click report with the link for each story, give it to ChatGPT or Claude with your candidate stories, and ask which ones to include based on what readers actually clicked. A click means little until the model can see the content behind it.

• Can AI write and run a local newsletter?

  Mostly, yes. I got a local newsletter to roughly 97% quality in about an hour with Claude's multi-agent workflows, and my working rule is 90% AI, 10% you. The hard part is sourcing local news, not writing. A person still edits, adds local knowledge and reads it before it sends.

• Can AI write local news from city council meetings and permits?

  Yes, and it's ready now. You can do it by hand by pasting a council transcript into a Claude project, or let larkfeed watch council meetings and permits and draft the stories overnight. It has adapters for the seven main platforms that host council content in the US. The person becomes the editor: review, edit or hit publish.

• Can Claude use my software for me, like Beehiiv, Canva or ClickUp?

  Yes, through connectors. Connect a tool once, then ask in plain speech: design four Instagram posts in Canva, create a ClickUp task, tell me how many clicks my last issue got. When I checked in March 2026, Beehiiv's own connector (MCP server) was read-only: data and reports, no changes. Assume read-only until a write path is proven.

• Can I build a website for a local business with AI?

  Yes, and fast. I built one for one of my first advertisers in under an hour by chatting with v0, and hosting about 20 small sites there runs about $20 total. Copy a design you like, expect a second round of fixes, and test anything that talks to customers. A chatbot I added made up a price.

• Claude vs ChatGPT vs Codex: which AI should I use?

  Use more than one, but pick one per job and stick with it. In mid-2026 I told companies ChatGPT was ahead for general use, yet consistency beats chasing the leader. For coding I lean on Codex because it works longer on its own and asks me for less, and on Claude for design-heavy builds and Notion. One AI coding tool per project.

• Do I still need Zapier, Make or n8n now that Claude Code and Codex exist?

  You don't need them to build automations anymore, but they can help keep them running. In May 2026 I said skip them and just describe the workflow to Claude Code or Codex. By July my from-scratch workflows kept breaking, so now Codex builds them inside n8n as a sturdy frame. I never open n8n myself.

• How can a non-technical founder build software with AI?

  Talk to the coding agent in plain English, let it interview you before it builds, and demand proof it works live. I can't code. I built three software products over about 18 months by voice with Claude Code and Codex, spending $600 to $800 a month on AI. Start with the $20 plan and ignore its time estimates.

• How do I automate the whole sales process after a call with AI?

  Let an agent work from the call transcript. After a sales call, mine reads the transcript, builds a client-specific proposal website, creates a Stripe payment link at the price we discussed, and once paid sends onboarding and an intake form. The goal is that everything comes to me in Slack for a one-click review, so I can take more calls.

• How do I build a personal daily AI news digest that's actually useful?

  Tell the AI who you are and what you work on, so it gives direct advice instead of a generic roundup. Then lock the format: two to four stories per section, smart brevity style (what happened, why it matters, your move), one image up top, items ranked by urgency, a 60-second version and a like or dislike button.

• How do I cut AI costs when using Claude or Codex all day?

  Run agents on flat-rate subscriptions, not pay-per-use keys. Use the top model for planning and a cheaper one for doing, start cheap and escalate only when it fails, and cap any API spend. On a $20 plan, default to Sonnet for simple jobs. The $200 plans hide thousands of dollars of API-equivalent usage.

• How do I get AI agents to give me updates I can actually read?

  Give every agent one fixed format: in plain English, what happened, why, the impact, whether it's fixed, what's next and its recommendation. Use markdown with titles, bullets and one sentence per line. Send briefs meant only for you to a private channel, and add one top briefing for days when you only have three minutes.

• How do I give AI context about my business without re-explaining it every time?

  Have the AI interview you once. Ask Claude or ChatGPT to ask you about 40 questions about you, your business, your processes and workflows, then turn the answers into a context document. Put that file in your project folders or upload it to any chat. For more than one brand, keep a separate business bible for each.

• How do I know an AI coding agent really fixed something?

