Here is the offer that sounds great in a sales meeting and quietly fails three months later: a training session, plus ten or twenty custom prompts written for your business, and you are away. The prompts get used for a fortnight. Then the document settles into a shared drive, and usage flatlines. The library was never the problem. Treating it as the deliverable was.
We see this pattern often enough that it is worth naming. A team buys prompts, feels productive for a few weeks, and then drifts back to old habits because nothing about how they work actually changed. This post is a constructive argument against that model, from people who run AI training for a living and want it to stick.
The prompt pack promise, and why it fails
Prompts matter. A well-written prompt is genuinely more useful than a vague one. But a pack of them is a starting point, not a product, and three failure modes show up almost every time a business buys the pack and stops there.
Prompts decay. Tools change, models change, and the business changes. The prompts do not. A prompt tuned for last quarter's process quietly stops matching this quarter's, and no one is watching for the drift.
Copy-paste treats Claude as a vending machine. Paste in, get output, move on. That misses the actual gain, which comes from giving Claude context and letting it work like a colleague rather than a slot machine you feed coins.
Nobody owns adoption. Once the trainer leaves, there is no named person responsible for whether the team actually uses the thing. Momentum fades and nobody notices until it is gone.
None of these are failures of the prompts themselves. They are failures of the assumption that handing someone better words is the same as changing how their work gets done. It is not, any more than handing someone a recipe makes them a cook.
What actually drives adoption
The businesses that get real value from Claude treat adoption as a change of process, not a handout. Four things separate them from the ones whose prompt pack is gathering dust.
Workflow ownership. Each high-value workflow has a named owner and a clear definition of done, so there is always someone accountable for the outcome.
Persistent context. Claude Projects hold the instructions, files, and tone for a piece of work, so staff start from the team's setup instead of a blank chat every time.
Progression to delegation. The biggest gains arrive when repeatable work moves to Claude Cowork with guardrails, not when people learn to word a chat request slightly better.
A 30-day checkpoint. Usage and time-saved numbers put in front of leadership at the one-month mark, so adoption is measured rather than assumed.
A worked example
Picture a mid-sized professional services firm in Melbourne. Proposal drafting starts as a twenty-prompt pack: useful, but each proposal still takes about six hours because every prompt is run by hand. Move the same work into a Claude Project with templates and precedent files, and the blank-page problem disappears. Move it again into a Cowork workflow that produces a first draft overnight, and a person spends their time editing rather than starting from nothing.
Hours per proposal fall from six to about two. At Australian charge-out rates, that recovered time is worth somewhere between $60,000 and $150,000 a year for a team of that size. The prompts were step one. The workflow was the product, and the difference between the two is the difference between a pilot that fizzles and one that pays for itself.
Where prompt libraries still fit
To be fair to the humble prompt pack: a good library is a useful on-ramp and a handy training artefact. It lowers the barrier for a nervous first-timer and gives a team a shared starting vocabulary. The failure is not building one. The failure is stopping there and calling it transformation.
So what should you ask for instead of a prompt pack? A short list of the workflows that eat the most time, an owner for each, a Project set up with the real files and tone, and a date in the diary to check whether usage actually moved. That is less exciting than a glossy library of clever prompts, and far more likely to still be running next quarter. The measure of good training is not how many prompts you leave with. It is how little the team needs them in six months, because the way they work has quietly changed.
We build Claude training around workflows rather than handouts, and we carry the high-value ones through to a working Claude Cowork implementation. If your last AI training left a document nobody opens, book a discovery call and we will talk about what adoption actually takes.
