This post is built from the live operator inbox of the Klaudius pipeline as of 2026-05-24. We've drawn on a sample of over a hundred recent business replies (with more landing every week); every quote, percentage and pattern below is pulled from those actual conversations, not invented.
It's deliberately long because the question "what do I do when someone replies?" is the one we get asked most, and the honest answer needs the texture of real conversations behind it.
Before we start
Two things to flag.
First, when a business replies, they've already seen a live preview of their website. The Klaudius pitch includes a link to a real, deployed preview built specifically for them. The reply you're handling is from someone who's already clicked the link and scrolled the homepage. They're not evaluating an idea; they're evaluating a finished thing.
Second, this playbook assumes you're running the conversation in writing. We default to avoiding phone calls because the pipeline is built to scale, and at the volume we're operating at, calls are simply too time-consuming. Calls might convert better; we genuinely don't have enough data points to say either way. What we can say is that every paid deal so far closed in writing. If you want to test calls, do — what follows describes what's worked for us, not the only way.
The replies you actually get
Looking across the sample, replies break down roughly like this.
52% explicit rejection. "Not interested." / "No thanks." / "Am fine at the moment." / "Keep busy enough without one." The single largest category by a wide margin. Half of every reply you get is a polite-but-firm no. Make peace with that on day one.
14% warm yes. "I like this mate ill have a proper look tonight when I get home ok." / "Nice. Really like it actually." / "Thats very nice and helpful drop me a call when your free." They like the work, they're comfortable with the price, they want to engage. This is the shortest path to money.
13% polite-curious. "I'll let you know." / "Leave this with me for now and I will come back to you." / "I'll have a look later today." No clear signal in either direction. About a third of these eventually re-engage, a third quietly evaporate, a third become rejections after one or two follow-ups.
10% price question or pushback. Even though the pitch names the price, roughly one in ten will still raise money in some form: "How much is it mate?" / "What can you do with that price?" / "To much for now" / "Can you do anything with 50£ mybbrother?". The most lucrative of these conversations isn't the polite ask. It's the one that opens with an absurdly low counter-offer (more on that below).
8% conditional yes. "Like it but would need a few small changes." / "Cheers Rohan, looks good but does need a few changes can you give me a call tomorrow afternoon." / "Thanks, looks good - there would be a few things I'd want to change if I went ahead." They've evaluated the site and have specific feedback. The more specific the change request, the warmer the lead.
3% "who are you?" / "where are you based?" check. "Hi there I don't know you …" / "Where are you based?" / "I'm a bit dubious as I have no idea who you are or what guarantees I'll have a website after paying the money." A small but real subset that won't engage until you've explained who you are in plain words. The fix is one sentence and it works.
What surprised us most when we tabulated this: there's no "I love it but I want a call first" category as a percentage worth talking about. Only about 3% of replies asked for a call.
The conversation arcs that actually lead to money
Reading the inbox end-to-end, every paid customer has followed a distinct arc. Knowing which arc you're in is more useful than knowing the reply's "shape" because the arc tells you what your next move should be.
Arc 1: The fast yes (Stoke roofer, Birmingham electrician)
A customer replies within hours, the messages are short and enthusiastic, they pepper you with practical questions ("Are you local?" / "Could you add more photos?" / "How does it go live?" / "Will it help get more work?"), and the deal closes within a few days.
The Stoke roofer sent 30+ replies over two days. Many of them were a single line: "Nice web site." / "Can you take the photo of the van off." / "Have you done web sites long." The operator's job in this arc is not to sell; it's to keep up. The selling already happened when they opened the link.
What makes the fast yes convert: replying within an hour or two, taking every micro-request seriously, and offering to handle the domain purchase yourself so they don't have to learn anything new. "Domains these days are very cheap so I don't charge anything extra for it" is the line that closed the Stoke roofer the same evening.
Time-to-money for this arc, in our data: 1-2 days from reply to bank transfer.
Arc 2: The slow yes with conditions (Watford builder, Portsmouth roofer)
A customer replies on day one with a question or a small reservation. "How you do this one off, no monthly payments?" (Watford builder). "May I ask how you add photos to each section? How it can be regularly edited? Who owns the domain?" (Portsmouth roofer). The conversation runs for a week or two, with the operator answering questions, making design tweaks, sending updated links, and eventually arriving at agreed scope and price.
