What 1,031 Upwork postings say about paid automation work right now
Titles and tags from roughly the last two weeks. The busiest lane is also the cheapest.
Titles and tags from roughly the last two weeks. The busiest lane is also the cheapest.
I went through 1,031 Upwork postings, titles and tags only, covering roughly the last two weeks. This is what people are actually hiring, how they want to pay, and which clients sit behind the postings.
A posting can land in more than one bucket. A GoHighLevel plus n8n plus voice agent build is common, not rare.
The top ten, ranked by volume
GoHighLevel / CRM automation: 216 postings (21%). The single biggest category by a wide margin. Funnel builds, sub-account setup, pipeline automation, syncing CRM data with other tools.
AI agents and assistants (agentic workflows): 155 postings (15%). Custom agent builds, often multi-tool (LangChain, LangGraph, RAG pipelines). Not just simple bots.
n8n workflow automation: 133 postings (13%). New builds, and people paying to fix or extend flows they already have.
AI chatbots (customer-facing): 93 postings (9%). Mostly new builds for WhatsApp, websites, or support.
Zapier / Make.com: 77 postings (7%). Usually simpler point-to-point app connections than n8n or agent work.
Lead generation and outreach: 48 postings (5%). Cold email and SMS systems, prospecting pipelines, GTM automation.
Appointment scheduling and booking: 31 postings (3%). Often bundled with a chatbot or a voice agent in front of it.
Social media content automation and scheduling: 29 postings (3%).
Voice AI / call automation: 27 postings (3%). Retell and VAPI-style voice agents and call handling. Small, distinct, and it sounds current.
Web scraping / data-entry automation: 27 postings (3%). Tied with voice.
The top three (GoHighLevel, AI agents, and n8n) account for roughly half of everything in the scrape. That is where the weight of current demand sits.
Voice AI is the one I would watch. Smaller in volume, but specific. Worth watching if it keeps climbing.
Chatbot and AI-agent work are really one blurred lane, not two. 67 postings match both. That is by far the largest overlap of any pair, more than double the next one (n8n plus AI-agent at 41). A lot of "chatbot" work is "build me an AI agent that happens to be conversational," and the other way around. If you ever planned to specialize in just one of those labels, in practice they are one continuous market.
That shows up in the search terms too. Of the eight terms used to pull this set, "ai_agent" is only 6% exclusive. Almost everything it finds also shows up under "chatbot," "ai_assistant," "n8n," or "automation." Compare that to "mobile_app" (95% exclusive, its own island) and to "gohighlevel" and "n8n" (55% exclusive each, largely their own territory). AI-agent is the label that bleeds into the others. Mobile app, GoHighLevel, and n8n are the ones that hold their own ground.
High volume is not high pay
GoHighLevel is the busiest lane and also the cheapest one. The median fixed price for GHL work is $50. The average hourly midpoint is $21. Both are the lowest of the five lanes we priced.
Mobile app development pays the best on both counts: median fixed $200, average hourly $29. Chatbot work is close behind on hourly, at $28.
Read that next to GHL sitting at number one in raw volume. The biggest, easiest-to-find lane is also the most commoditized one. High demand here does not mean high pay. It means more competition for lower-ticket work.
"Senior" and "Expert" language is common, and it pays a bit more. 18% of all postings use that framing in the title. Just 1% signal junior or entry-level. Jobs with "Senior" or "Expert" in the title average a $28/hr midpoint versus $24/hr overall. That is a real premium, and a modest one. The bigger point is the near-total absence of beginner framing. This market talks to experienced freelancers by default.
The $5 to $20 fixed-price jobs are not lowballs. 70 postings sit at $20 or under. 86% of those come from a client who already passes a strong-client bar: verified, a good rating, real spend history. One has $20K+ spent, another $10K+, another $6K+. Titles like "CRM Automation Expert Needed" at $12, or "Fix N8N Workflow Issue" at $5, from proven clients, are almost certainly milestone or placeholder pricing (a $5 first milestone that gets renegotiated once work starts). A flat price cutoff would throw those out for the wrong reason. Client quality first, then the number on the post.
