run four 75-minute evenings, one week apart. Give each evening one job: start together, unstick in a lab, apply AI to a real task, then finish and score. Replace "self-paced" with a written deadline for every pair of courses. Keep the same faces in the room each week, and pair every learner with a check-in buddy. Learners still do most of the course on their own, roughly 2 to 4 hours a week between evenings.
Version 1. ed.cafe has not run this plan with a cohort yet. Circle #1 is still forming. Treat every timing below as a starting estimate, and expect a revised version after the first cohort.
Outcome, time and roles
Outcome: each learner leaves with the Google AI Professional Certificate (or a dated plan to finish it) and one scored work sample.
Learner time: Coursera's program page headlines 8 hours, but its per-course estimates add up to 13: 3, 1, 1, 1, 2, 1, 2 and 2 hours. Plan for 13. Tell learners that number on day one.
Facilitator time: four evenings of 75 minutes, plus about 30 minutes of prep and follow-up per week.
Roles you need:
- One facilitator who has finished the certificate. Jobs for the Future's implementation field guide describes one college hiring "career coaches who have completed a Certificate."
- One license administrator. At a MassHire Career Center this is the SPOC named under DCS 14-623, who sends invites and runs Coursera progress reports.
- A second scorer for Evening 4. Any colleague will do.
Use this plan if, do not use it if
Use it if:
- You have 6 to 30 adults with licenses for the same certificate.
- You can hold the same room, or the same video link, at the same hour for four weeks.
- Learners can reach a laptop or tablet with a keyboard. Phones make the labs hard.
Do not use it if:
- Learners have not yet decided they want the certificate. Have the goal conversation first.
- You cannot secure four consecutive weeks. A broken calendar defeats the design.
- Your funder expects a WIOA credential outcome and your state has not ruled on it. Read the TEGL 10-16 analysis first.
Why this shape
The evidence favors structure and human contact over reminders. The JFF guide calls a course calendar with milestones "critical for persistence and completion" in asynchronous programs. It reports South Texas College reaching "a completion rate of at least 85% in one semester" when the certificate counted for half a course grade. A large MOOC study led by René Kizilcec found that light-touch nudges shrank at scale: the effects were "an order of magnitude smaller," per Cornell's summary of the 2020 PNAS paper. So this plan leans on deadlines, a room and a person, not on emails.
Pre-flight checks (one week before Evening 1)
- Confirm every license. Ask the SPOC for the invite list. Verify each learner has accepted the invite. Failure trap: an unaccepted invite looks like a learner who quit.
- Test the room. Check Wi-Fi with a full group's worth of devices. Check power outlets.
- Print the pacing calendar below with real dates.
- Print the rubric. Use the four criteria in the 12-point rubric guide.
- Send one message to each learner: date, room, "bring your device and charger," and the 13-hour estimate.
Verification: every name on the roster shows an accepted license. Do not proceed to Evening 1 with unknown license status.
The pacing calendar
Fix these deadlines in writing. Hours are Coursera's own estimates.
WEEK OPENS WITH COURSES THIS WEEK EST. HOURS CUMULATIVE DUE
1 Evening 1 1 AI Fundamentals, 2 Brainstorming 3 + 1 = 4 4 before Evening 2
2 Evening 2 3 Research, 4 Writing 1 + 1 = 2 6 before Evening 3
3 Evening 3 5 Content Creation, 6 Data Analysis 2 + 1 = 3 9 before Evening 4
4 Evening 4 7 App Building, 8 App Deployment 2 + 2 = 4 13 7 days after Evening 4Week one carries the heaviest load because Course 1 is the longest. Expect your biggest drop there.
Evening 1: start together
Job: every learner leaves with Course 1 started.
- 0 to 10 min. Welcome. Names. Pair each learner with a check-in buddy.
- 10 to 20 min. Everyone logs in to Coursera. Everyone opens Course 1. Verification: you see Course 1 on every screen. If not, the SPOC fixes it now.
- 20 to 55 min. Work on Course 1 in the room. Walk the room. Answer login and navigation questions only.
- 55 to 65 min. Each learner writes two deadlines on the calendar: Courses 1 and 2 before Evening 2.
- 65 to 75 min. Buddies swap a phone number or email. Each learner names one task from their job, or the job they want, that AI might help with.
Exit check: Coursera shows progress above zero for every attendee.
