An automation pack for one real business, with a person approving every send
The capstone project for AI Agents for Business Workflows: Automate Five Back-Office Jobs on Free Tools, with a Person Approving Every Send — a free AI & Emerging Tech course. Pass it and you earn a certificate anyone can verify.
- Most courses free to start
- 10 steps
- Pass mark 65%
- Graded within minutes
- Beginner-friendly; a phone is enough to learn and to run a finished automation; writing the script and setting the triggers need a laptop or borrowed computer, in two sessions booked with the business owner, who is present and signed in to the owner's own Google account
The project unlocks once you complete the lessons. The workbook is free to download now so you can see exactly what is expected.
What you will submit
One shared Google Doc link with the task audit and five-workflow map with budgets, Build A and Build B described and tested on sample data, the checkpoint and alert design, the time-saved sheet, the data and spam compliance lines, the test sheet and handover sheet, and the priced offer with a Skill Market listing draft; plus a write-up (5,000 characters at most) that pastes the words proving each part.
What to do, step by step
- Choose one real business: your own, or one whose owner has agreed in writing. If you have neither, use your own automation service, and make a separate free Google account for that service to build in, never your personal one. Build everything in the business's own Google account (never your own), with no card and no paid plan; every computer step is done in a session booked with the owner, who is present and signs in to the owner's own account. Test only with your own email, your own phone and made-up customers: no real customer may receive anything from your builds. Put no customer's personal data in your Doc or write-up (write a customer as 'Customer C' and a number as 080X XXX XXXX), and never write a password, key, login code or recovery link anywhere.
- Part 1. The task audit and the five-workflow map: a table of the business's routine jobs from a two-week time log (or, if the owner could not keep one, at least five timed occasions of each job that you watched, labelled 'watched sample, not a two-week log'), with a decision for each and at least two 'never automated' lines; then a map of each of the five workflows (enquiry to first reply; quote to invoice to reminder; order to update; the week in numbers; a staff or opening checklist) with its trigger, steps marked plain or AI, the data given to the AI, decisions, the named approver, how it knows a row is already done, what happens if it fails, and its free-plan budget (your busiest day against the limit).
- Part 2. Build A: the enquiry form with its privacy note and a separate, unticked offers box; the sheet; the AI-drafted first reply in an Outbox. Describe each screen in words and give the results of your tests: one with an email, one with a WhatsApp number, one with the AI key removed for a moment, and what address the test reply came from.
- Part 3. The checkpoint and alert design: your approval rules (who approves which messages, with which initials, and what approving means); the result of running the sender twice; the WhatsApp link tapped on two phones; and the daily 'I ran' email with its named person and what happens if it does not arrive.
- Part 4. Build B: one of quote to invoice to reminder, order updates, the week in numbers, or the opening checklist, built and tested on sample data the same way, with its trigger, its templates or prompt, and at least two test results.
- Part 5. The time-saved sheet: your two-week before log, then a trial on the same ten sample rows for each job you built: the owner does the rows by hand the old way, timed, then the same rows through your build doing only the person's part (reading each draft against its row, ticking and initialling, sending by hand to your own contact), timed. Nothing real is timed and no real customer is involved. Label it 'a trial on sample data, not a saving', show minutes each as well as totals and your maintenance time, and claim nothing beyond what you measured.
- Part 6. The data and spam compliance lines: the privacy note, how consent for offers is recorded and how STOP is honoured, the transfer record (where each piece of data goes, why, and exactly what is sent, with the minimum going to the AI), the processing note between you and the owner, the NDPC question (two facts and where to ask) and your one-line breach plan; if Build B is the checklist, the notice to staff.
- Part 7. The test sheet for both builds (run twice, no initials, own contact and, where there is an AI step, AI fails, each with its result) and the handover sheet: what it does, what it never does, whose account it is in, where the key lives, who approves, who gets alerts, what is safe to run again, the limits with the date you checked them, the records kept, what to do if something leaks, and a 'verified on' line.
- Part 8. The priced offer: the build worked from your own estimated hours and your own hourly figure, the deposit, the monthly care and what it covers (including re-checking the free plans' limits), three things you do not promise, and a Skill Market listing draft for one build.
