Dify
Menu

Solutions

Assessment &
Testing

AI for assessment organisations.

Hand drafting and first-pass review to AI; editors focus on the final call. Decades of editorial rigour scale without the team scaling with it.

Contact Sales
Contact Sales

Deployed in Production at.

ETS / Educational Testing Service (United States)

ETS, founded in 1947 and headquartered in Princeton, New Jersey, is one of the world's leading educational assessment organizations. It runs Dify Enterprise as its unified internal AI platform, supporting around 2,000 employees across the organization.

With a 12-person AI platform team, ETS supports 50 to 60 production-grade AI applications across test item generation, content review, quality control, and business innovation.

Using Dify, ETS built a test item quality review workflow covering grammar checks, content review, and fairness evaluation. It reduced the time from business requirement discussion to prototype delivery to around 30 minutes, helping more than 300 employees build and use AI applications directly.

This model redefines how IT and business teams collaborate: IT owns the platform, governance, and infrastructure; business teams own the application logic and innovation.

Read the full storyHow ETS Built Its In-House AI Platform on Dify Enterprise
Contact Sales

  • 50+

    Production apps on Dify Enterprise

  • 2,000

    AI platform for 2,000 staff, run by just a 12-person team

  • 30 min

    Business interview → deployable prototype

  • 300+

    Staff onboarded to build their own apps

Source: ETS, quote authorised by ETS for Dify use, 2025

“Dify's intuitive interface lets our teams rapidly design and deploy complex natural-language pipelines, improving the quality of our assessment products while reducing cost and time to market.”
Gary FengDirector of AI, ETS

Why
This Industry, Why Now.

The constraints

Assessment organisations (standardised testing providers, certification bodies, educational publishers) operate inside a particular set of constraints.

They produce high stakes content at scale, where every test item, scoring rubric or certification question has to clear an editorial review process built up over decades.

They sit on top of legacy infrastructure such as item banks, scoring pipelines and candidate data systems that nobody wants to rebuild from scratch but that increasingly slow down operational pace.

And many of them face fresh pressure from new entrants whose native technology stack is much younger, which puts a hard ceiling on how long the legacy approach can carry the business.

The role of AI

For this industry, AI gets used less as a chat assistant and more as the way to scale a high quality editorial review process without scaling headcount or backlogs.

It modernises legacy systems by wrapping them as workflow callable APIs.

And it repositions the internal AI platform as a strategic capability owned by the IT organisation.

Patterns We See.

AI-assisted item generation at scale
Assessment organisations use Dify workflows to generate test items across multiple item types, including language, mathematics and other standardised test question formats, and feed them into existing item bank systems. The pattern replaces a heavy manual authoring process and lets editorial teams focus on judgement work, not first draft production.
Multi-stage editorial review workflows
The same workflow approach handles the review process that assessment organisations have built up over decades. One stage handles language checks. Another handles content correctness and logical consistency against institutional standards. A third handles fairness inspection, checking items for cultural bias or discriminatory content. Each stage produces structured output that the human reviewer signs off on.
Wrapping legacy systems as callable APIs
Assessment organisations of any age run a substantial estate of older internal systems. The Dify approach is to wrap the underlying capabilities of those systems as APIs and call them from Dify workflows that orchestrate the modern logic on top. The institution gets the benefit of new workflow design without paying the cost of replacing the legacy core.
IT as platform, business teams as builders
In mature deployments, IT owns the runtime, governance, integration patterns and standards. Business teams (editorial, scoring, candidate operations) own the application logic for the workflows they understand best. A small platform team can support tens of production applications because the platform lets the work scale past the team itself.

Discuss Assessment & Testing Workflows with Dify.

Contact Sales