The Watford builder thread is the canonical example: it went 37 messages over two evenings, included a hard negotiation ("Can you do anything with 50£ mybbrother?" → eventually £120 with a referral hook), six design changes, an Instagram button add, a logo placement debate, the operator buying the domain mid-conversation, and a custom admin panel built and deployed before bedtime. The customer paid the next morning and has since said he'll refer his construction-trade contacts.
What makes this arc convert: treating every question as a real question, not a stalling tactic. The Portsmouth roofer asked the exact same three questions the Watford builder did ("how do I add photos? how do I edit it? who owns the domain?"). Both became paying customers. The customers asking these questions are doing due diligence, not making excuses.
Time-to-money for this arc: 1-3 weeks.
Arc 3: The credential correction
The customer replies politely with one specific request: remove something the site says about them that used to be true and isn't anymore.
Real examples from the inbox:
- "I'm not gas registered" (Cleethorpes plumber, after we'd pulled "Gas Safe" from an old van photo)
- "could you please remove any reference to the NICEIC from the draft please, as I'm not currently with them" (Chesterfield electrician)
- "I'm not doing joinery any more, just a laminate flooring and engineering click flooring" (Nottingham carpenter)
- "I haven't got that van no more or the landline" (Stoke roofer)
- "Could you take the photo of the van off" (Stoke roofer, same conversation)
These are not the customer being picky. They are the customer correcting a factual error within the website we just put in front of them. The operator's reply should always be a one-liner: "No problem, I've removed that, same link", with the change deployed within ten minutes. Almost every paid customer has had at least one credential-correction request in the conversation. None of the rejections did.
This arc is also a quiet test of operator speed. If you take 24 hours to fix a credential request, the conversation cools off. If you fix it in 20 minutes, it warms up. The pipeline's Claude Code wrapper makes 20-minute edits standard.
Arc 4: The price negotiation that becomes a relationship
The most interesting arc in the data.
Klaudius's pitch states the price up front. Most replies don't negotiate. But for the ones that do, the right move — counter-intuitive as it sounds — is usually to give the customer the discount they're asking for. Klaudius operators in particular might as well negotiate down, because the economics make refusal an unforced error.
Here's why. By the time a reply lands, the pipeline has already done the expensive work: it found the business, gathered the photos and information, designed the website, built it, and deployed a preview. All of those costs are sunk. The marginal cost of closing a deal at this point is essentially zero. £120 from a willing customer is £120 of profit you would not otherwise have had. Holding the line and walking away with £0 is, in pure cash terms, leaving money on the table for no reason.
And the upside doesn't stop at the cash. The Watford builder negotiated his price down from £299 to £120, and shortly after, sent us a referral lead from his construction-trade contacts. The Bath carpenter is in flight at £150 down from £199. Across the inbox, customers willing to negotiate are also, consistently, the most-engaged customers we have, and the ones most likely to talk you up to their network afterwards. The discount you give them today is the cheapest source of warm referrals you'll ever find.
The pattern these conversations follow, almost every time:
- Customer opens with an unprompted price challenge. "Can you do anything with 50£ mybbrother?" or "Hi Rohan, what can you do with that price?"
- Operator does not counter with a number. Operator asks "What's your offer?" or "What's the best you can do?"
- Customer names their floor. Usually well above the absurd opening. (£50 became £100, £150 became £150.)
- Operator meets near the customer's floor with an upsell hook. "120 if you can refer me to your friends."
- Customer accepts the new number and the conversation shifts immediately into design-and-deploy mode.
The line that works is "I could be flexible, what's your offer?". The floor is roughly half the asking price — below that the maths starts to feel insulting to the operator's time even though the cash is technically still profit. Above that, the customers worth keeping land naturally.
The "who are you?" / "where are you based?" reply
This is the single most reusable conversation pattern in the data. It comes through repeatedly, in identical wording, and the right answer is always the same.
The replies look like this:
- "Hi there I don't know you …" (Wadhurst mobile mechanic)
- "Where are you based?" (Salisbury tiler)
- "I'm a bit dubious as I have no idea who you are or what guarantees I'll have a website after paying the money" (Norwich decorator)
The reply that works, almost word-for-word every time:
"Sorry [name], my name is Rohan, I'm a software engineer based in London who helps build websites for businesses without one. I just came across your business on Google and thought your reviews were great so thought I'd build a website for you and see if you like it / would be interested?"
What this sentence does, that no other reply in the dataset does as cleanly: it names the operator, places them geographically, states the volunteer nature of the build (which is the real source of confusion: customers can't believe someone built them a thing before being asked), and offers an opt-out ("see if you'd be interested") that removes any pressure.
