AI builders solved building. Agencies still need a funnel.

TL;DR

The line, up front

Building a site is now free. Building a business is not.

That's the entire thesis. The rest of this post is what it means for web agencies, why it doesn't reduce to "use Lovable instead of Prewiro," and how we plan to plug into the AI-builder generation rather than against it.

What the AI-builder generation actually shipped

The last eighteen months produced a category of tools that didn't exist before:

• Lovable — type a sentence, get a deployed React app. Took the friction out of "I have an idea" → "I have a working URL" for non-developers.
• Bolt — StackBlitz's in-browser AI builder. Lets a designer prototype a full-stack app without ever opening a terminal.
• Emergent — the Indian-origin AI app builder, growing extraordinarily fast. Verifiable proof that this category is global, not Silicon Valley alone.
• Replit — Agent mode turned a coding IDE into a project-doer. You describe what you want; the agent ships it.
• Bubble — the no-code original, now augmented with AI generation.

The honest read: these are extraordinary products. Pre-2024, building a small-business marketing site was a 2 to 4 week deliverable. With these tools, the same site is 30 minutes of prompting. That's not incremental — that's a category collapse.

We are huge fans. Genuinely. Our team uses Lovable and Bolt for our own internal tooling, we've been watching Emergent's growth with admiration, and Figma Make is changing how design hands off to development.

So this post is not a takedown. The opposite. It's an explanation of why agencies still need something more on top of these tools — and why that something is the layer Prewiro is building.

The agency math: building used to be 80% of the work

Here's the quietly important part. Before AI builders, an agency's economics looked roughly like this:

Activity	Share of effort
Find lead	5%
Qualify lead	5%
Pitch / win client	5%
Build site	70%
SEO + content + ongoing	10%
Other (admin, billing)	5%

Building dominated. So when an agency owner thought about "how do I scale," the answer was always "hire developers" or "find a way to deliver faster." That's why the entire web-agency category for twenty years has been about templates, page builders, WordPress, Wix, and now AI builders — each generation reduced the cost of delivery.

Lovable, Bolt, Emergent, Replit, Bubble took it to its logical endpoint. Delivery is now nearly free. Same site as before, four minutes instead of four weeks.

But here's the thing nobody talks about — after that compression, the agency math doesn't get easier. It gets harder. Because the bottleneck moved.

After AI builders, the bottleneck is the funnel

Run the same table again, with the build collapsed:

Activity	Share of effort (post-AI-builder)
Find lead	20%
Qualify lead	15%
Pitch / prove value	35%
Build site	10%
SEO + content + delivery	15%
Other (admin, billing)	5%

Building dropped from 70 percent to 10 percent. Everything else expanded to fill the space. Specifically:

1. Finding the right prospect — agencies need 50 leads on the top of the funnel to land 5 paying clients. With building free, more agencies enter the market, and finding the right SMB to pitch becomes harder, not easier.
2. Qualifying — does this lead have budget? Are they actually a buyer for what we sell? Are they not just shopping?
3. Proving value before the pitch — every other agency now has a Lovable account too. A pitch deck is not enough. The lead needs to see the work in their voice before they'll pay attention.
4. Multi-channel outreach + follow-up — first email gets ignored. WhatsApp lands better in many markets. Then a competitor brief. Then a tailored social kit. Then a reel. Each touchpoint has to feel custom.
5. Closing — by the time the lead is warm, they want to talk to a human. The funnel still ends in a sales conversation; it just earns it.

An AI builder solves point 4. It doesn't touch points 1, 2, 3, 5.

That's not a flaw. It's by design — those tools are built for the founder shipping their own product. The shape is one person → one app → one company. An agency's shape is fundamentally different.

The model mismatch for agencies

When a founder uses Lovable, the workflow is:

Founder has idea → prompts Lovable → gets working app → tests with users → iterates.

Linear. Singular. The builder serves the builder.

When a web agency uses Lovable, the workflow they want is:

Agency picks city + niche → finds 50 prospects → scores who's worth converting → builds for top 10 → ships work to each in their voice → follows up across 6 days → closes 5 → bills 5.

Different shape entirely. The builder is one step inside a much larger flow that the agency owner is running manually — usually with a spreadsheet, a calendar, four different SaaS tools, and a virtual assistant.

That manual orchestration is where the time goes. It's also where most one-person agencies hit a ceiling — at some point, you can only run so many parallel prospects before the operations break. Adding a faster builder doesn't lift that ceiling; the ceiling is in the orchestration.

Where Prewiro fits — the distribution layer

Prewiro is built for that other shape. The agency-shape.