  Make it prove it. I don't accept done, fixed or live without evidence: screenshots of the agent using the feature like a human, the live link, a real end-to-end run. AI agents regularly say they tested things they didn't, patch symptoms instead of causes, and call a deploy done before the job ran. A merged badge proves nothing.

• How do I set up alerts for my automations without learning to ignore them?

  Alert only when something is truly broken, and make silent failures impossible. My rule: either the system works and I hear nothing, or it's broken and I need to know. Four or five bug emails while the site was up taught me to ignore them. A daily weekday check catches the quiet failures.

• How do I turn a YouTube video or podcast into a newsletter with AI?

  Paste the YouTube link into Gemini, which reads the video itself, and run a prompt you had AI write once and reuse every time. Ask for a short overview, the top three takeaways and an action step. I built one on my phone in about five minutes. For podcasts, feed transcripts to a custom GPT.

• How do I turn my call recordings into content automatically with AI?

  Connect your call recorder to an AI model with instructions about who you are and how you write, and have it draft content from every transcript: a blog post, four LinkedIn posts and three tweets, delivered to Notion or Slack. It isn't advanced AI. The win is putting it on autopilot so it happens after every call.

• How do I use ChatGPT deep research to find content for a local newsletter?

  Don't ask deep research directly. Have AI write a detailed prompt first: your towns, a default date window, five news stories, five restaurants, three local businesses, and two or three sentence summaries in your style. Keep it in a doc and rerun it weekly. Treat the output as a starting point. It shines on restaurants and businesses.

• How do you get AI agents to do useful work without being prompted?

  Treat each agent like an employee and give it the access to act. My bar is one valuable thing a day per agent without me asking: build something, redo something, or propose an idea and do it when I say yes. A report alone doesn't count. The biggest blocker is logins and permissions.

• How do you run AI agents inside Slack with a team?

  Treat Slack as the agents' office. Each agent gets a role, teammates message it in shared channels, and it acts on what it hears and reports back in the thread. My rules: replies in threads, three roundups a day instead of constant pings, no agent-to-agent DMs, plain-English recaps. Expect setup pain.

• How do you stop AI from producing generic slop?

  Don't let it freestyle. Give it an approved system to follow: a locked design system for visuals, decided copy before any design, a banned-words list and real local knowledge for writing, and a reviewer that checks before you see it. Make it look at the rendered result, not the code. AI defaults are where slop comes from.

• How should I set up an AI second brain for my business?

  Keep one source of truth, split only into business and personal. Make it plain markdown files your agents read, feed it automatically from call transcripts and AI chats, and only show agents the curated, high-value part. I built mine five times before realizing almost all of it was company knowledge everyone can share.

• How should a complete beginner start learning AI?

  Pay for Claude or ChatGPT, then make AI your first stop for every digital task: before you do it yourself, ask how AI can help. Learn by doing on something you actually care about, and go to YouTube only when you're stuck. You're not too late. Even if you're behind, you can get ahead in about two weeks.

• Should an AI agent start with read-only access?

  Start read-only where an agent can wreck real data, then widen fast. One AI-assisted job wrote 725,000 rows into production because test and live weren't separated. But after three months of agents, the bigger problem has been agents blocked from obvious work. My rule: they can draft and edit almost anything; only a human hits publish.

• Should an AI agent update my CRM and track client payments?

  Yes, but have it propose changes before it writes them. When we checked my inbox for leads, only 3 of 35 candidates were real, so my agent now asks in a DM before adding anyone. For payments, it reads Stripe, sees who's on autopay and on what date, and tells me the day after a payment is missed.

• Should an AI reviewer check work before a human sees it?

  Yes. Nothing from my agents should reach me until a separate reviewer agent checks it for design, language, facts, links and banned words. The reviewer knows the goal, not the code. Cap it at two or three rounds. If it still fails, it comes to me with what's wrong. A person still catches what the agents miss.

• Should you plan with one AI and build with another?

