What kind of automation projects should I start with on Upwork to build a credible portfolio using Claude code?
Go to Upwork.com, click 'Find Work', type 'automation', and apply to recent posts—raw dog the GH LX expert for full snapshot design, GH LX form and automation setup, or an AI automation engineer role (10K spent). Build a fully spec'd email triage AI drafting service (200K spent); there's money everywhere if you know where to look. Apply to many, maybe five, set an hourly alarm to reload the page for new posts. Don't worry about proposals with 50+ bids—80% are junk like 'Hello dear, I'm an automation developer, I'll help you.' They're not customized, have no value, and people glaze over at m‑dashes or AI junk. There's still room to boost your posts, which many overlook. (Side note: I've been testing clickbait vs. high‑value long‑form, trying different terminals, etc.—but that's unrelated.) My Upwork profile is solid: $1K+ earned, top‑rated, five‑star reviews, 100% job success. Competition is growing, and my theory rate feels lower despite following tips. If I were you, I'd aim for a 10K/month milestone with some upfront capital, focusing on input goals rather than output. Upwork is full of people making 10K/month with little upfront investment; the niche is exploding, with a million new niches monthly as tools evolve rapidly. You could become a Higgs field expert or learn Open Claw—even a mediocre tool can sell if you know how to position it. The idea that things are getting harder is false; our niche hasn't grown as fast as the number of people, so competition is actually easier now than before. If supply and demand were equal, competition would be the same; if demand is lower, competition is higher; if demand is higher, competition is lower. Anyway, Leto, son of Paul Atreides—thanks.
First, make sure your Upwork profile is fully optimized. Start with a clean, well‑framed profile picture – avoid cutting off half the face or having a badge that obscures it. Zoom out a bit when you upload so the whole head is visible. Next, set your status to "online for messages" so clients see you’re available; this can boost response rates by 30‑50%.
Your title is the first piece of text a client sees, so craft it to match what they’re searching for. If you’re an AI automation specialist, include that, and also list the platforms you’re proficient with (e.g., Make.com, Figma).
Add as many relevant skills as possible – aim for around 15 – and make sure they overlap with the keywords in the job description; this improves your SEO within Upwork’s search.
If you’re new and have no completed jobs, focus on getting your first couple of contracts quickly. Those early wins will make your profile stand out against other zero‑job freelancers.
Another quick hack is to create a mock job post as a client, watch how other freelancers pitch, and use that insight to shape your own proposals.
Overall, treat your profile like a landing page: professional photo, clear title, comprehensive skill list, and early completed jobs to build credibility. Follow these steps and you’ll start seeing more replies to your applications.
You're sending about four proposals per day on average. If you increased that to 10 per day (2.5× more), you'd see roughly 2.5× more results. Your main bottleneck is the view rate, currently around 12 %; improving your CV, profile, how it stands out, and your writing to avoid generic templates could push that to 20‑25 %. Make sure the first 150 characters show you read the post, give something free, and present yourself as a professional. With a 12 % view rate, 16 viewed proposals translate to about five replies, which is rough but not hopeless. In my early days I sent 20‑25 proposals a day; I would have completed your 30‑day volume in about five and a half days, outperforming you six‑fold on views alone. Even without improving my view rate, that would have yielded roughly 90 views, around 30 replies, and three or four calls—enough to eventually land a deal. You can make up for any shortfall in volume with strategic effort, something I've done since day one. I'd still recommend you do it more effectively, but the principle holds.