The platform takes one input from the agency owner: city + niche. Everything else runs autonomously:

• Day 0 — Prewiro pulls 20 candidate businesses from Google Maps in that city + niche, scores them on conversion likelihood (rating quality, review velocity, current digital presence, vertical-fit signals), and selects the top 10.
• Day 1 — for each of the 10, generates a tailored marketing site, deploys it to a public URL, and ships the URL to the lead via email and WhatsApp. The lead opens real work in their voice, not a deck.
• Day 2 — sends a set of brand-voice blog posts, structured for both classic SEO and AEO (the answer-engine surface where ChatGPT, Perplexity, Claude, Gemini, and Google AI Overviews now live).
• Day 3 — competitor gap brief: who ranks for the queries this lead should rank for, and the specific gap to close.
• Day 4 — Instagram social media kit (logo, captions, post templates, brand-tailored images), delivered via WhatsApp, ready to publish.
• Day 5 — a 15 to 30 second AI-generated reel built around the lead's service. WhatsApp again.
• Day 6 — local influencer shortlist: micro, mini, macro creators in the lead's city × niche, sized to budget, partnership-ready.

By day seven, the conversation is already earned. Out of 10 leads, our internal target is one paying client per pipeline run — a 10× lift on the 1 to 2 percent conversion rate cold web-agency outreach typically achieves.

See the full agency cost breakdown vs Prewiro → for the line-by-line economics.

The agency owner did none of the manual work. They picked a city. They picked a niche. The funnel ran.

That's the distribution layer. It's not a builder. It's not a workflow tool. It's the layer between "I want to grow" and "I have a paying client."

Concrete contrast

Two scenarios. Same week. Same one-person operator.

Scenario A: solo founder + Lovable.

Builds one app. Spends days 1 to 3 on the build. Spends days 4 to 7 on Twitter trying to get attention. Hopes someone signs up. End of week: 1 app live, ~10 to 50 sign-ups if launch landed well, 0 paying customers (typical for early-stage product).

Scenario B: solo agency + Prewiro.

Picks city (Bangalore) + niche (dental clinics). Day 0: Prewiro scrapes 20, scores 10, builds tailored sites for the top 10. Day 1 to 6: cadence runs autonomously. End of week: 10 sites deployed, 10 cadences run, ~3 to 5 leads in active conversation, 0 to 2 closed. Bills first paying client by week two.

Both operators worked the same hours. They got categorically different outcomes — because they were running categorically different workflows. The AI builder fits scenario A. Prewiro fits scenario B.

This is what we mean by the distribution layer not being a competitor to the AI-builder generation. It's a different shape of tool, for a different shape of customer, doing a different shape of work.

Why building "another builder" would have been wrong

When we started Prewiro, we knew the AI-builder space was crowded and getting better every month. The temptation was to ship our own builder and compete on raw build quality. We deliberately didn't.

Three reasons:

1. The category was already going to be commoditised — Lovable, Bolt, Emergent, Replit, and Bubble are all racing to make building cheaper and faster. We could not have outrun any of them on that axis, and we shouldn't try.
2. The customer's problem moved. As building got cheap, the part that didn't get cheap became the bottleneck. We chose to live where the bottleneck is.
3. We can be a partner instead of a competitor. Builders need a layer that brings them volume of real-business use cases. Agencies need a layer that brings them volume of real prospects. Prewiro can sit between the two.

That third point is the part that makes us most excited about the next year.

Where this is going — interoperability with the AI-builder stack

We're already storing generated design assets in Figma-compatible structure today. The directional roadmap:

• Export to Figma — every Prewiro-generated site available as a Figma file, so an agency can take a generated site and extend it with Figma Make, or hand it to a designer for polish before delivering to the client.
• Export to Lovable / Bolt / Emergent / Replit project formats — for the agency that wants to keep iterating in their preferred AI-builder environment after Prewiro has done the lead-gen, scoring, and Day-1 site generation. Use Prewiro to win the client; use the builder of your choice to expand the engagement.
• Bring-your-own-builder mode — eventually, an agency with its own Lovable / Bolt / Emergent workspace will be able to plug it into Prewiro, and the platform will route the Day-1 site spec into that environment instead of using our default renderer.

The principle: the agency's choice of build environment is the agency's choice. Prewiro takes care of everything around the build — finding the lead, deciding it's worth building for, wrapping the build in proof artefacts, running the 6-day cadence, closing the deal. The build itself can be ours, or Lovable's, or Bolt's, or Emergent's, or Figma Make's. The funnel is the same.

Closing — distribution is the new moat

Two years ago, the moat for a web agency was being able to build well. Templates and Wix had not yet caught up; technical chops differentiated.

The AI-builder generation closed that moat. Lovable, Bolt, Emergent, Replit, and Bubble made building well into table stakes for everyone, including non-developers. That's done. It's not coming back.

The new moat is distribution. Whoever can find the right businesses, prove value before the pitch, and run a tailored cadence that earns trust — wins. Whoever can't, gets stuck delivering for whoever sent them work.

Prewiro is the distribution layer. It's not an AI builder. It's not a competitor to the AI-builder generation. It's the funnel that makes those tools useful for agencies — and for the next thousand solopreneurs who want to start a web agency without hiring eight specialists first.

If you're already shipping with Lovable, Emergent, Bolt, Replit, or Bubble — keep doing it. Prewiro brings you the right clients to ship for. See the cost breakdown vs traditional agencies →, or start with 5,000 free credits → — that's 50 pipeline runs, ~500 leads outreached, before you spend a rupee.

The build is solved. Let's solve the rest.