One Web Designer Uses AI Now and Then, the Other Has Built It Into His Day: Eric Karkovack and Pravin Kumar
Two web designers who run their businesses alone take very different approaches to AI. A WordPress designer with more than 25 years of experience uses ChatGPT occasionally to get past coding problems, and still writes every word himself. A Webflow developer has wired Claude into his planning, publishing and site audits, cutting the time to publish a blog post from two hours to about 45 minutes. Both keep some parts of their work firmly in their own hands.
Founders in this article
- Eric KarkovackUnited States
- Pravin KumarUnited States
Building a website for a client used to mean either knowing every piece of code yourself or paying someone who did. For a web designer working alone, the parts outside their strengths, a tricky function, a custom interaction, a stack of blog posts waiting to be published, could stall a project or eat into their evenings.
Two web designers who run their businesses on their own show two different ways of using AI to close that gap. One uses it occasionally, as a patient helper when he is stuck. The other has built it into almost every part of his working day.
This article is for web designers, developers and other freelancers who build things for clients. We look at what each of them hands to AI, where each draws the line, and then, in a separate section, at what their experience still has to do.
Eric Karkovack: a patient helper, used sparingly
Eric Karkovack is a web designer and WordPress specialist; his website says he has more than 25 years of experience. His business, Eric Karkovack Web Design Services, LLC, is a Pennsylvania company, and his website describes the work entirely in the first person. A 2020 WordCamp speaker profile says he lives in Carlisle, Pennsylvania. He also writes about web design for the design publication Speckyboy, where he described how AI changed his workflow in an article first published in January 2025. In that article, he describes himself as working solo and used to doing everything himself.
How he uses AI
- Getting unstuck on code. He describes himself as not the world's best coder. When a piece of PHP or JavaScript will not work, he pastes it into ChatGPT, which explains the error and offers a corrected version, instead of searching the web for answers.
- Small technical tasks. He uses AI to write WordPress functions or JavaScript that manipulates a page, tasks that used to hold him back when he had neither the time nor the budget to hire an expert.
What it changed
- A colleague when working alone. He writes that for freelancers who work solo, having AI available is like having a colleague on hand around the clock, even when its answers are not fully accurate.
- More time for quality. With less time spent debugging, he can think about how a whole project should look and work, add details that improve the experience for users, and spend more time on quality control.
- More confidence. He says AI makes him feel he can take on projects he might have turned down before, and that he worries less when a client asks for something he has not done before.
Where he draws the line
He does not use AI every day. Most of his design and development work comes from his own thinking, and all of his writing is his own. He also writes that people are still much better than machines at planning and critical thinking. He feels guilty about contributing to falling traffic to Stack Overflow and similar communities, and he notes that there are open questions about where AI models get their data.
Pravin Kumar: AI built into the working day
Pravin Kumar runs his own Webflow development practice, building websites for clients and handling sales, marketing and accounting himself. His website describes the business in the first person, and his Webflow partner profile lists his location as New York City.
How he uses AI
According to an April 2026 article on his blog:
- Planning each day. He keeps a file of notes for each active client and spends 20 to 30 minutes each morning with Claude Code to look back at the previous day and plan the day ahead.
- Publishing and audits. Through a connection between Claude and his Webflow sites, Claude publishes blog posts directly to the site with the right details. On client sites, he also uses it to check every post in a collection for missing meta descriptions, posts that need new internal links and outdated content, which he says now takes minutes instead of hours.
- Starting points, not finished sites. For new projects, he uses Webflow's AI site builder to generate an initial page structure, which he says is never ready to publish but saves starting from a blank page. He uses v0 to generate individual interactive components.
- Design reviews. He shares screenshots of work in progress with Claude, which often points out accessibility, hierarchy and spacing problems he has missed, though he notes its feedback is not always correct.
- Sales preparation. Before a call with a prospective client, he asks Claude to review their website and suggest where his services could help.
What it changed
He says a blog post that used to take him two hours from research to publication now takes about 45 minutes. He writes that Claude drafts his blog posts following his own style rules, and that he reviews and edits them before they are published, which lets him publish far more often than he could alone.
Where he draws the line
He does not use AI to write client emails or project status updates, because his client relationships depend on genuine communication. For proposals, he lets Claude produce a first draft and then reworks it heavily in his own words. He uses AI to prepare for client calls, but the conversations themselves are his. In his view, the skill that has grown most in value is judgment: deciding what to build and what a client really needs.
What experience gave them that AI could not
This section is our analysis, not a claim made by either designer.
- Knowing when an answer is wrong. Karkovack notes that AI's code is not always right, and Kumar that its design feedback is not always correct. Recognising the bad suggestions takes experience.