The highest-paying individual postings are wildly inconsistent on that same quality bar. Six of the ten highest fixed-price jobs ($24K, $20K, $10K, $9K, $7K, $4.5K) come from clients with a rating of 0 and $0 spent. The biggest headline numbers are disproportionately unproven clients, not the safest bets.
The standout exception is a $265/hr posting: "Build a Claude tool that enforces a law firm's drafting playbook." Verified, 5.0 rating, real spend history, and a Claude-specific ask. That one row is a good argument for not chasing headline dollar figures without running them through the same client-quality check as everything else.
Claude is showing up by name. "AI" is in a third of the titles.
Claude is a live keyword, not a rounding error. 37 postings (3.6%) name Claude. That almost matches the 31 postings (3.0%) that name GPT, ChatGPT, or OpenAI.
Some of the titles are specific: "Claude Code Developer for GoHighLevel CRM AI Agents." "Claude + Wave + Dropbox Automation." "Claude & n8n Automation Expert for GoHighLevel CRM System."
Clients are asking for Claude-based builds, often paired with the same lanes already in this list (GoHighLevel, n8n). Positioning around Claude is not a niche against the market. In explicit client demand right now, it sits roughly at parity with GPT.
"AI" itself is in 32% of all 1,031 titles. Claude and GPT together are only named explicitly in 6.6% of postings. Most clients are asking for AI in the generic sense. They want something AI-branded, not a specific model or vendor. You do not need to lead with a specific tool to match what the title is asking for.
n8n connects lanes. GoHighLevel sits in one.
Looking at tags that travel together, n8n postings co-mention "ai agent development," "api integration," and "zapier" most often. It is the tool clients reach for when they need to stitch other things together.
GoHighLevel postings co-mention "CRM automation," "marketing automation," and "customer relationship management." That is a narrower, agency and marketing cluster.
Practically: n8n skills transfer across almost every other lane. GHL skills are more self-contained.
Zapier still beats Make.com more than two to one in explicit mentions (55 versus 26), with almost no overlap. Only 4 postings name both. They are competing for largely separate audiences, not interchangeable asks.
When an automation job asks for a coding language at all beyond the no-code platforms, Python is the default. It shows up in 5% of postings, ahead of JavaScript at 3% and PHP at 2%.
Outside the five defined lanes and Zapier/Make, almost no other tool gets real traction. Salesforce, Monday.com, ClickUp, Slack, and Stripe are each under 1%. WordPress (3%) and Airtable and Shopify (1% each) are the only other tools showing up at all. There is not a hidden sixth lane being missed. The five lanes plus the automation platforms cover what is being asked for.
How they want to pay tells you how well the work is scoped
AI agent and assistant work is 73% hourly. Clients do not trust themselves to scope an agent build as a fixed deliverable. They want to pay for iteration.
n8n is the most balanced lane: 53% hourly, 47% fixed. About as often it is "build me this specific flow" as it is "be my ongoing automation person."
Almost a quarter of the whole market is a real long-term retainer. 232 of 1,031 postings (23%) are "More than 6 months." That is large enough to treat as its own track, not leftover noise.
Mobile app postings churn the fastest. n8n postings sit the longest. 20% of mobile app postings were scraped within hours of being posted, versus 2% for n8n. That is a proxy for competition speed. Mobile app posts get seen, and presumably applied to, almost immediately. n8n postings sit around longer before they show up as fresh, which could mean less competition per posting, or a slower-moving buyer pool. Either way, if speed-to-apply matters, mobile app is the lane where being fast actually matters most.
The clients behind the postings
GoHighLevel has by far the most reliable clients of any lane. 95% are payment-verified. Only 24% are $0-spend accounts. Both numbers are noticeably better than the other four lanes.
Compare that to mobile app (68% verified, 61% $0-spend) or chatbot (73% verified). GHL pays the least. It is also the lane where you are least likely to run into a flaky or brand-new client. Lower ticket, lower risk. Mobile app and chatbot pay more, and you are rolling the dice on the client more often.