- If you could reach no computer during the course, build every part a phone allows (the form, the tabs, the formulas, the Outbox, the WhatsApp link, the templates), write the computer steps you would have followed and what stopped you, and say so at the top of Parts 2 and 4. Put everything in one Google Doc (try it on your phone; if a screen will not let you continue, use a computer), set sharing to 'Anyone with the link' as a viewer, and submit the link. In the write-up box (5,000 characters at most) paste the words that prove each part, because your write-up is what gets marked: three rows of your audit with their minutes, your enquiry map in its six boxes, one line for each other workflow (trigger, approver, never-twice mark) and your busiest day against the limit; your AI prompt's 'Do not' sentence, one test draft with your check of it, and the address the test reply came from; your approval rules in one line, your run-twice result, your two-phone tap result and the first and last sentences of your 'I ran' email; which Build B you made, its trigger and two test results; your before and trial totals and minutes each, and the maintenance minutes; two rows of your transfer record, your offers wording and your NDPC two facts; your handover sheet's 'never does' and 'alerts' items and its 'verified on' line; and your offer's working with the three things you do not promise.
Files to work with
Project workbook (fill in, then submit)The whole brief, an evidence checklist, the grading rubric as a self-check and the link-sharing steps in one file. Opens in Word, Google Docs or WPS.Word · 9 KBHow it is graded
| Criterion | Weight |
|---|---|
| The audit comes from real timing (or a labelled watched sample) and rules out money, bulk promotions, deletions and decisions about people; the enquiry map has all six boxes (steps marked plain or AI, the minimum data given to the AI, a named approver, a never-twice mark, a failure path) and a free-plan budget, and the other four maps are named with trigger, approver and never-twice mark | 20% |
| Two workflows are built in the business's own account and tested on sample data with the learner's own contacts: the enquiry build and one Build B, each described in words with real test results (including run twice and, where there is an AI step, the AI failing). A learner who could reach no computer earns credit for the phone-built parts and a clear note of what stopped them, but not for runs that never happened | 25% |
| The checkpoint works as designed: only named people approve, with initials; the sender sends approved, unsent messages once; WhatsApp is sent by a person; a daily 'I ran' email goes to a named person; and the list of what the automations never do is clear | 15% |
| The time-saved sheet shows the before log and a trial on the same sample rows (the owner's hand-done run against the build), with minutes each, totals and maintenance, labelled 'a trial on sample data, not a saving', with no real customer timed and nothing claimed that was not measured | 15% |
| The compliance lines, handover and offer fit the business: a privacy note and separate offers consent, a transfer record with the minimum sent to the AI, a processing note, the NDPC question and a breach line; a handover sheet a stranger could follow; and an offer worked from the learner's own hours, with monthly care and what is not promised | 25% |
You need 65% overall to pass. A failed submission comes back with feedback and can be revised and resubmitted once free; further attempts come with the All-Access Pass or as a paid retake.
Before you submit: make your link public
If your work lives in Google Drive or Google Docs, open Share → General access and change “Restricted” to “Anyone with the link” (Viewer). Then paste the link into a private browser window: if it opens without sign-in, you are ready. A private link cannot be graded — it is the single most common reason a good project fails.
Frequently asked questions
Do I have to finish AI Agents for Business Workflows: Automate Five Back-Office Jobs on Free Tools, with a Person Approving Every Send before submitting the project?
Yes. The submission screen opens once every lesson in AI Agents for Business Workflows: Automate Five Back-Office Jobs on Free Tools, with a Person Approving Every Send is marked complete. The lessons are where the methods, the local context and the worked examples the project depends on are taught.
How is the project graded?
An examiner scores each rubric criterion from 0 to 100 and weights them as shown on this page; you need 65% overall to pass. Most submissions are graded within minutes and you get written feedback on what was strong and what to improve.
Can I resubmit if I fail?
Yes. A failed project comes back with feedback; revise the weak parts and resubmit. A passed project is final — the certificate is issued and cannot be re-rolled.
What do I actually submit?
A write-up of what you did (under 5,000 characters) plus a public link to your work — a Google Drive folder, Google Doc, spreadsheet, GitHub repository or video. The link must open without sign-in; a private link cannot be graded. The free project workbook on this page walks you through all 10 steps.
Is the certificate real?
Yes. Passing this project issues a certificate with a unique verification code. Anyone — an employer, a client, a school — can open the verification page and see that the certificate is genuine and which project earned it.
More AI & Emerging Tech projects
- Your personal prompt pack for real work
AI Prompting: Get Real Work Done with AI, Safely and Honestly
- An agent pack for one real chore
Agentic AI: Build and Supervise AI Agents That Take Action, Safely
- Your first robot, tested honestly: a build-and-test log
Robotics: Run and Test Your First Robot on a Simulated Bench
- A live small-business website, handed over properly
AI Website Building: Plan, Build, Publish and Hand Over a Real Small-Business Website
Ready to earn this certificate?
AI Agents for Business Workflows: Automate Five Back-Office Jobs on Free Tools, with a Person Approving Every Send is self-paced. Finish the lessons, complete this project, and the certificate is yours to share.
Start learning free