- Seeing the whole project. Both describe AI freeing time for the bigger picture, how a site should look, work and serve its goal, which is where their expertise lies.
- Keeping their own voice. Karkovack writes all of his articles himself, and Kumar writes his own client emails and reworks AI-drafted proposals in his own words.
- Choosing how much AI to use. One uses it rarely and the other constantly, but each has decided deliberately where it fits.
The two approaches side by side:
| Eric Karkovack | Pravin Kumar | |
|---|---|---|
| Work | WordPress web design and development | Webflow web development |
| Base | Carlisle, Pennsylvania (per a 2020 WordCamp speaker profile) | New York City (per his Webflow profile) |
| How often he uses AI | Occasionally, when stuck | Throughout the working day |
| Main AI tools | ChatGPT | Claude Code, Webflow's AI site builder, v0 |
| What AI handles | Explaining and fixing code; small functions | Planning, blog drafting and publishing, site audits, design feedback, sales research |
| What stays human | All writing; planning and critical thinking | Client emails and conversations; final proposals; judgment |
| Reported result | More confidence, more time for quality | Blog posts in about 45 minutes instead of two hours |
What you can copy
These are our suggestions, drawn from the two stories.
- Use AI where you are weakest. Karkovack uses it for the coding that slows him down, while most of his design work still comes from his own head.
- Give AI context, not just questions. Kumar's daily notes for each client let AI help with real decisions.
- Start from a structure, then make it yours. AI-generated page layouts and proposal skeletons are starting points, not deliverables.
- Keep client communication human. Decide which messages always come from you.
- Pick one painful task first. Kumar suggests starting with the part of your work that causes the most friction.
Notes on the sources
The facts about Karkovack come from an article he wrote for Speckyboy, first published on January 13, 2025 and later updated, and from his own website, which gives his years of experience and describes his business as a limited liability company formed in Pennsylvania. His home town of Carlisle comes from his speaker profile for WordCamp Lancaster 2020, as his website does not give an address. The facts about Kumar come from an article on his own blog published on April 16, 2026, his website and his Webflow partner profile; he says in the article that Claude drafts his blog posts, which he reviews before publishing. The same article says he does not use AI to write proposals, but also that Claude produces first drafts of his proposals, which he then rewrites; we have reported both. Kumar's location comes from his Webflow profile, as his own website does not state it. Neither source gives revenue figures.
Sources
- How AI Changed My Web Development Workflow, Eric Karkovack, Speckyboy, January 13, 2025 (updated 2026)
- Eric Karkovack Web Design Services
- Eric Karkovack, speaker profile, WordCamp Lancaster 2020
- My Daily Workflow Building Client Sites with Claude Code and Webflow MCP, Pravin Kumar, April 16, 2026
- Pravin Kumar
- Pravin Kumar, Webflow partner profile
Related articles
Running a Photography Business Alone with AI: A Full-Time Wedding Photographer and a Weekend Photographer with a Day Job
Wedding photographers often spend more hours at the computer than behind the camera. A documentary wedding photographer in North Carolina and New York lets AI cull thousands of photos while he sleeps. A photographer who works a full-time job and shoots on weekends uses AI to cull, edit and retouch so she can deliver galleries faster. Here is what each one hands to AI, and what stays with the photographer.
A One-Person Operations Partner in France Who Lets AI Draft but Checks Everything: Sophie Kazandjian
Sophie Kazandjian runs a one-person digital operations business in the south of France, setting up systems, automations and websites for small business clients. She uses AI for transcripts, automations, dashboards, documents and drafts, and rebuilt her own website with it. She has also documented where it fails: confidently reporting a rise in email opens that was actually a fall, and reporting success when a third of the items were missing from one of its results. Her rule is that identifiable client information does not go into the AI tools she uses.
A Handyman Who Uses AI to Diagnose Jobs, Find Rare Parts and Build His Own Tools: Ray Duke
Ray Duke runs a handyman business near Tucson, Arizona, working for property managers. He uses AI to work out thermostat wiring from photos, reset a luxury fridge drawer instead of replacing it, track down discontinued parts, soften emails to unhappy clients and build his own pricing tools without knowing how to code. He is also clear about the limits: AI can show him the price range in his market, but the price is his decision.
Notes Done Before the Patient Leaves: A Solo Physiotherapist and a Solo Psychologist on AI Scribes
For clinicians in solo practice, documentation is unpaid work that often spills into evenings. A physiotherapist in Queensland, Australia, now completes most of her notes while she treats, and says she ended her after-hours paperwork within two weeks. A clinical psychologist in Singapore saves up to 45 minutes on each new intake note and finds she can stay more present with clients. Here is what each one hands to AI, and what stays with the clinician.