AI agent and assistant work is where the enterprise-scale clients show up. 7% of those postings come from clients with $100K+ lifetime Upwork spend. That is more than double any other lane. n8n and GoHighLevel sit at 2%. Mobile app sits at 5%.
It is also the lane most likely to be a genuine long-term retainer: 26% of AI agent postings are 6+ month engagements, roughly double every other lane. This is the place where landing one good client and turning it into an ongoing account is most plausible. It is also the mixed pool, with the lowest verified rate of client quality overall among these lanes, so screening matters most here.
Client spend is very top-heavy for n8n and mobile app. Both have 44%+ $0-spend clients (n8n 44%, mobile app 61%). Close to half, or more, of the postings in those two lanes come from accounts with zero proven willingness to pay. Highest of the five lanes. Worth keeping in mind before you treat every posting in those lanes as a real job.
Of the 227 postings that fit a lane but fail a client-quality check, 43% fail because the client is not payment-verified at all. That is the hard no. 57% are verified but have zero spend and zero rating history. That is the soft no: a real account, just unproven. The second group is the majority of what gets screened out for client quality, not a rare edge case. An unusually strong job post can still be worth a look there. Most of the time, the screen is doing the work.
Ratings, when they exist, are heavily grade-inflated at the top. Of the 582 postings where a client has a rating at all, 73% sit at 4.9 to 5.0, and 92% are 4.5 or above. Only 2% are below 4.0. "Has a rating" is nearly the same signal as "has a great rating." The real dividing line is rated versus unrated (about 44% have no history), not good-rated versus badly-rated.
True whale clients are vanishingly rare. Only 3 of 1,031 postings come from a $1M+ lifetime-spend client. The finding that AI-agent work has the most $100K+ clients still holds. Even that tier is a small minority. This market is small-to-mid clients, not big accounts.
Two industries show up by name
Scanning titles for named verticals, real estate (26 postings) and e-commerce / retail (24) are the two clear leaders. Well ahead of healthcare (13), fitness (10), and legal (4).
Only about 11% of postings (114 of 1,031) name an industry at all, so this is a minority signal. Real estate and e-commerce work is common and specific enough (CRM sync, lead routing, inventory, POS-to-CRM integration) that it is worth searching on its own, rather than folding it into a generic sweep.
WhatsApp is quieter than the anecdotes
Only 14 postings (1.4%) mention WhatsApp. Just 4 mention Telegram. If WhatsApp automation feels like a hot niche right now, that impression is not showing up at volume in this set. It may be real and under-indexed by these particular 8 search terms. Or it may not be as large a slice of current demand as it seems.
Economy-wide, new-build language outweighs fix-or-rescue language nearly 10 to 1. 28% of all titles read as "build." Only 3% read as "fix" or "rescue." That holds across every lane, not just chatbot and AI-assistant. This is a make-something-new market, not a repair-something-broken one.
This is as far as titles, tags, budget, duration, rating, and spend can take you. What is actually in scope on a given job would need the full descriptions, which this scrape did not capture.
What I would do with this
If you want paid automation work right now, the volume sits in GoHighLevel, AI agents, and n8n.
If you want the work that is easiest to find and least likely to flake, that is GoHighLevel. You will compete on price.
If you want a shot at one client that turns into a long account, that is AI agents. You will need to screen. Chatbot work is mostly the same market.
If you want skills that travel, that is n8n. Those postings also sit longer. Mobile app is its own island, and it moves fast.
This is mostly new-build work. A rating, when it exists, is almost always a good one. The filter that matters is whether the client has any history at all. Whales are not the market. Three postings in a thousand come from a $1M+ account.
Treat "Senior" or "Expert" in the title as the default language of this market, not a special filter. Do not throw out a $5 or $12 post from a proven client. Do not chase a $24K post from a $0 account.
Voice AI is still small. I would keep an eye on it.