For the last year, the AI inside the MX8 Labs Research Platform has been an excellent advisor. Ask it to review a questionnaire and it would tell you where the wording was leading, which scales were inconsistent, and where the logic might trip a respondent. Useful, but it stopped at the suggestion. You still had to make the change yourself.
Today that changes. MX8 Labs Assist still reviews and advises, but it can now draft the work too: rewrite survey code, clear checklist issues, build a report, refine an insight, and fix a data-prep script in the project you have open.
That is the useful shift: less copying advice out of a chat window, more reviewing proposed changes where the work already lives.
From advice to action
The reason this matters is simple. Most of the time lost in a research project is not the thinking. It is the mechanical translation of a decision into the platform. You know the scale should be a consistent five points. You know the checklist wants soft validation on that numeric question. You know the report needs the latest tracking questions added. None of that is hard. It is just fiddly, and it adds up.
Assist closes that gap. Because it runs inside the platform, it can read the live object you are working on: the current survey, report, insight, checklist, media table, translation status, or recode script. You describe the change in plain English, and Assist drafts it for review.
A few things it can do today:
- Read and rewrite survey code, add skip logic and validation, refine quotas and branching, and insert calculated variables.
- Work through the go-live checklist, explain each issue, and draft the exact fixes needed before launch.
- Create cross-tab, time-series, lift, weighting, and verbatim reports, and update existing ones.
- Create and sharpen Insights from a research question.
- Inspect and update the data-preparation recode script after fielding.
It also does the unglamorous but valuable work of orienting you. Ask it to summarize where a project stands, what reports and insights already exist, or whether a survey is ready to launch, and it answers from the actual state of your work rather than your memory of it.
Why every change is still yours to approve
Here is the part I care about most: Assist never saves a change to your survey on its own.
It proposes. You decide. When Assist rewrites a question, fixes a checklist item, or restructures a report, it shows the proposed change and writes nothing until you review it and choose to apply it. Until you do, your survey is untouched.
A questionnaire is the foundation of a study. A single unreviewed change to wording, logic, or a quota can bias your results or break fielding in ways that are expensive to catch later, sometimes only after the data is in. The researcher who understands the objectives should always have the final say.
That review step is what makes larger changes practical. You can ask for a bold edit, check the diff, and accept or discard it without putting the live survey at risk.
The same partnership, lower down the stack
Readers of this blog will recognize the philosophy. I have written before that the future of research is not an autonomous dashboard that runs without humans. It is a workflow where the agent handles execution and the researcher handles judgment.
Assist applies that idea to the day-to-day mechanics of building a study. The agent does the typing. You do the deciding. It takes the survey programming, the report configuration, and the recode wrangling that sit between a good decision and a fielded study, and makes them faster without moving the judgment call.
This is also why Assist does not touch the numbers. Cross-tabulation, weighting, and statistical testing are still performed by the platform as configured, with significance and thresholds preserved. Assist helps you set up, interpret, and act on results. It does not recalculate them.
Try it on your next study
MX8 Labs Assist is available now inside the platform. The best way to understand it is to open a survey you are working on and ask it for something concrete: "Review this survey for leading questions and show me the fixes," or "Go through the checklist and draft the changes to resolve it," or "Create a cross-tab report for awareness by age and gender." Watch what it proposes, then apply what you agree with.
The MX8 Labs Assist documentation walks through what it can do, and the example actions page gives you prompts to start with.
We have spent years automating the slow, mechanical parts of research so that researchers can spend their time on the parts that actually require a researcher. Assist drafting changes is the next step in that work, and you stay in control of every one.