Are we talking no clients in general, in any of my businesses, or specifically automation? You might be talking about automation. The way I did automation was I just started out on Upwork — applied to like 10 to 20 Upwork jobs per day along with my business partner at the time, and the vast majority of them said no to us. We used a custom Loom-video-based approach: I'd record a Loom of me talking through the problem and send out a ton of those. I got really good at explaining automation problems consistently, and we scaled that automation agency pretty quickly to $15-20K a month. We languished around $20-40K a month for a while because my partner wanted to operate it as a software consultancy taking on custom projects with a big hired team, but that didn't really work — the engineers we hired could drag-and-drop in tools like Make.com but didn't understand the business side, so the automations they built were brittle. That's when I learned how difficult hiring is for something as complicated as a software/automation project. My partner and I eventually split up amicably — he wanted to focus on the consultancy side and building resellable equity, I didn't. I kept the business, focused on high-leverage applications of systems I'd already built, scaled up Upwork applications and cold email (reinvesting most of my income into cold email), and simplified the offer down to mostly cold email systems and one-click CRM templates. Around $35-45K/month I couldn't keep scaling that way, so I started pitching a fractional-COO model: I took a few smaller clients paying $3-4K/month and gave them massive 20-30 page proposals saying, 'I want $10K/month plus 7% of your business, and here's how I'll justify it by delivering $100K/month in value.' That worked for a couple of clients, my revenue shot up, and combined with old clients coming back that same month I hit $72K in revenue (not all of it was recurring).
Yeah, so for big builds, don't just do it all in one shot; avoid big all-encompassing systems because scope creep usually leads to a situation where you haven't clearly defined what you’ll do for the client, which is impossible to specify ahead of time, causing the client to expect something you didn’t agree to for payment, breeding resentment and harming the relationship. Instead, start with a small starter project that’s tightly scoped and agreed upon by both sides, then vibe each other out to learn what the client likes and demonstrate your aptitude—this lets you pitch more later and gives you insight into their business before committing to a massive scope. If you’re doing this for a company and need to send 10,000 depersonalized emails, think logically: using Make, first you get the leads (one op), then you iterate through the leads (n ops), generate icebreakers (another n ops), and optionally upload to Instantly (another n ops) or bulk‑upload to Google Sheets (effectively one op). Roughly, that’s about three n ops, or two n ops if you use the bulk‑upload shortcut. For 10,000 records that works out to roughly 20,000 operations, give or take. As for the Make plan, the $34‑per‑month Core plan gives you 40,000 ops, plenty of headroom; you don’t need the Pro plan unless you want better error handling, in which case the Pro plan is $62. You can start with Core and upgrade later, paying only the difference. On a personal note, I’m about twenty minutes into this video and feeling locked in; I’ve gotten more done today than in the past week because I’ve committed to stop daily posting on my main channel. My growth has slowed—my first‑24‑hour view‑to‑subscriber ratio dropped from about 15% (≈15,000 views on 100k subs) to around 5% (≈5,000 views). Comments suggest this is because my niche‑specific content has saturated its audience, and broadening the topic would attract less‑interested viewers. Elrico noted that my recent videos repeat earlier material, which may also be hurting growth, and suggested exploring other tools besides Make, like voice agents and live builds. Parker joked about throwing me on while cooking to start a riot over quitting daily, and shared ideas such as adding an automation bucket for publicly traded codes to save Coinbase money, using customer‑support headcount as an input for percentage assessments, and targeting big logos to pull in the right audience. My daily updates channel now has about 5,600 subscribers, while the main channel sits at 110,750. I’m checking follower‑count tools like Trend Hero, though I’m not sure about the daily‑limit metric. Reflecting on all this feedback, I’ve decided to stop posting daily and instead pursue a multifaceted strategy: focus on what makes me unique—my calm, low‑production‑value, high‑signal‑to‑noise conversational content—and develop a new, less‑replicable format. With my team, we’ve decided to pursue the “complete and utter uncopyable thing”: building a business with this stuff in front of everybody, doing it extraordinarily well by spending time generating leads, selling them, securing proposals, onboarding clients, fulfilling projects, and growing the company. Nobody else can replicate that live business‑building demonstration. My earlier live video of starting and selling an AI service in ten hours went viral precisely because it proved I can generate leads on camera; it’s a litmus test for knowing how to do this.