Lesson 2.6: Exponential Intelligence, practical task
The skill being assessed is not "using AI"; it is reviewing what the assistant wrote before anything reaches the site. In this task each learner uses the assistant on their own article, keeps what is right, rejects what is not, and shows the trainer where every draft went.
Note
This is the trainer's own assessment, not an official certification. Use a training installation with the assistant switched on and a provider the trainer controls: a local model, or a provider key with a small budget and limit. Never put personal data into prompts, and never use a live site.
At a glance
| Course | Assessments and practical exercises |
| Module | Module 2: Assessing the editorial courses |
| Duration | 65 minutes (5 briefing, 40 task, 20 scoring) |
| Level | Intermediate |
| Audience | Learners at the end of the Exponential Intelligence course |
| Prerequisites | Lesson 2.5; an Editor account with the assistant's policies on a training installation |
Learning objectives
Learners who pass this task have shown that they can:
- use field actions in the edit form and accept or reject each suggestion;
- check facts and tone of a suggestion before storing it;
- let the assistant write a draft from the Intelligence tab and find it again;
- discard a draft they do not want;
- (administrators) read the EI: Summary page and explain its sections.
Trainer's notes
How to run it
- Before: switch the assistant on as described in
expai: Exponential Intelligence on the training
installation: the extension active for the admin siteaccess, the key in
settings/override/expai.ini.append.php, the tables created, the policiesexpai/writeandezjscore/callfor the learners' role, and theexpqueueandexpaicron parts run by hand when needed. Set limits per user and per day. Prepare one article per learner with a deliberately weak intro and one factual error in the body (for example a wrong opening date), and an image without alternative text. - Briefing: the task is about review. Say that a suggestion stored without checking is scored as an error.
- Scoring: read each learner's article, its versions and the drafts list in the Intelligence tab.
Common questions
The model is slow or unavailable. Stop the learner's clock and give the time back, as for any broken lab (Lesson 1.5).
Different learners get different suggestions. That is expected. The rubric scores the review, not the wording the model produced.
Pitfalls
- Cost surprises. Watch the cost line at the top of the edit form during a rehearsal and set limits before class.
- Learners pasting their own data into prompts. Say before the task: only the training article's text.
Hands-on exercise
The task, as handed to learners. Use your training installation and your Editor account.
- Open the article your trainer prepared for you and press Edit. Read the box at the top of the form: write down which service is used and what a click costs.
- On Intro, use Improve (or Rewrite). Read the suggestion. Accept it only if it is correct and keeps the meaning; otherwise reject it and write the intro yourself.
- Read the Body. It contains one factual error your trainer planted. Use Shorten or Rewrite on the body if you like, but make sure the error is corrected in the version you keep.
- Press Store draft and exit. Do not publish yet.
- Open the image without alternative text, edit it and let the assistant suggest an alternative text. Correct it if needed and publish the image.
- On your article's page, open the Intelligence tab and ask: "Write a short teaser article for the newsletter based on this article." Look at the steps it shows.
- Find the draft it wrote in the list of drafts on the tab. Open it, read it, and decide: keep it as a draft, or discard it. Write one sentence why.
- Go back to your own draft from step 4, check it once more, and publish it.
- (Administrators only) Open EI: Summary. Write down the names of its sections and, for one provider, its region and retention.
Expected result
- The article is published with a correct intro and without the planted error; its Manage versions page shows the new version.
- The image has a sensible alternative text.
- The Intelligence tab lists the teaser draft as kept or discarded, matching the learner's sentence.
- The learner's notes name the service and cost from step 1, and (administrators) the sections of EI: Summary: Master switch and surfaces, Providers, Lanes and their retention, Data policy, Limits, budget and usage this month, Tools that reach outside the site, What this site keeps, Roles with Exponential Intelligence policies.
Scoring rubric
| Criterion | 0: Not shown | 1: Partly | 2: Mostly | 3: Fully | Evidence |
|---|---|---|---|---|---|
| Review of suggestions (must) | Planted error published | Error found, but a wrong suggestion kept elsewhere | Error fixed, tone or meaning slightly off | Error fixed, every kept suggestion correct | The published article |
| Using field actions | None used | One action, result not checked | Actions used, one suggestion stored without reading | Actions used and each decided deliberately | The learner's account, the versions |
| Alternative text | Missing | Generated, not checked, wrong | Present, vague | Present and describes the image | The image's page |
| Drafts from the tab | No draft made | Draft made, not found | Found, no decision | Found, kept or discarded with a reason | The Intelligence tab, the note |
| Cost and policy awareness | Nothing noted | Service only | Service and cost | Service, cost, and (administrators) the summary sections | The learner's notes |
Pass: 10 of 15 points or more and Review of suggestions at 2 or 3. For editors, score the last criterion on service and cost only.
Quiz
Debrief questions to ask after scoring.
- Who published the article in step 8: you or the assistant?
- Why was the planted error the most important part of the task?
- What happened to the published article while you worked in steps 2 to 4?
- Where would an administrator see that you used the assistant today?
Answers
- You. The assistant never publishes on its own (expai).
- A model can write fluent text that is wrong; the editor remains responsible for what is published.
- Nothing changed until you published; suggestions change nothing until accepted and stored (Creating and editing content).
- In the audit, channel
ai, and in the usage log, which holds user ids only (expai).
Further reading
- expai: Exponential Intelligence: what to switch on before class.
- Running an assessment session: broken labs and lost time.
Related pages
- Lesson 2.5: Exponential Intelligence, written test
- Module 2: Assessing the editorial courses
- Assessments and practical exercises: course overview
- Trainers: start here
- Free documentation: this page is free to copy, translate and improve under the GNU FDL 1.3.
Become an Exponential Partner
Partners share their information, provide funding, contribute back and develop the products with us, and that is how Exponential grows. A partner website and dashboard are planned for mid-2027: please become a partner today to speed up our growth rate and land new features faster.