Teech AI for faculty
We interviewed faculty at leading universities about where their week actually goes, and identified the pain points that come back every time. The teaching itself is fine. The work around it is not: repeating emails, exams marked one by one, and material that never quite fits the class in front of you.
Three building blocks. Pick one to see it working.
Save at least 2 hours a day on email
Replies drafted in your own words, routine emails answered or filed by your rules, and one WhatsApp summary at the end of the day.
Check exams fast and bring every student up to speed
A grade and feedback for every script, then a class dashboard: who is falling behind, what they should work on, and what to tell the family.
Tailor every lesson to the class in front of you
A course plan and material built from your books and your class, adjusted after every lesson from a one-tap WhatsApp check-in.
Save at least 2 hours a day on email
Some faculty take more than 100 emails a day, and most are repeats. Choose what the assistant does with each kind of email, then run a day and see the result.
Getting credit for work already finished
Both the publication record and the literature search are structured problems wearing the costume of knowledge work.
Check exams fast and bring every student up to speed
Put the exam, your model answer and the students' work in your Drive or the school's shared folder. Teech AI proposes a grade and feedback for every script, written answers included, then builds a class dashboard: the most common mistakes, who needs help, what each student should work on, and a draft message for the family.
Tailor every lesson to the class in front of you
Describe your class, or upload the plan and material you already have. Teech AI builds the course plan and the material in your Drive, from the books you use. After each lesson it asks you on WhatsApp how it went, and adjusts the material so the next session engages the class better.
Two conditions we were given
Both came up unprompted in every interview. Everything above is built around them, and you will see both in every demo.
The professor sends. Not the system.
One wrong email to the entire department costs more than the hour it saved. So nothing sends, posts, submits or grades on its own.
Drafts are prepared and queued. The professor reads, edits if needed, and approves. Approval is one click, and it stays a click a person makes.
The data stays where the university wants it.
The hesitation we heard was not about quality, it was about connecting a mailbox and student data to something outside the institution.
So: deployment inside the university's own cloud account, content never used for training, and each connection approved separately rather than all at once. A workflow can also run on files the professor hands over, with no live connection at all.
Where the money for this already sits
None of this needs a new budget line to start.
| Route | What we found | Fits |
|---|---|---|
| Institutional licences | The universities we spoke to already pay for general AI assistants, and for specialist legal tools on the law side. Paid AI tooling is established, so the question is which workflow, not whether. | Department wide rollout |
| Research funds | Faculty hold personal research funds, often in the region of 10,000 euro a year, and are reimbursed for subscriptions they buy themselves. No central agreement needed. | Single professor pilot |
| Procurement | Several universities run an innovation procurement unit whose job is finding and testing solutions like this, which is the route to more than one department at a time. | Multi department contract |
How we would start
Six weeks, one professor, two workflows, measured.
Pick two workflows.
Usually admin mail and exam marking, because both repeat weekly and both have a clean before and after.
Two weeks to build.
Built on that professor's real cases, their templates, their phrasing, inside the university's own environment.
One month running.
The professor reviews and approves every output. We count hours returned, drafts accepted without edits, and anything the review caught.
Then decide.
Extend to the rest of the department's admin portfolios, or stop. The pilot is sized so stopping is cheap.