codeit's biggest update in a decade: AI-assistance, not AI autopilot

AI has changed the world of verbatim coding. Work that once took hours of manually reading, sorting and categorizing responses has been accelerated dramatically.

But not all AI-assisted coding tools are created equal.

That’s one of the reasons we’ve just launched the biggest update to codeit in a decade.

The new version of codeit has been rebuilt from the ground up, with a completely redesigned coding screen and a much broader range of AI-assisted features. Researchers can now use smart theme-splitting, semantic AI search and duplicate-theme detection alongside faster filtering, more flexible codeframe editing, and the ability to compare verbatims from another wave, task or project without leaving the coding screen.

These improvements build on something that has always been a core strength of codeit: keeping the researcher in control of the coding process.

And with our new version, that strength just got better.

AI assistance, not AI autopilot

There is a lot of excitement around AI verbatim coding at the moment. And understandably so.

But look underneath the surface of many products and you’ll find that they are little more than an LLM wrapped in a simple user interface.

Send your verbatims to an AI model. Ask it to generate a codeframe. Let it assign codes. Display the results in a table. Add a few controls around the edges.

It can be useful. But it isn’t a proper human-in-the-loop workflow.

For professional researchers, that distinction matters.

Researchers have to make judgements about nuance, context, terminology and relevance. They need to decide whether subtly different responses belong under the same theme, whether a coarse theme needs to be split, where codes overlap, and whether an unexpected result warrants further investigation.

The AI can help with all of this.

But the researcher should be the one making the decisions.

The difference between AI-powered and AI-assisted

Any AI-enabled coding tool can produce an answer.

However, a good coding tool should help a researcher arrive at an answer they completely trust and understand.

That means giving researchers practical ways to investigate, review and refine what AI is doing rather than blindly accepting its output.

• Can you see the responses behind a theme and investigate them?
• Can you search by meaning rather than relying on exact keywords?
• Can you easily adjust a codeframe to capture subtlety, granularity and domain expertise?
• Can you quickly override AI autocoding to apply the unwritten rules that a generic LLM can’t know?
• Can you filter the data precisely enough to investigate a particular subgroup?

These aren’t cosmetic features. They are what lifts AI output to a deliverable standard of quality.

Without features to support all of this it’s not really human-in-the-loop, it’s just human-on-the-end-of-the-pipeline.

Keeping experts in the driving seat

This has always been central to how we think about codeit.

AI should remove repetitive work without removing professional judgement.

That’s why our new version expands AI assistance throughout the coding workflow while keeping the researcher firmly in the driving seat.

Smart theme-splitting can help identify where a broad theme actually contains multiple ideas. Semantic AI search makes it possible to find relevant responses based on meaning rather than exact wording. Duplicate-theme detection can highlight potential overlaps before they become problems.

The point isn’t for the AI to make every decision.

The point is to make it easier for the researcher to make the right decisions.

Good AI shouldn’t ask experts to stop thinking.

It should help them move faster, explore more deeply, and produce better work, while keeping them firmly in control.

Find out how to do AI correctly with a free trial of codeit