AI Operations Emergency Kit
A complete method for implementing AI in a small business: the 7-day roadmap, 50 prompts, 10 operating templates, an ROI calculator and 50 documented use cases.
With one difference: every saving figure in this kit is anchored to published studies with a control group, cited by author, year and URL. You can check them before you buy.
time on professional office writing, with assessed quality going up, not down
Noy & Zhang · MIT / Science, 2023
n = 444 professionals
productivity for novice customer-support staff, in a real operation
Brynjolfsson, Li & Raymond · NBER, 2023
n = 5,179 agents
accuracy when the task falls outside what AI does well: worse than not using it
Dell'Acqua et al. · Harvard / BCG, 2023
n = 758 consultants
of all hours worked across a whole economy: the real aggregate saving today
Bick, Blandin & Deming · NBER, 2024
US national survey
The problem
Almost every small business already knows AI can help. What they don't know is where to start without breaking something.
And it isn't a shortage of information: it's that the available information is written for companies with a data team, or it's a list of prompts with no method around it. The questions that actually stall an implementation are different ones.
It gives you a seven-day method to answer those six questions about a real process in your business, with the templates, prompts and calculator that method needs. It doesn't teach you a specific tool and doesn't depend on one.
Why this kit is different
Most AI products promise saving percentages nobody can trace. This one brings the whole table, with sample size and study, so you can verify it and cite it in front of your own management.
| What was measured | Result | Sample | Study |
|---|---|---|---|
| Professional office writing: announcements, short reports, delicate emails | Time −37% Quality +0.45 SD |
444 professionals | Noy & Zhang, MIT / Science, 2023 |
| Customer support: cases resolved per hour, in a real operation | +14% average +34% for novices |
5,179 agents | Brynjolfsson, Li & Raymond, NBER WP 31161, 2023 |
| Consulting tasks inside the AI capability frontier | −25% time +12% tasks |
758 consultants | Dell'Acqua et al., Harvard Business School WP 24-013, 2023 |
| The same tasks, designed to fall outside that frontier | −19 percentage points accuracy | 758 consultants | Dell'Acqua et al., Harvard Business School WP 24-013, 2023 |
| Exposure of administrative work to generative AI | 24% high 58% medium |
Global analysis by occupation | Gmyrek, Berg & Bescond, ILO WP 96, 2023 |
| AI adoption gap between large and small firms in the European Union | 17% small 55% large |
157,000 firms | Eurostat, ICT usage in enterprises survey, 2025 |
When the task falls outside what AI does well, the teams that used it were 19 percentage points less likely to reach the right answer than the teams that didn't. It isn't that AI doesn't help: it helps on some tasks and gets in the way on others, and the difference isn't obvious from the outside. That's why the kit devotes a whole document to deciding where not to use it.
What's inside
It isn't a folder of loose files. Each document produces the input for the next, and together they build the file of a pilot you can audit.
The full 7-day roadmap: what happens each day, what the deliverable is, how long it takes and what that day's trap is. Plus the five rules and the five most expensive mistakes.
13 pages · editable Word
The ten-step operating model, the six AI patterns with their characteristic failure mode, and the information classification that decides what may be pasted into which tool.
11 pages · includes a workflow design canvas
Fifty business prompts ready to copy. Each with the literal text, what information you have to paste, and the mandatory verification before using the output.
35 pages · 8 functional categories
Process assessment, project charter, RACI, risk register, KPI sheet, weekly operating review, action tracker, SOP, experiment log and automation specification.
24 pages · with space for your logo and sign-off
Six Excel sheets: a calculator with status colours and an automatic verdict, an indicator dashboard with charts, and an assumptions sheet carrying the evidence behind them.
Excel · no macros, opens in Excel, Sheets and LibreOffice
Fifty cards by department with current pain, inputs, what the AI does, expected output, indicator and human control. Plus three cases worked through with the full calculation.
37 pages · 7 functional areas
What it looks like inside
No mockups, no renderings: this is what opens when you unzip the file.
One case, with the full calculation
One analyst spent 45 minutes on each reconciliation run, 60 times a month. Eight to ten invoices a month were paid with discrepancies caught only afterwards.
The error rate fell from 6% to 1.5%, because the prompt flags inconsistencies that used to slip through. Three low-quality scanned invoices had to be processed by hand and were documented as an exception.
It was scaled with two conditions written into the procedure: mandatory verification of every amount above $500 against the source document, and scanned invoices excluded from the flow. The documented exception is what made approval possible, not the saving.
The kit includes this case and two more worked the same way — among them one that was rejected, and why that is also a result. They are scenarios built with the kit's methodology to show the full calculation, not results measured at a specific company.
The method
Between 45 and 90 minutes a day, one single person. If the day ends without its deliverable, you don't move on: you repeat the day.
Then you repeat. The second process costs half the effort, because the method has already been proven inside your company and with your people.
Who it's for
Who wrote it
Price
One-time payment, no subscription. Secure checkout by Stripe. Price in US dollars.
Digital product. Because it is delivered instantly, it is non-refundable once the file has been sent. Subscriptions to AI tools are not included and are not required to use the method: the kit works with any general-purpose conversational assistant.
Questions
No. The kit is written for business users. None of the workflows require programming: you work with a conversational assistant, Word and Excel.
Any serious general-purpose conversational assistant. The method is deliberately designed not to depend on a platform: what it teaches — which process to tackle, how to measure it, where to put human review — doesn't change when the tool changes.
No. It's six working documents: guides, prompts, editable templates and a spreadsheet. There's no video and no platform to learn.
Yes. The ZIP contains two folders, English and Spanish, each with all six documents. It isn't a partial translation or a summary: it's 120+ pages per language, with the example currency adjusted in each version.
No — and be wary of anyone who guarantees it. The result depends on the process you choose, the quality of your data and how far your people adopt it. What the kit does is give you the method to identify it, test it on 20 real cases and measure it against a baseline, so that on day 7 you have a number rather than an impression.
From public studies with a control group, cited by author, year and URL inside the product. The reduction bands on each case are planning ranges built from those studies, not results measured at your company: the kit says so explicitly and teaches you to replace them with your own measurement.
The licence is internal use for one company: you can adapt, customise and print the documents inside your organisation. It doesn't include resale or public redistribution of the kit or its components. If you're a consultant and want to use it with several clients, write to me before buying.
After payment you get the ZIP download link automatically. There's nothing to wait for and no platform to register with.