The geographic placement matters more than you'd expect, and it's worth pulling out separately. "Remote" sounds like a scam; a named real city sounds like a person. A clear pattern in our UK data: businesses are warmer to operators they perceive as being in the same country. In rare cases we've even had customers reject us specifically because we were in London rather than in their local town — rare enough that we still happily sell from London into every part of the UK, but real enough to flag. The takeaway isn't that you need to be local; it's that location matters more than the post-internet reflex would suggest. If you're inside the country your customers are in, lead with that. If you're not, lead with the actual city you're in anyway — any honest geographic answer reads better than "remote".
In all of the cases where we used this reply, the customer engaged further.
If you're a non-technical operator running Klaudius, save this paragraph as a snippet. You will reuse it weekly.
When they want changes: this is what Claude is for
Almost every paid customer asked for at least one change before paying. Across the warm-yes and conditional-yes replies, the change requests cluster into five categories.
Adding their logo. The Klaudius gather step pulls a logo if one's publicly available. Most small trades don't have a logo online. The customer texts it over, the operator drops it into the site within an hour. Edinburgh plasterer is the canonical example: "I like it although I'd like my logo included I do have a logo pack."
Removing a credential they no longer hold. Already covered in Arc 3 above. The most important one to action quickly.
Removing a service they no longer offer. Same shape as credential correction. "I'm not doing joinery any more" is a common variant from carpenters and tradespeople whose scope has narrowed.
Adding a contact channel. WhatsApp button (Bromley landscaper: "the only thing I dont like is that it dosnt have a direct access to my WhatsApp. Most clients will prefer to WhatsApp not email or call"). Instagram link (Watford builder, multiple). Direct phone call link. A surprising number of trades route their leads through channels that are not phone or email.
More photos. Customers regularly want to add photos of their work that aren't anywhere on the open web — they live on the trader's phone camera roll. That's why the Klaudius gather step didn't pull them in the first place: it scans public sources, so anything that exists only in the customer's pocket has to come to us directly. The customer sends them over, the operator drops them into the site within an hour. The Lancaster tree surgeon sent in ten photos, annotated with which section heading each one belonged to. All added same-day.
The rule for handling all five: build the change yourself, in Claude Code, in five minutes. This is what the operator pipeline is built for. Just copy the customer's message straight into Claude Code, or screenshot it and paste the image in — Claude reads both, figures out what they're asking for, and ships the change. The Watford builder asked for six changes across one evening (logo placement, Instagram button, photo swap, navbar tweak, footer logo, tagline retention) and all six shipped before bedtime. The fact that we can absorb six change requests in under an hour, for a £120 deal, is the entire reason this business model works at this price point. A traditional web agency cannot do this and stay solvent.
The craziest single change request the operator has handled: a self-serve admin panel so the Watford builder could upload photos to his own site without messaging us. Building a custom CMS is normally a week of engineering for a senior developer; a traditional web agency would quote it as a separate project, easily four figures. Claude Code one-shot it in fifteen minutes, deployed to /admin with a memorable password. The customer's been managing his own portfolio since. That capability shift is the reason the entire £120 deal is even possible at this price.
Taking the money
This is the question that scares non-technical operators the most: "How do I take payment? I don't have a business."
The honest answer: you don't need one for a long time.
The vast majority of customers are happy, willing, and quietly expect to pay you by direct bank transfer. The operator pastes their personal bank details into the conversation, the customer transfers the money the same day or the next morning, and the site goes live. No invoice required, no payment processor, no fees skimmed off the top.
A small minority will want a formal invoice with a correct billing address on it. For those, build a basic PDF invoice template before you need it, and double-check the customer's address before sending — getting the address wrong on the first try is a small friction that costs you a day. Stripe Invoicing or any similar tool will produce one in two minutes, free, no setup.
You won't need a registered company for the first stretch of deals. In most countries you can take money into a personal account up to a tax-free trading threshold before you owe anything. In the UK that's £1,000 of revenue per tax year, beyond which you'd register with the tax authority. Most other countries have an equivalent: the threshold and the regulator change, but the principle (small revenue stays personal, larger revenue gets a company) is universal. Check your own country's rules early so you know roughly where the line sits, then go close some deals.
Getting the website live
Once you've agreed that the customer is keen to go ahead, two pieces of work remain.
Before anything else: ask Claude to do an SEO pass on the website. The pipeline doesn't do this by default, and that's deliberate — SEO is invisible, behind-the-scenes work, and there's no sense spending it on a site nobody's committed to buying yet. But the moment a customer's keen and you're about to take it live, it's well worth it: just ask Claude "can you make sure the website is fully optimised for SEO?" and a few minutes later you're handing over not just a website but an SEO-optimised one, structured to rank for the local searches the customer's customers actually type. Mention it in the next message you send them; it's a credible value-add that costs you nothing.
