Skip to content
Coyotiv home

Getting an AI study featureready to launch.

An education technology company had an AI study feature it could not launch. We worked across product, AI engineering, and leadership to examine the experience, address the underlying issues, and help the team move forward.

Book a callSee all work
Client
Education technology company
Sector
Education technology
Our role
AI product & engineering transformation
Today
Live, after product and leadership aligned

The starting point

The study product generated notes, quizzes, and mock exams from learning material. But the feature could not get into production.

Testing the product as a student exposed a central problem: mock exams did not reliably reflect the uploaded document. Other questions involved the generation approach, the user experience, evaluation, and how the team owned the work.

The outcome

Leadership supported an immediate experiment giving uploaded documents more weight and welcomed the preview approach. Product and leadership aligned around the changes, and the feature went live.

The work connected the behavior of the AI product with the decisions and responsibilities needed to move it into production.

How it went

Work through the experience with the people building it.

  1. Experience

    Start with the student’s experience

    We tested the product as a student and brought specific failures to product and engineering. That made the discussion concrete: what the student expected, what the product produced, and where the two diverged.

  2. Generation

    Examine generation and evaluation

    With AI engineers, we reviewed context handling, generation design, and the tradeoffs between quality, latency, and cost. Recommendations included generating questions individually, validating them with bounded retries, and evaluating mock exams with input from educators.

  3. Product

    Connect product choices to engineering work

    We investigated experience gaps such as streaming display and course selection. We also discussed prioritizing uploaded documents and showing a five-question preview before generating a full exam.

  4. Team

    Address how the team worked

    The engagement covered development setup, team ownership, and AI-assisted development. We recommended small teams that could own the whole experience and worked with leadership and product on the direction.

The result

The result

Leadership supported an immediate experiment giving uploaded documents more weight and welcomed the preview approach.
Aligned
Product and leadership aligned around the changes, and the feature went live.
Live

More work

More work.

See all work

Contact

Book a call.

Free, 30 minutes, with an engineer. Tell us what you are working on.

  • StartupsFree architecture review
  • EnterprisesAI review
  • SchoolApply for the next cohort
hello@coyotiv.com
What brings you here?
How did you hear about us? Optional

An engineer replies by email.