Every experienced regulatory affairs leader carries a private mental model of what the agency will ask. Before a submission goes out, they can tell you: expect stability questions on the new site, expect a pushback on the dissolution spec, expect clinical to probe the subgroup imbalance. That instinct is real expertise — and it is also trapped in a handful of heads, applied inconsistently, and unavailable to the team at 2am while they assemble the dossier.

Predictive HAQ Readiness in DnXT Planner turns that instinct into an explainable, data-grounded capability: before you submit, the platform forecasts the health authority questions your specific submission is most likely to receive, organized by discipline, with recommended preparation actions for each.

Where the Predictions Come From

The feature is grounded in an analysis of more than 1,000 real health-authority questions from actual regulatory programs. Mining that corpus produced two foundational findings:

  • Questions concentrate. Agency questions are not uniformly distributed across the dossier. They cluster heavily in predictable territories — stability, specifications, impurities, manufacturing changes, clinical safety analyses — with long-tail questions being genuinely rare.
  • Structure predicts discipline almost deterministically. The CTD section a question targets maps to the responsible discipline with 98–100% consistency. This sounds obvious, but its implication is powerful: a submission’s content profile — which CTD modules and sections it contains, and what kind of application it supports — is a strong statistical predictor of the shape of the question set it will attract.

Predictive HAQ Readiness combines your submission’s tracker profile (application type, submission type, therapeutic context) with the CTD modules actually present in your plan, and projects the corpus base-rates onto that profile. The output is not a vague risk score — it is a concrete list: these disciplines, these question themes, this relative likelihood.

What You Actually See

For an upcoming submission, the readiness view presents:

  • Predicted question themes grouped by discipline — CMC, clinical, nonclinical, biostatistics, labeling — so each functional lead sees their own exposure at a glance
  • Likelihood indicators for each theme, derived from corpus base-rates for submissions with your profile
  • Recommended response preparation — what a strong answer typically contains, so teams can pre-draft while the source authors are still available
  • Preparation actions you can convert directly into tracked tasks in the submission plan

Because the predictions are driven by base-rates and structural mapping rather than a black-box model, every prediction is explainable. Click into a predicted theme and you can see why it appears: submissions with this profile historically attract this class of question at this rate. Explainability is not a nice-to-have in regulated environments — it is the difference between a tool RA leadership will actually use and one they will politely ignore.

Why Pre-Submission Beats Post-Submission

The economics of answering agency questions are brutally asymmetric. After submission, a question arrives with a clock attached: teams reconstruct context, chase authors who have rotated to other programs, and draft under deadline pressure. Industry experience puts the cost of a full response cycle at roughly 500–900 hours of team effort, and each additional review round adds months to approval.

Before submission, the same preparation is dramatically cheaper. The authors are still on the program. The data is still fresh. The response can be drafted calmly — or better, the anticipated weakness can be addressed in the dossier itself, converting a would-be question into a paragraph the reviewer never needs to write. Teams using predictive readiness aim to shift review dynamics from three rounds to two — and on a major application, eliminating a review round is worth more than almost any other operational improvement available to a regulatory team.

From Prediction to Preparation

Predictions only matter if they change behavior. Predictive HAQ Readiness is wired into the rest of DnXT Planner so that each predicted theme can become work:

  • Convert a predicted question into a pre-drafted response using the same grounded drafting workbench used for live agency questions — retrieval from your approved-response knowledge base, citations included
  • Convert a preparation gap into a tracked task in the submission plan, with an owner and a date, alongside all your other workflow-managed submission activities
  • Feed lessons back: when the real questions eventually arrive, they join the corpus, and the predictions for your next submission get sharper

That last loop matters. The system is not a static oracle — it is a compounding asset. Every program your organization runs through it deepens the base-rates with your own therapeutic areas, your own agencies, your own history.

What This Is Not

Honesty matters with predictive features, so three clarifications:

  • It does not claim to predict the exact wording of agency questions — it predicts themes and disciplines with calibrated likelihoods.
  • It does not replace regulatory judgment — it gives your experts a structured starting point, and they remain the authority on what the agency will care about in your specific scientific context.
  • It does not send anything to an agency — it is a preparation tool. All AI-generated draft material stays on the draft side of a hard draft-versus-record boundary until a human approves it.

Part of a Larger Regulatory Intelligence Stack

Predictive HAQ Readiness sits alongside DnXT Planner’s guidance tracking, portfolio-aware regulatory news scoring, and correspondence management — a regulatory intelligence layer built into the same platform that handles eCTD publishing and document management, rather than a separate subscription bolted on. The submission plan that drives the predictions is the same plan your team executes against daily. If you are evaluating how AI fits into regulatory operations generally, our AI document intelligence overview covers the platform’s broader approach: grounded, explainable, auditable.

See Your Own Submission’s Question Forecast

The most convincing demo is your own program. Book a demo, tell us the application type and modules you are planning, and we will show you the predicted question landscape for a submission like yours — and what preparing for it in advance looks like.