For a gene therapy developer, an FDA INTERACT meeting can provide an important opportunity to obtain early regulatory feedback before major development decisions are locked in. INTERACT, or INitial Targeted Engagement for Regulatory Advice on CBER/CDER ProducTs, is intended for early-stage development of novel therapies. FDA currently recommends considering an INTERACT when the investigational product has been identified for clinical evaluation and preliminary preclinical proof-of-concept studies have been conducted, but definitive toxicology studies have not yet been designed and conducted. [1]

For sponsors, the challenge is not simply requesting the meeting.

The greater challenge is deciding what FDA needs to know, which questions are worth asking, what evidence supports each question, and how the answers could change the development programme.

This is where AI can support preparation.

Used appropriately, AI can help teams organise evidence, identify gaps, compare regulatory guidance, structure briefing materials, and refine questions. It should not replace regulatory judgment or become an unchecked source of regulatory conclusions.

A strong workflow combines AI-assisted research with experienced regulatory, CMC, nonclinical, clinical, and quality experts.

For sponsors building an integrated pharma clinical strategy, this approach can make early FDA engagement more structured and evidence-driven.

What Is an FDA INTERACT Meeting?

An INTERACT meeting is an early engagement opportunity with FDA for sponsors developing novel therapies.

According to FDA's current guidance, the appropriate timing is generally after the sponsor has identified the investigational product and completed preliminary preclinical proof-of-concept work, but before definitive toxicology studies have been designed and conducted. [1]

This timing is important.

The purpose is not to obtain detailed FDA agreement on a complete clinical programme.

Instead, INTERACT can help sponsors obtain early feedback on important development questions while there is still time to incorporate that feedback into the programme.

FDA explains that INTERACT discussions can cover areas such as:

• Chemistry, manufacturing, and controls

• Pharmacology and toxicology

• Clinical development

• Product characterisation

• Early clinical development considerations

For gene therapy programmes, these areas can be tightly connected.

A manufacturing decision may affect the clinical product.

A vector characteristic may affect the nonclinical strategy.

A proposed first-in-human study may depend on available natural history information.

That interconnectedness makes preparation particularly important.

Why INTERACT Preparation Is Difficult for Gene Therapy Sponsors

Gene therapy programmes can involve highly specialised scientific and regulatory questions.

Depending on the programme, sponsors may need to consider:

• Vector design

• Transgene characteristics

• Manufacturing process

• Product characterisation

• Potency

• Identity

• Purity

• Safety testing

• Biodistribution

• Immunogenicity

• Toxicology

• Patient population

• Natural history

• Dose selection

• First-in-human study design

FDA's August 2026 final FAQ guidance on developing potential cellular and gene therapy products addresses common questions across regulatory review, CMC, pharmacology/toxicology, clinical, and clinical pharmacology. [2]

The challenge is therefore not a lack of information.

It is the opposite.

There can be too much information.

The sponsor needs to determine which evidence actually matters for the questions being presented to FDA.

Step 1: Confirm That INTERACT Is the Right Meeting

Before using AI to draft anything, confirm that an INTERACT meeting is appropriate for the programme's development stage.

FDA's current OTP information indicates that INTERACT is intended for early development and before definitive toxicology studies have been designed and conducted. [1]

FDA also distinguishes INTERACT from Pre-IND meetings.

A Pre-IND meeting is generally more appropriate when the sponsor has defined the manufacturing process for clinical studies, developed assays and preliminary lot-release criteria, completed proof-of-concept and possibly preliminary safety studies, and has questions concerning IND-enabling CMC, pharmacology/toxicology, or clinical trial design. [3]

This distinction should be documented internally.

AI can help create a comparison matrix:

The final determination should remain with the sponsor's regulatory team.

Step 2: Build an Evidence Inventory

Before drafting questions, create an inventory of what is actually known.

AI can help organise information from:

• Internal research reports

• Preclinical studies

• Manufacturing documents

• Published literature

• Natural history studies

• Previous regulatory interactions

• FDA guidance

• Relevant scientific publications

• Clinical development plans

The inventory can be organised into five categories:

Product

What exactly is being developed?

CMC

How is the product manufactured and characterised?

Nonclinical

What preliminary evidence supports development?

Clinical

What is known about the disease and proposed first-in-human population?

Regulatory

What guidance and previous regulatory experience are relevant?

