← Back to Hub research

What to automate on social. And what to keep human.

The fastest way to look fake on social is to automate the wrong things. The fastest way to compound your time is to automate the right things. Here's where the line goes for an owner-operator running their own feed.

Most owner-operators land in one of two ditches. Ditch one: you post sporadically, watch reach collapse the week you go quiet, and conclude social does not work for your business. Ditch two: you buy a tool that auto-generates 30 posts, schedules them across five channels, and auto-replies to comments. Six weeks later your feed reads like a marketing intern's audition tape and your DM box is full of people thanking a bot for thanking them.

There is a middle path. The owners who actually win on social automate the mechanical parts (drafting from raw notes, scheduling, repurposing one piece into many) and keep humans on the parts where a human voice is the entire product (replies, DMs, hot-takes, comments on other people's posts). That split is the thesis. Everything below is what it looks like in practice.

Automate the mechanical parts

Four jobs are safe to hand off. They are tedious, they follow rules, and the customer never sees the seam.

Drafting from your raw notes. You already think out loud in voice memos, Slack, your newsletter, customer calls. A drafting workflow takes those raw inputs and produces channel-native first drafts. Tools that do this well right now: Typefully (X-first, clean threads), Hypefury (X plus repurposing), and any custom workflow built on Make or Zapier that pipes a Notion page or a transcript through an LLM and into a draft queue. The output is a draft, not a publish. You read it, you edit it, you ship it. The hour you saved was the hour staring at a blank box.

Scheduling. Posting at 9:14 a.m. on a Tuesday when you remember beats posting nothing, but it loses to posting at the time your audience actually reads. Buffer, Later, and Hootsuite all do this competently. Pick the cheapest one that supports your channels. Buffer's own data across 52 million posts shows businesses posting fewer than three times per week per platform see a 52 percent drop in reach versus those posting five or more times¹. Scheduling is what makes five-per-week sustainable for one person.

Repurposing one piece into eight. You wrote a newsletter Sunday. That newsletter has a thesis, three sub-points, a quotable line, a chart, a closing. That is six to eight posts across X, LinkedIn, and an Instagram carousel without writing a new word. Hypefury, Typefully, and Buffer all do some version of this; Make and Zapier let you build a custom flow if your source format is unusual. The repurposing rule: each piece must stand alone. A LinkedIn post that says "thread on X" is a tax on your reader and a vote against your own feed.

Reposting evergreen content. Your best post from nine months ago is unread by 90 percent of your current audience. Queuing it to recirculate every three to six months is free reach. Hypefury was built around this; Buffer added it later. Cap the queue at 20 to 30 truly evergreen pieces so the feed does not feel like a rerun channel.

Keep humans on the conversational parts

Three jobs you do not automate. Not because the tech can't, but because the moment a customer realizes a bot replied to them, the trust you built across 200 posts evaporates in one screenshot.

Replies to comments on your own posts. When someone takes 30 seconds to comment, they are giving you a free conversation. A two-line human reply is the highest-leverage marketing minute in your week. An LLM reply, even a good one, signals you do not care enough to type. Read it, reply once, move on.

DMs. Same logic, higher stakes. A DM is usually a buying signal or a complaint, and both deserve a human read in the first response. Auto-responders that say "thanks for reaching out, we'll get back within 24 hours" are tolerable for a Fortune 500. For a two-person business they read as theater.

Comments on other people's posts. This is the single biggest distribution lever for an owner-operator on LinkedIn or X, and it is the one most often outsourced to bots. Do not. A thoughtful 40-word comment on a relevant founder's post puts your name in front of their audience. A generic "great post!" from your account, especially if it lands at 3 a.m. local time, is worse than silence. There is also a hard line here: LinkedIn does not expose a third-party API for commenting on other users' posts at all. Every tool that claims to (Taplio, Aware, AuthoredUp's automation mode) is running browser automation against LinkedIn's terms of service, and accounts do get restricted².