  Yes. I use a planning chat as my product manager and a coding agent as the developer. One model writes the plan and a second model redlines it, because the first one always misses things. Cheaper tools can do the building if a smart model reviews each step. Chats forget, so the plan has to land in files.

• Should you send an AI-written newsletter without reading it?

  No, not until it consistently earns it. My rule today is a human read before every send: check it fits your content, get the facts and dates right and test the links. AI does about 90% and a person does the last 10%. I'd only skip the person if an automatic reviewer scored issues 9 out of 10, and early tests scored 5 and 6.

• Should you talk or type to AI tools like ChatGPT and Claude?

  Talk. Speaking is at least three times faster than typing by my estimate, and the bigger win is that you stop self-censoring, so the model gets more context. I use Wispr Flow for dictation, don't type more than a sentence anymore, and code, write articles and brainstorm by voice. Type only exact strings like URLs, numbers and code.

• What AI tools should a non-technical business owner pay for first?

  Three paid tools and one free account: Claude at $20 a month, ChatGPT at $20 a month, a voice dictation tool at about $15 a month, and a free GitHub account. If you're not paying, you don't see what's possible, because the free tiers don't give you access to it.

• What are multi-agent AI workflows, and are they better than one agent?

  A multi-agent workflow has a coordinator agent hand pieces of a job to specialist sub-agents running at the same time, then review and assemble the results. I got a local newsletter to roughly 97% quality in about an hour this way. The risks are agents approving each other in loops, and trying to build a whole team at once.

• What is meta-prompting, and why does it improve AI output?

  Meta-prompting means you never prompt AI directly for anything serious. You ask the AI to write the prompt first, have it score and improve that prompt, then run it. I use it for deep research, app builds and anything that matters. Paired with voice dictation, it turns three rambling sentences into far richer instructions.

• What research tools give AI agents better results than plain Google?

  Give agents AI-native search and scraping tools, and keep them on free tiers until they prove out. My growth agent has Exa for search and Firecrawl for scraping, plus a prepaid $50 keyword-data API. For people: Perplexity when you want an answer or real sources, Google when you want to find a specific thing.

• What should go in AI agent instruction files like CLAUDE.md and AGENTS.md?

  Short, clear rules that stop the mistakes you keep seeing, not a growing pile of restrictions. I spent about 40 hours over four months tuning my CLAUDE.md, AGENTS.md and skill files, mostly to stop AI tools inventing problems. My rule of thumb: fewer rules, no conflicts, and delete before you add.

• What should you automate with AI first in a newsletter business?

  The back end first: the grunt work and annoying admin tasks, not the writing. Most of it isn't hard, just time consuming, so it never gets done. My test: if a reader or advertiser will experience the output, keep a human in the loop. If it's internal, automate it. Start with recurring steps that add up across every brand.

• Which AI app builder should a beginner start with: Replit, Lovable, v0 or Claude Code?

  Start with a hosted builder like Replit, Lovable or v0, then graduate to Claude Code or Codex. Claude Code is level three, four or five, and I couldn't have gone from nothing straight to it. For front-end designs, v0 won about 90% of my side-by-side tests. Getting to 95% is easy. The last 5% is hard.

• Why do AI coding agents write too much code, and how do you fix it?

  Because agents add rules, restrictions and code far more readily than they delete. My team's custom agent system grew to 70,000 lines duplicating an open-source tool that already worked, which explained the random failures. Told to start over from zero, the coding agent cut less than 1% in two days. You have to insist on deletion.

• Why does ChatGPT always agree with me, and how do I get honest feedback from it?

  It defaults to agreeing, even when you tell it not to, so never treat its yes as validation. Ask it to review its own answer, have it score ideas and suggest better ones, tell it its job is to improve your thinking, and ask the pointed questions it won't raise on its own. Then make the call yourself.

• Why does ChatGPT or Claude forget my work, and how do I stop it?

  Regular chats don't reliably save things, and long chats run out of context and start claiming work they didn't do. I lost a four-hour website build that way. For anything you're building, work in tools that read and write folders on your computer, like Claude Cowork, Claude Code or Codex, so the work lives in files.