This creates the foundation for question development.

Step 3: Use AI to Map FDA Guidance to the Programme

Gene therapy developers now have a growing body of FDA guidance to consider.

FDA's cellular and gene therapy guidance collection includes guidance on areas such as preclinical assessment, human genome editing, CMC information, rare diseases, potency, manufacturing changes, and early-phase clinical trials. [4]

In 2026, FDA has also published additional guidance covering cellular and gene therapy development.

For example, FDA issued final guidance in August 2026 addressing frequently asked questions across multiple CGT development disciplines. [2]

FDA also issued May 2026 guidance describing CMC flexibilities for human cellular and gene therapy products being developed for BLAs. [5]

AI can help create a guidance-to-programme map:

FDA requirement → Guidance → Programme issue → Existing evidence → Gap → Regulatory question

This is much more useful than asking an AI system to simply "summarise FDA gene therapy guidance."

Step 4: Create a Regulatory Question Bank

The quality of the INTERACT meeting depends heavily on the questions.

FDA currently states that the meeting package should contain clearly worded and targeted questions that directly address concerns about the product development programme. For a 60-minute meeting, FDA considers a maximum of 10 questions, including sub-questions, to be reasonable. [1]

That means sponsors need to prioritise.

AI can help transform broad concerns into specific question candidates.

For example:

Broad concern:

"We are uncertain whether our proposed nonclinical programme is adequate."

Better question structure:

"What is the sponsor proposing?"

"What evidence supports the proposal?"

"What specific feedback is needed from FDA?"

"What decision will change depending on FDA's response?"

The AI system can generate candidate formulations.

The regulatory team should decide which questions are scientifically and strategically appropriate.

Step 5: Prioritise Questions by Decision Impact

Not every uncertainty deserves space in an INTERACT meeting.

A useful prioritisation framework is:

High Priority

FDA's answer could materially change the development programme.

Medium Priority

FDA's answer could influence execution but is unlikely to change the overall strategy.

Low Priority

The question is useful but can be resolved through existing guidance or internal work.

This prevents the briefing package from becoming a catalogue of every unresolved issue.

A strong INTERACT question should have a clear relationship between:

Evidence → Uncertainty → Regulatory question → Development decision

Step 6: Prepare the CMC Section

FDA recommends that the CMC section of an INTERACT package provide a high-level description of the product, manufacturing process, proposed characterisation, and lot-release tests. The sponsor should also provide its position and justification for questions and relevant published information. [1]

For gene therapy, the CMC discussion may need to consider:

• Product description

• Vector or construct characteristics

• Manufacturing process

• Starting materials

• Process controls

• Product characterisation

• Identity

• Purity

• Potency

• Safety-related testing

• Proposed lot-release strategy

FDA's 2020 guidance on CMC information for human gene therapy INDs explains that CMC information should provide sufficient information to assure the safety, identity, quality, purity, and strength, including potency, of the investigational product. [6]

AI can help organise CMC information into a consistent structure.

It can also identify where a proposed regulatory question does not appear to have enough supporting evidence.

But AI should not independently determine whether a CMC strategy is scientifically adequate.

That requires qualified CMC and regulatory experts.

Step 7: Structure the Pharmacology and Toxicology Evidence

FDA's preclinical guidance for investigational cellular and gene therapy products provides recommendations on the substance and scope of preclinical information needed to support clinical trials. [7]

For an INTERACT package, the focus should remain appropriate to the meeting's early-development purpose.

AI can help create an evidence map covering:

• Proof-of-concept studies

• Relevant animal models

• Pharmacodynamic findings

• Biodistribution information

• Safety observations

• Immunogenicity considerations

• Dose-related findings

• Known limitations

• Outstanding questions

The purpose is not to generate a definitive toxicology strategy automatically.

In fact, FDA specifically notes that questions regarding definitive preclinical safety studies should generally be addressed in a Pre-IND meeting rather than included in the INTERACT package. [1]

This distinction is exactly where an AI-assisted regulatory workflow can be useful: it can flag questions that appear inconsistent with the intended scope of the meeting for human review.

Step 8: Build the Clinical Development Context

FDA states that the clinical portion of an INTERACT package should generally contain high-level information rather than detailed protocol design.

The package should describe:

• Disease of interest

• Target study population

• Available natural history information

• Available treatment options

• Brief outline of the first-in-human study

[1]

AI can help synthesise the relevant literature into a structured evidence map.

