For L&D, enablement and engineering managers
Most of the English your team reads is now written by AI.
Ticket summaries, PR descriptions, call recaps, variance commentary, candidate summaries. It is longer than what a colleague would have written, denser, and it arrives faster than anyone can read it. Your existing English training was built for writing emails and speaking in meetings. This is a different skill.
Reading level of typical LLM output. Your team was hired at roughly B2. [21]
The whole program, in the working day.
A pilot, and a report at the end you can act on.
Ten minutes a day. One text. Ten questions. A chart.
Each seat gets a three-minute benchmark, then one generated text a day in their own field, with comprehension questions in their first language. Some texts contain a planted error, so the program measures care as well as speed. Every fifth text is read with no help at all — that unassisted number is the one your report is built on.
Start with twelve weeks and a report.
Week 0
Domain verification, SSO, invitations, groups. About an hour of admin.
Weeks 1–12
Ten minutes a day per seat. A weekly email to the admin: seats active, team medians, nothing individual.
Week 12
The report: team medians and spread for no-help speed at benchmark, Day 10 and the last checkpoint; the share of seats that gained 20% or more; catch rate before and after; the instrument and its limits.
Employees join programs they trust.
What an admin can see
- Seat status: invited, active, inactive
- Team and group medians from checkpoints only
- Aggregates for groups of five or more
What an admin can never see
- A learner's daily readings or their speed on any one text
- Any text a learner pasted in themselves
- A name next to a number
- Gloss taps, translations used, or error marks
A seat ending never deletes a learner's account or record. People keep their results if they leave.
Borrowed from reading research. Tested by us.
The method is assembled from published work on reading rate, repeated reading, reading-while-listening and L1 glossing — not invented for this product. Where we are running our own experiment, we say so and we mark it.
Everything procurement usually asks for, in one place.
Data
DPA, subprocessor list, EU (Frankfurt) or US (Virginia) hosting by billing country, works-council annex.
Access
SAML SSO, SCIM provisioning, domain verification, per-seat invoicing in your currency.
Exit
Export everything as CSV or JSON at any time. Seats end without deleting learner records.
What a faster reader is worth, in minutes.
We will not put a currency figure in your business case, because we do not know your salaries. This is the arithmetic we would use, with your numbers in it:
| Input | Where it comes from |
|---|---|
| Words of AI-written English read per person per day | Estimate it, or measure it for a week |
| Effective speed at benchmark, and at week 12 | Measured, unassisted, in the report |
| Minutes saved per person per day | Words ÷ old speed − words ÷ new speed |
| Cost of a missed number or condition | Yours. The catch-rate change is in the report. |
We would rather hand you the formula than a number we made up.
Questions
Talk to us first
Tell us the team and the languages. We will come back with a pilot scope, a price per seat and the DPA — not a demo booking.
Sent.
We have your request and will reply within two working days, from a person.