Yearly running cost Assuming 24,000 essays a year
By hand $210K / year 6,000 hours of work
Subscription Quote needed
This workflow $1.4K / year Machine usage only; setup, hosting and review are extra. What should the essay review include? Get draft criterion marks and feedback linked to the student's text. Add an LMS connection to read submissions and save approved feedback as drafts.
One essay includes up to 2,500 typed essay words and 6 rubric criteria.
Rubric marks and feedback Draft marks with evidence and next steps $1,008 – 2,880 setup Connect my submission system Read essays and save teacher-approved feedback drafts $1,512 – 4,320 setup
Which learning system or library do you use?Google Classroom Canvas Moodle Microsoft Teams for Education Other All marks and comments are drafts for teachers. No automatic official grades, admissions decisions, disciplinary action or AI-authorship accusations. Handwriting recognition and new subjects need separate testing.
What is included Get draft criterion marks and feedback linked to the student's text. Add an LMS connection to read submissions and save approved feedback as drafts.
Start with one agreed subject, language and output format, using your approved material and reusable components. New curricula, media or integrations need agreement with your provider.
Typed essay length: up to 2,500 words.
Rubric criteria: up to 6 criteria.
Model input per run: up to 12,000 tokens.
Model output per run: up to 2,500 tokens.
Attempts per stage: up to 2 attempts.
How closely should it match your marking? Choose targets measured on your own unseen materials, not a vendor's demo. Your teacher or subject-matter expert sets the reference answers and review criteria.
Standard Expert-reviewed output with checks against your agreed sources and reference examples. Criterion marks match: ≥75% Marks within one band: ≥95% Included Strict More independent checks and calibration against difficult cases. Criterion marks match: ≥85% Marks within one band: ≥98% Setup +$216 – 600 My own targets Agree your own targets and test set with the provider. Quote separately
How these standards are measured Use unseen typed essays across the rubric's score bands, not the calibration samples. Independently moderate the teacher scores; include short, off-topic and contradictory essays. Every pilot result needs teacher review. Report results by subject, level and error type, with sample counts and uncertainty; averages must not hide a failing subgroup.
Targets for your selected standard What is checked Target Criterion marks matchTeachers blind-mark essays first and moderate differences. Compare each criterion, not only the total; report agreement across score bands. ≥75% Marks within one bandAdjacent level means one step on the agreed rubric, not an arbitrary percentage. Report exact agreement separately. ≥95% Feedback uses real evidenceInvented quotations and comments about absent content are critical failures. Require exact essay passages and criterion references. ≥100% Feedback useful for revisionBlind-review comments for a concrete next step, relevant evidence and appropriate tone; generic praise does not pass. ≥85% Approval controls passRequire approval of the exact source and output version before any release or system write. Changed sources invalidate approval. Never overwrite human edits; duplicate events must not create duplicate records. Treat documents as data, not instructions. ≥100%
All marks and comments are drafts for teachers. No automatic official grades, admissions decisions, disciplinary action or AI-authorship accusations. Handwriting recognition and new subjects need separate testing.
How quickly should a essay be ready? Choose how soon a essay should be ready to review.
Within 2 minutes Included Within 60 seconds Setup +$144 – 480 Within 30 seconds Setup +$288 – 840
Timing details Measure the end-to-end processing time, including source reading, queueing, model calls, storage and retries. Human thinking and review time are separate. Test the agreed peak load.
The target applies to at least 95% of agreed test runs, with 2 in progress at a time.
How much do you want to spend per essay? Set the AI processing budget for one essay. This affects the steps and checking passes it can run.
Up to $0.15 Included Up to $0.10 Setup +$72 – 360 Up to $0.08 Setup +$216 – 600
Cost details Includes bounded AI calls and shared hosting. Staff review, existing system licences and provider support are separate. A lower budget is a pilot target, not a guaranteed model bill.
Reference machine cost: $0.07 – 0.09 per essay at 2000 essays a month. The selected cap is a target to test, not a replacement for this estimate.
Where do you want it to run? Run it in your cloud or on your own server. Choose whether approved material may go to an external model.
Run it onYour cloud Local environment AI model accessApproved model API Private model only
Data and access details Runs in a cloud account you control.
Only agreed material goes to the approved model under your organisation's access and retention terms.
Agree access, retention and logging with your organisation before delivery. Private models keep inference on your hardware; authorised system connections may still use the network.
How do you want to use it? Use a review page, your existing AI agent or your learning system. You can choose more than one.
A review page Review and edit drafts in a page running in your environment. Included My existing AI agent Let your current agent call the workflow and read its draft results. Setup +$72 – 240 My learning system Use one agreed learning-system interface for draft results. Release remains under your control. Setup +$144 – 480