For example:

Disease → Natural history → Current treatment → Unmet need → Target population → Proposed intervention → First-in-human rationale

This is especially useful for rare diseases where natural history information may be limited.

FDA's guidance on human gene therapy products for rare diseases specifically discusses challenges associated with small study populations, feasibility and safety issues, and interpretation of bioactivity and efficacy outcomes. [8]

Step 9: Use AI to Compare the Programme Against Precedent

Regulatory precedent can be useful, but it must be handled carefully.

AI can help identify publicly available information about:

• Similar gene therapy products

• Comparable vector platforms

• Relevant clinical programmes

• Published FDA guidance

• Public regulatory communications

• Clinical trial designs

• Published safety findings

The objective should not be to say:

"FDA accepted this approach for another programme, therefore FDA will accept ours."

Instead:

"What similarities and differences exist between this programme and relevant precedents?"

That distinction matters.

Regulatory decisions are product-specific.

Step 10: Create a Question-to-Evidence Traceability Matrix

One of the most useful applications of AI is traceability.

For each proposed question, create a record containing:

This creates a defensible chain between the evidence and the question.

It also makes cross-functional review easier.

Step 11: Keep the Briefing Package Focused

FDA states that INTERACT briefing packages should not exceed 50 pages and that voluminous packages are discouraged. The agency can cancel a meeting if the package is grossly inadequate or lacks sufficient information to address the questions. [1]

This creates a balancing problem.

The package must contain enough information.

But it should not become a comprehensive development report.

AI can help identify:

• Repeated information

• Redundant sections

• Unanswered questions

• Unsupported claims

• Excessive background

• Inconsistent terminology

• Missing references

The regulatory team should make the final editorial decisions.

Step 12: Use AI for Consistency Checks

A briefing package can contain contributions from several functions.

CMC may use one product description.

Nonclinical may use another.

Clinical may describe the programme differently.

Regulatory may use different terminology.

AI can perform a consistency review across the draft.

Useful checks include:

• Product name consistency

• Vector terminology

• Dose terminology

• Study population

• Indication

• Manufacturing terminology

• Development-stage descriptions

• Abbreviations

• References

• Question numbering

• Cross-references

This is a relatively low-risk AI application because the system is supporting quality control rather than making the regulatory decision.

Step 13: Review Every AI-Generated Claim

AI-generated regulatory content should never be treated as authoritative simply because it sounds plausible.

Every important claim should be checked against the underlying source.

This is particularly important because FDA guidance changes over time.

In 2026 alone, FDA has published or updated several CGT-related resources, including:

• Final CGT development FAQs

• CMC flexibilities guidance

• Genome-editing prior-knowledge draft guidance

• Genome-editing safety-testing draft guidance

• Other CGT development materials

[2][4]

The regulatory team should therefore establish a source hierarchy.

A sensible order is:

Current FDA regulation → Current FDA guidance → FDA meeting information → Official FDA communications → Peer-reviewed literature → Other secondary sources

AI should help navigate these sources.

It should not silently substitute one source for another.

Step 14: Track Regulatory Changes During Preparation

An INTERACT package may take weeks or months to prepare.

Regulatory information can change during that period.

An AI-enabled monitoring workflow can flag new:

• FDA guidance

• Draft guidance

• Final guidance

• CBER announcements

• Meeting procedures

• Relevant clinical-development recommendations

• Gene therapy safety updates

For example, FDA's current CGT guidance page lists several 2026 publications, demonstrating how quickly the regulatory environment is evolving. [4]

A change-monitoring process can prompt the team to review whether a new publication affects the briefing package.

Step 15: Build a Pre-Meeting Simulation

AI can also support internal rehearsal.

A sponsor can use its approved evidence base to simulate questions that an FDA reviewer might reasonably ask.

For example:

CMC

"What is the justification for the proposed characterisation strategy?"

Nonclinical

"What evidence supports the relevance of the selected model?"

Clinical

"Why is this patient population appropriate for first-in-human evaluation?"

Regulatory

"What is the specific decision for which FDA feedback is being requested?"

The objective is not to predict exactly what FDA will ask.

It is to expose weak assumptions before the meeting.

Step 16: Convert FDA Feedback Into Development Actions

The value of an INTERACT meeting comes from what the sponsor does with the feedback.

