The Project You Already Priced Out
Recost the build, hire, or production you shelved as too expensive. The number changed and you never went back to check.
By James Schramko · Updated August 2026
You quoted a project two or three years ago, flinched at the number, and shelved it. That number is still sitting in your head as the true cost of doing it. It is probably wrong now, and it is wrong by a lot more than a normal price move explains.
The cost of building things with software has fallen fast enough that decisions made on old numbers are quietly stale across the whole business. Not "AI is interesting," a specific operational fact: the quote you got in 2023 for a documentary, a membership platform, a set of client playbooks, or a piece of custom software is not the quote you would get today, because the labour behind that quote has partly moved to a tool that costs a subscription instead of a team.
This playbook is about running that recheck deliberately, instead of finding out by accident.
The Pattern Across Four Different Projects
A filmmaker researching a documentary got quotes near a million dollars and a Hollywood-length production timeline to do it properly. He studied several existing documentaries, fed his notes into an AI system to work out story structure, and produced his own version for a small fraction of that figure, on an eight-week runway that ended in a multi-city tour. The film did not get worse because it cost less. The cost estimate had simply been built on an assumption about how many people and how much time the work required, and that assumption stopped being true.
A business coach's client cloned an entire coaching business in about two weeks: platform, payment processing, a video sales letter, and a set of forty to fifty pieces of content built from eighteen source pieces. That is normally a quarter of agency work, sequenced properly with a developer, a copywriter, and a video editor. It ran in fourteen days because one person with the right tool did the developer's job, the copywriter's job, and most of the video editor's job in sequence, on the same afternoon they had the idea.
A brand-new coaching client received close to fifty thousand dollars of build value in his first ten days: a full coaching program built out, eighteen playbooks pulled from his own call transcripts, a cloned website, a sales video, a funnel, ad creative, and a ninety-day plan. He had booked four calls in two weeks and implemented within a day of each one landing. None of that was a discount on agency pricing. It was work that used to require an agency retainer, produced instead inside the coaching relationship itself.
A tradesman running an estimating business paid roughly six thousand dollars to build a one-button AI tool that reads a job and returns a tender analysis, with an offshore team finishing the details overnight. His coach's read on the finished tool: wrap a real business around this, because the tool alone is worth building a company on. Compare that six thousand dollars to what a custom estimating platform would have cost built the conventional way, spec, dev team, months of testing, and the gap is the same gap running through every example above.
Why The Old Number Sticks
You do not re-quote something you have already filed as "too expensive." Once a project gets a number and a no, it drops out of the part of your attention that reconsiders things. It sits there as settled, the same way a shelved offer or a closed line of business sits settled, except this kind of shelving has a silent expiry date attached that nobody wrote on the file.
The founders in every story above did not discover the new cost by researching pricing. They discovered it by trying the thing directly, usually because someone showed them a tool working on a related problem first. The recost happened almost by accident. That is the part worth fixing. You should not need an accident to find out your own numbers changed.
What Actually Collapsed
Not the whole cost of everything. Three specific categories moved the most, and they are worth naming separately because they call for different action.
Structured production work. Video editing, documentary assembly, content repurposing, transcript-to-playbook conversion. Work that follows a describable process on inputs you already have. This moved from a team-and-months cost to a tool-and-days cost, because the process itself is what the tool now runs.
Repetitive analysis at volume. Reading eighty-plus videos for one detail, reading every transcript for one recurring theme, running the same triage logic across hundreds of records. A task that used to justify a part-time hire or a monthly VA retainer, done in an afternoon once the pattern is captured properly.
One-off custom builds against a known spec. A funnel, a platform clone, a specific automation tool. If you can describe exactly what it needs to do, the build itself has gotten dramatically cheaper, even when the idea and the judgment behind it have not.
What did not collapse: judgment about which project is worth doing at all, the taste to know when the output is actually good, and the relationships that make people trust what you built. Those are still the expensive, scarce part. The collapse is specifically in the labour of production and analysis, not in deciding what is worth producing or analysing.
The Recost Pass
Pull the list of everything you have shelved as too expensive, too slow, or too big a project for now. For each item, run three questions instead of guessing.
What category is this? Structured production, repetitive analysis, or a custom build against a known spec. If it is none of those three, the old number was probably about right and the shelving stands.
Can I describe the exact steps a person would take to do this by hand? If you cannot describe it, you cannot hand it to a tool yet, and the old quote likely still holds. If you can describe it in plain steps, that is usually the sign the cost has moved.
Has anyone I know actually rebuilt something like this recently, and what did it cost them? Do not estimate from a blog post about AI capability in general. Get one real number from one real build, close to your own situation, and use that as your new baseline instead of your old quote.
Anything that clears all three gets re-quoted properly before you decide it is still out of reach. Do this once across the whole shelved list, then repeat it every few months, because the categories that collapsed keep collapsing further and the list restocks itself with new candidates as old assumptions harden again.
What This Is Not
This is not a pitch to automate everything or to distrust every quote you receive. Plenty of shelved projects are correctly shelved. A project that needs deep relationship trust, original judgment under ambiguity, or a genuinely novel creative act was never primarily a labour-cost problem, and no tool changes that. The recost pass exists to catch the other kind: projects you filed as too expensive purely because of the hours a person would have needed, in a year when that specific assumption was still true.
Quick Reference
- Pull your shelved-as-too-expensive list. The number you are carrying is probably stale.
- Sort each item into structured production, repetitive analysis at volume, or a custom build against a known spec. Anything outside those three probably still costs what you think.
- Get one real cost from one real recent build close to your situation before you trust any new estimate.
- Re-quote before you re-shelve. Do this pass on a standing cycle, not once.
- What did not get cheaper: judgment on what is worth doing, taste in what is good, and the relationships that make the output land.
Working out which of your old no's are actually a yes now, with your own project list in front of me, is exactly the kind of pass I run with founders inside Mentor.