Then comes the domain. Most customers don't already own a domain, and they don't really want to think about it. The right operator move is to handle the whole thing yourself: it signals professionalism, removes a chunk of decision-fatigue from the customer's side, and is one of the highest-leverage conversion moves in the entire playbook.
The domain flow takes about ten minutes of your time and is, mechanically, just four steps.
1. Pick a domain. Ask Claude to suggest 2-3 strong options that fit the business. Claude has the reasoning to come up with strong brand fits based on the business name and industry, and can autonomously check availability across registrars, so what you get back is a real, available shortlist worth sending. Pass them to the customer, ask which they prefer.
2. Register it yourself. A .com runs around $5-6 for the first year at any major registrar — IONOS, Namecheap, Cloudflare Registrar, Porkbun, Spaceship all do the job. Annual renewals are roughly $10-13/year. Bake both costs into the price you've already quoted; almost no customer notices, and it keeps the relationship clean: they never have to log into a registrar dashboard, and the domain is something you can transfer to them later if they ever ask.
3. Tell Claude to point the website at the new domain. Modern hosting providers (the kind every Klaudius site sits on by default) are free at the volumes a small-business website will ever produce. Once you've registered the domain, just tell Claude "I've purchased [domain.com], please point the website to it." Claude attaches the domain to the hosting project, generates the SSL certificate, and tells you the exact two custom nameservers your registrar needs to use.
4. Update the nameservers at the registrar. Log into the registrar you bought the domain from, find the domain's DNS settings, and replace the default nameservers with the two custom ones Claude gave you. Save. That's it — your part is done.
DNS propagation can take up to 48 hours, but in practice most domains go live within an hour. Once they do, the website is live at the customer's own URL with HTTPS already working.
Two final touches.
Submit it to Google Search Console. This is what actually tells Google the website exists and asks it to start crawling. Without this step the site eventually gets indexed organically, but it can take weeks; with it, you're usually findable within a day or two. If you've never done this before, just ask Claude to handle it using Claude in Chrome — it'll verify domain ownership, submit the sitemap, and request indexing in about five minutes, with you just watching the browser do the work.
Send the link to the customer and tell them to add it to their Google Business Profile. That single action on their side is the most valuable thing they can do to start getting traffic — their existing Google reviews and map listing now link straight to a real homepage.
That's it. Conversation closes, money already cleared, customer has a working website on their own domain, indexed by Google, with their phone number on it. Done.
Two things we genuinely didn't expect
The customers asking the most questions are the ones who pay. Almost every paid customer opened with "how does this work?" or "who owns the domain?" or "can you do this for £50?". The customers who paid asking only "yes please" are zero. Engagement velocity, not initial warmth, is the conversion signal.
Holding the line on price isn't always right. The natural instinct is to refuse to negotiate. The data flatly contradicts that. Customers who negotiated their price down have been the highest-engagement customers we have, including a referral-generator. Hold the minimum (your floor), not the asking price.
A working model for every reply you get
Stop trying to classify replies by tone. Classify them by what comes next.
- Will money clear this week? → Fast yes. Cancel everything else, finish the conversation now.
- Will money clear this month? → Slow yes with conditions. Schedule it; check in on a regular cadence.
- Are they negotiating? → Price arc. Ask their offer; meet near it; don't go below the floor.
- Are they correcting a fact? → Credential arc. Fix it in 20 minutes; re-share the link.
- Do they want proof of who you are? → Identity check. One-paragraph reply, then carry on as normal.
- Have they said no? → Rejection. Respect it the first time; "stop contacting me" the second.
- Is it neither yes nor no? → Polite-curious. One short reply, then patience.
You will get every one of these arcs in your first week of running the pipeline, probably more than once. The customers worth paying attention to are the ones who keep messaging. The customers worth letting go are the ones whose first message is also their last.
A closing note
The numbers underneath this whole conversation are small ones: around an hour per close, around £10 in cash cost per deal. They look thin in isolation. They look very different when you remember that the pipeline did almost all of the expensive parts (found the business, scraped the public information, designed the site, built the site, sent the cold message) and the human in the loop only spent an hour per close on the parts that actually need a human.
The conversation is the easy part. The pipeline was the hard part. The pipeline is the thing you bought.
Team Klaudius