After the meeting, create a structured record:

FDA feedback → Interpretation → Development impact → Action → Owner → Deadline

For example:

FDA feedback

Additional information may be useful to support a proposed approach.

Internal interpretation

The CMC package requires additional evidence.

Action

Generate or analyse the relevant data.

Owner

CMC team.

Impact

Update development plan and future regulatory interaction.

AI can help maintain this action map and connect each action back to the original question.

Where AI Helps Most in FDA INTERACT Preparation

The strongest AI applications are generally those that reduce information-processing work while keeping regulatory judgment with experts.

High-value applications

• Regulatory literature search

• FDA guidance comparison

• Evidence extraction

• Question drafting

• Evidence-to-question mapping

• Document consistency review

• Reference checking

• Gap identification

• Meeting simulation

• Action tracking

• Regulatory change monitoring

Activities requiring strong human control

• Final regulatory strategy

• Scientific interpretation

• CMC decisions

• Toxicology strategy

• Clinical development decisions

• Risk acceptance

• Final FDA questions

• Interpretation of FDA feedback

The distinction is important.

AI should increase regulatory team's capacity.

It should not become the regulatory decision-maker.

Pienomial and Pharma Clinical Strategy

For complex regulatory preparation, the challenge is often not finding one document.

It is connecting many pieces of evidence.

Pienomial's Life Sciences solution is designed to support connected intelligence workflows across clinical, scientific, competitive, and regulatory research.

Its platform provides an enterprise knowledge layer designed to connect information and context across AI workflows.

For regulatory teams, this type of architecture can support a workflow such as:

FDA guidance → Programme evidence → Regulatory question → Expert review → Meeting package → FDA feedback → Development action

Pienomial's Knol AI can support AI-assisted research and synthesis, while KnolComposer can help turn validated research into structured documents and briefing materials.

The important principle is that the AI output should remain connected to its underlying evidence.

A Practical FDA INTERACT Preparation Checklist

Before submitting an INTERACT meeting request and package, sponsors should confirm:

• The investigational product has been clearly identified.

• Preliminary preclinical proof-of-concept work has been completed.

• The programme is at an appropriate stage for INTERACT.

• The development team has confirmed that INTERACT is preferable to another meeting type.

• The disease and target population are clearly described.

• Natural history information has been reviewed.

• Current treatment options have been documented.

• The first-in-human study has been outlined at an appropriate level.

• The CMC section provides the required high-level product and manufacturing information.

• Relevant preclinical evidence has been organised.

• Each question has a clear rationale.

• Each question is supported by appropriate evidence.

• Questions are prioritised based on development impact.

• The package remains concise.

• References have been verified against primary sources.

• Terminology is consistent across sections.

• AI-generated content has been reviewed by subject-matter experts.

• Regulatory claims have been checked against current FDA sources.

• A process exists to track FDA feedback after the meeting.

Conclusion

Preparing for an FDA INTERACT meeting is fundamentally an exercise in early regulatory decision-making under uncertainty.

For a gene therapy programme, the sponsor must bring together product information, CMC considerations, preliminary preclinical evidence, clinical context, natural history information, and a focused set of regulatory questions.

FDA's current guidance makes clear that the INTERACT package should be targeted and appropriately concise. For a 60-minute meeting, FDA considers up to 10 questions, including sub-questions, reasonable, while briefing packages should not exceed 50 pages. [1]

AI can make the preparation process more efficient by helping teams search guidance, organise evidence, identify gaps, compare precedents, draft question candidates, perform consistency checks, and maintain traceability from evidence to regulatory questions.

But the most important part of the workflow remains human.

Regulatory experts must determine what matters.

Scientists must assess whether the evidence supports the proposed approach.

CMC specialists must evaluate manufacturing and product-quality considerations.

Clinical experts must assess the development strategy.

And the sponsor's regulatory team must decide what it actually needs from FDA.

The most effective model is therefore not AI replacing regulatory strategy.

It is:

AI finds and connects evidence → experts interpret it → regulatory teams formulate questions → FDA provides feedback → the programme adapts.

For gene therapy developers, that can turn an early FDA interaction from a general discussion into a focused decision-support opportunity.

And as FDA's gene therapy regulatory guidance continues to evolve, an evidence-connected approach can help sponsors remain prepared for the next regulatory question as well.