BidReady

A productised catalogue of proposal and RFx templates I built as a side hustle. The interesting part wasn't the templates themselves, which are the easy bit when you've written proposals for a living. The interesting part was using Claude as the entire team behind the product.

The objective

I write bids and proposals for a living. The objective for BidReady was specific and commercial. Turn that practitioner experience into a productised asset that generates revenue without ongoing client delivery. Mortgage-covering income, low touch, evergreen. No retainers, no SOWs, no project meetings.

The constraint was time. I have a full-time job and a family. The whole build needed to be compressed into evenings and weekends, which meant AI had to do real work, not just brainstorm.

Why this is an AI project

Building a digital product business traditionally means stitching together a copywriter, a graphic designer, a brand strategist, a developer for landing pages and a virtual assistant for store admin. Each of those is a real cost in money and elapsed time.

I used Claude as the entire team. The interesting part of the project wasn't "can AI write a product description". That's table stakes now. The interesting part was orchestrating Claude across very different disciplines (commercial copy, document templating, visual identity, Python image generation, ffmpeg video production, legal copy) in a way that produced a coherent brand and a sellable product, not a pile of disconnected outputs.

What I asked Claude to do

The work split across five workstreams, all running through Claude.

Product design. Structuring the Word-document templates around what evaluators actually look for in proposals, with embedded placeholder conventions and instruction notes.

Document generation. Producing the .docx files programmatically using Node.js and the docx library so I could regenerate the entire catalogue across multiple colour styles without manual formatting. Generating those by hand in Word is a nightmare. Style drift creeps in, tables get misaligned, fonts inherit unpredictably. Code is the only sensible way to keep the catalogue consistent.

Visual identity. Logo, monogram stamp, listing images, store banner and a 15-second brand video. All generated with Python, Pillow and ffmpeg under Claude's direction.

Commercial copy. Every Gumroad and Etsy listing, the About sections, the post-purchase email, the privacy policy.

Operations. GDPR and Australian Privacy Act-compliant policies, FAQ copy, refund handling.

What I learnt

First-pass visual design is generic. Every visual workstream needed two or three iterations before it looked distinctive. The first cut always looked like an AI-generated marketing asset. The fix is to push back specifically and re-brief with constraints. My first attempt at product images showed a stack of mockup pages in the relevant colour palette. They looked professional in isolation. The problem only became obvious when I lined up all the thumbnails in a row on the Etsy storefront. They all looked identical. I went back to Claude and reframed the problem from "make a nice product image" to "make thumbnails that are visually distinct from each other when seen as a grid of 200-pixel squares on a phone". The redesign came out of that reframe.

Vague briefs produce vague output. The single most important lesson. The quality of any AI-generated artefact is bounded by the specificity of the request. "Make a brand video" produced a busy mess. "Five scenes, three seconds each, one sentence per scene, no overlapping elements" produced something usable.

AI pushes back when you ask it to do the wrong thing. I asked Claude to write a Reddit post designed to look like authentic user content for a thread that explicitly prohibited AI-generated copy. Claude declined and offered to brief me on what to write myself. That was the right call. Disguising AI marketing as authentic user content is the kind of thing that gets storefronts banned and reputations damaged. Worth recording because it changed how I think about what AI is actually useful for in marketing. It's a force multiplier on legitimate work, not a way to fake credibility I haven't earned.

What it took, what it would have cost

Compressed working time across the whole project, including the visual iterations, the failed brand video, the listing image redesign and the policy work, was roughly the equivalent of a long weekend. All templates in all colour styles, both storefronts populated, every listing written, the privacy policy, the brand video.

A conservative estimate of what the same scope would have cost through traditional channels sits between $6,400 and $12,200 in copywriter, designer, motion designer, legal and templating consultancy fees. The actual cost was the Claude subscription plus my time.

The takeaway

AI isn't a replacement for the practitioner experience that makes the templates valuable in the first place. The reason BidReady works as a product is that the templates are structured around what evaluators actually look for. That knowledge came from writing real bids for real clients over real years.

What AI replaces is the friction between having that knowledge and turning it into a sellable product. The copy, the design, the document generation, the legal scaffolding, the storefront setup. All of the work that would normally sit between an idea and a live product, compressed from months to days. That's the actual value.

Your expertise is the moat. AI just removes the cost of expressing it.