The X tier reality

If you are going to ship anything custom on X, the pricing changed in early 2026 and the math is now ugly for small accounts. The legacy Free tier capped writes at roughly 17 posts per 24 hours at the user level, shared across every app using your token. On February 6, 2026, X moved new developers to pay-per-use as the default; the old $200-a-month Basic tier and $5,000-a-month Pro tier are no longer available to new sign-ups³. If you got grandfathered onto Basic, you have 100 posts a day and 50,000 a month at the app level. If you did not, you are now paying per call.

The practical takeaway for an owner running their own feed: stay inside a scheduling tool like Buffer or Typefully that already holds an enterprise relationship with X. You do not need your own API key. The volume caps on those tools are well above what one human can write anyway. Only build direct against the X API if you have a specific reason (custom queue, internal approval flow, multi-account orchestration) and price the cost into the project before you start.

The LinkedIn API reality

LinkedIn is the opposite problem. Posting to your own feed via the official API is fine and supported through the w_member_social permission, which is what Buffer, Hootsuite, and any custom integration uses. The token expires every 60 days and there is no refresh token in most setups, so plan for a calendar reminder to re-auth. What does not exist: any sanctioned third-party path to comment on other people's posts at scale. The Comments API only lets a consenting member act on their own behalf, with daily call caps and short data-retention windows, and lead-generation automation is explicitly out of policy². If commenting on others' posts is your growth lever (it should be, for B2B), the answer is to do it yourself in 15 minutes a day, not to find a tool.

What the workflow actually looks like

Here is the shape that works for a one or two-person team. It is not exotic, it is not novel, and it ships.

Source layer. One Notion page, one Google Doc, or one newsletter per week as the raw thinking. Everything downstream pulls from this. If you do not have a source layer, automation has nothing to chew on and you get the bland posts.

Draft layer. A flow (Make, Zapier, or a custom LLM call) takes the source and produces channel-native drafts: one X long-tweet, one LinkedIn post, one carousel outline. Drafts land in a queue, not a published feed.

Approval layer. The drafts land somewhere you actually look. For us, that is a Telegram bot that pings the owner with the draft and two buttons: ship or kill. For others, it is a Buffer queue with a daily 9 a.m. email. The principle is the same. One human, one minute per post, one approval. Nothing publishes that an owner did not green-light.

Publish layer. Whatever tool holds the API relationship: Buffer, Typefully, Hypefury, or a custom publisher. Stagger send times by 15 minutes to dodge race conditions if you are running multiple channels off the same draft.

Conversation layer. Human. Twenty minutes a day, split morning and afternoon, for replies, DMs, and outbound comments on other people's posts. Put it on your calendar. The drafting and scheduling layers exist precisely to free this time.

Key takeaways

• Automate the mechanical parts: drafting from raw notes, scheduling, repurposing one piece into many, recycling evergreens. Tools: Buffer, Typefully, Hypefury, Make, Zapier, Hootsuite.
• Keep humans on the conversational parts: replies to your posts, DMs, comments on other people's posts.
• X moved to pay-per-use for new developers in February 2026. Stay inside a scheduling tool unless you have a specific reason to hold your own key.
• LinkedIn allows posting to your own feed via the official API. There is no sanctioned API for commenting on other people's posts. Do that by hand.
• Posting fewer than three times per week per platform costs you about half your reach versus posting five or more (Buffer, 52M posts analyzed).
• The workflow that works for one or two people: source layer, draft layer, approval layer, publish layer, human conversation layer. Approval is one button, not one rewrite.

Sources

1. Buffer. "The State of Social Media Engagement in 2026" (52M+ posts analyzed across Instagram, TikTok, LinkedIn, X, Threads, Facebook, Bluesky; January 2024 to December 2025). buffer.com/resources/state-of-social-media-engagement-2026
2. LinkedIn / Microsoft Learn. "Comments API" and API Terms of Use (consenting-member only, daily call caps, no lead-generation automation, no third-party comment on others' posts). learn.microsoft.com/en-us/linkedin/marketing/community-management/shares/comments-api
3. X Developer Platform. 2026 pricing migration: Free and tiered plans replaced with pay-per-use for new developers as of February 6, 2026; legacy Basic ($200/mo) and Pro ($5,000/mo) closed to new sign-ups. developer.x.com/en/portal/products