Build Assessments You Can Actually Trust, in Minutes

Build custom, evidence-based assessments in minutes, and now author them in the AI tools you use every day.

Every talent decision your team makes—who to hire, who to promote, who is ready for a critical project—is only as good as the evidence behind it. Yet building a custom assessment you can actually trust used to take your experts weeks. 

Compose is changing that. During beta, teams used Compose to turn a plain-language description of what they wanted to measure into a complete, quality-reviewed assessment, built in a conversation. Today we are taking the next step: Compose is generally available, with a set of enhancements and a second way to author.

Any AI can write a quiz. Compose makes it count.

AI is good at writing questions that read well. But questions that sound right are not the same as questions you can trust to measure the right thing. Compose adds a layer of measurement science on top of the AI. The AI drafts the questions, and Compose checks each one against Workera's standard before it is saved, so every assessment is evidence-based and holds up when a decision or an auditor challenges it.

Those checks are built into Compose itself, so they apply the same way for every author, on every assessment. Every question maps to a defined skill. No question tests two skills at once. Every skill gets enough coverage. Before a draft lands for your final review, a quality review agent checks it for common construction flaws and the more nuanced best practices that are easy to miss. It encodes the standard item-writing rules, but goes further: our team of experts has distilled decades of combined experience, building hundreds of assessments and tens of thousands of questions, into the checks it runs. 

The result is measurement you can stake a decision on, built at conversation speed.

What's new in general availability today?

Compose now builds the whole instrument with you, not just the questions:

• The full assessment, start to finish. Compose drafts the skills framework, the proficiency levels, the assessment design and lead scenario, and the questions, all in one guided conversation with Elo, our in-product authoring agent. From there, you can link the assessment to custom learning content, so measurement and development live in one place.
• All seven question types. Single selection, multiple selection, short answer with a scoring rubric, ordering, matching, grouping, and visual, plus code answers. Any of these can be artifact-based, where participants review a real work product such as a document, dataset, or diagram and answer questions about it, the way they would on the job.
• Built-in quality review agent: A dedicated review agent runs on every assessment and flags the issues AI tends to miss in its own work, so what you review is a refined draft rather than a first draft. 
• A validity view on request. Compose can produce an evidence-centered design map that lays out what the assessment claims to measure, the evidence each question provides, and how it is scored, so the justification already exists when a result is questioned.
• Right-sized length. Test length and duration are modeled from real completion-time data, so you can fit an assessment to a target testing time and respect participants' time.
• Multiple languages. View and test-run questions in multiple languages to see how the assessment reads for participants before it goes live.
• A living instrument. Once an assessment is live, Compose keeps measuring how questions perform on real participants, watches for drift, and helps you sharpen weak items over time. The quality compounds with use.

Two ways to author, one standard

Compose now meets your team where it already works. Choose the path that fits your setup. The measurement standard is identical on both.

In-app on the Workera platform. A guided conversation with Elo and a live workspace panel that shows the framework, proficiency levels, design, and questions taking shape as you build. Nothing to install, nothing to connect.

In your own AI tools, through the Workera connector. Prefer to work inside the assistant you already use? Connect Compose to Claude and author there, pulling in context from the documents and sources your assistant can already reach, such as a role definition or a project spec. Your work saves back to Workera as a draft to finish and publish. More assistants will follow over time.

Whichever path you choose, you get the same enforced rigor and the same defensible result.

Get started today

Compose is available now to Workera enterprise customers. Interested in learning more? Request a demo today! 

Describe what you want to measure. Compose builds an assessment you can defend.