Evening 2: the lab
Job: clear the exercise that stopped each learner, and catch a first AI error.
- 0 to 10 min. Progress round. Each learner says which course they are on. No judgment.
- 10 to 30 min. Unstick. Learners open the exact exercise that stopped them. Pair a learner who is ahead with one who is stuck.
- 30 to 55 min. First hallucination check. Each learner asks an AI tool a factual question from their field and requests sources. Then they check the answer against a primary source and find one thing that is wrong or unsupported.
- 55 to 65 min. Choose the work-sample task: a real task from their field, done with invented or public data only.
- 65 to 75 min. Set deadlines: Courses 3 and 4 before Evening 3.
Exit check: every learner has written down one error the AI made, or one claim they could not verify.
Evening 3: use it on a real task
Job: draft the work sample without private data.
- 0 to 10 min. Progress round.
- 10 to 20 min. Privacy rule. No customer names, patient records, account numbers or confidential figures go into any AI tool. Use invented values.
- 20 to 55 min. Draft the work sample. Each learner records what they asked, what they checked, and what they fixed.
- 55 to 70 min. Buddies read each other's draft against the four rubric criteria. Comments only, no scores.
- 70 to 75 min. Set the deadline: Courses 5 and 6 before Evening 4.
Exit check: every learner holds a draft with at least one documented correction.
Evening 4: finish and score
Job: score every work sample and lock a finish date for the certificate.
- 0 to 10 min. Progress round. List who is within one course of finishing.
- 10 to 35 min. Finish block. Learners close to the end work on their last course. Others polish the work sample.
- 35 to 65 min. Two scorers rate each work sample on the 12-point rubric. Pass is 8 of 12 with no zeros. Record certificates and rubric results separately.
- 65 to 75 min. Each learner writes their finish date for Courses 7 and 8, no later than 7 days out. Set a follow-up check 30 days out.
Exit check: every attendee has a rubric score and a written finish date.
Failure modes and exact fixes
1. License not active on Evening 1. Symptom: a learner cannot see the course. Fix: the SPOC re-sends the invite during the session. Prevention: the pre-flight license check. Do not let the learner sit idle; pair them with a neighbor to watch Course 1.
2. The Course 1 cliff. Symptom: learners arrive at Evening 2 still in Course 1. Fix: split Course 1 into two sittings on the calendar. Use the first 20 minutes of the lab for it. Do not move the Evening 2 date.
3. A learner misses an evening. Symptom: an empty chair. Fix: the buddy texts that night. The facilitator offers one 15-minute call before the next evening. The learner rejoins at the next session on the current schedule. Do not drop them and do not restart them.
4. Private data pasted into an AI tool. Symptom: a draft contains a real name or account detail. Fix: stop. Remove the data from the draft. Rewrite with invented values. Tell the learner to follow their employer's policy on what happened. Score criterion 3 as zero if it reaches the final sample.
5. Certificate done, no work sample (or the reverse). Fix: report both counts separately. Never merge them into one completion figure.
6. Week-3 slump. Symptom: three or more learners show no Coursera activity for 7 days. Fix: the SPOC flags them the same day. The facilitator sends one text with one question: "What stopped you?" Offer a 15-minute slot.
Make-up path and post-cohort check
Use this path for anyone behind after Evening 4:
- Agree a finish date within 14 days.
- Book one 15-minute check-in for day 7.
- Score their work sample at office hours if they missed Evening 4.
- At 30 days, the SPOC re-runs the Coursera report. Record late certificates in a separate column from on-time ones.
Health check: you can state, for the whole group, day-one enrollees, first-session starters, certificates earned and work samples passed. If you cannot, fix your records before you report.
What will change after Circle #1
This is version 1. After the first cohort, ed.cafe will publish the real timings, where people stalled and what failed, then revise these agendas. If you run this plan yourself, the free ed.cafe Staff Room is where facilitators can compare notes once monthly sessions begin.
Sources
- Google AI Professional Certificate program page, Coursera, accessed 15 October 2026
- Google Career Certificates Implementation Field Guide, Jobs for the Future, 31 August 2023
- DCS Information 14-623: Google AI & Digital Skills Partnership, MassHire Department of Career Services, 27 February 2026
- Study: No single solution helps all students complete MOOCs, Cornell Chronicle, 15 June 2020


