Answering DDQs / RFPs in Norm Ai
How Answers Are Sourced
All responses are generated exclusively from firm-approved sources to ensure accuracy, consistency, and compliance. Norm Ai will never scan the internet for potential responses. Answer generation can be customized on a per-questionnaire, and per-question basis, giving each submitter control over how answers are prioritized and generated, while ensuring consistency across users.
Norm Ai is transparent about the source of every answer to ensure that you always know what you’re reviewing.
Answer Sources
Answer Bank
The Answer Bank is a repository of approved responses your team has already vetted. Each entry is tagged, sourced, and reusable across funds, teams, and questionnaires. Approved edits can be pushed back in, so the bank compounds with every DDQ.
All firmwide responses are stored in the Answer Bank and tagged as such.
Think about using the Answer Bank for stable, reusable content: business unit / strategy / sub-strategy / fund history, investment philosophy, governance, and policy- or compliance-driven responses where consistency is paramount. It supports images like org charts and variables like AUM, performance figures, and headcount, so foundational answers can stay current without manual rewrites across dozens or hundreds of responses. Lean on it when the same question appears across funds or BUs, and promote any foundational reviewer edits into the bank so the canonical answer keeps improving.
Supporting Materials
Use Supporting Materials to give Norm Ai the source documents it needs to draft or extract answers: Fund PPMs, prior DDQs, policies, compliance manuals, and other reference content.
Like the Answer Bank, Supporting Materials are tagged so the right documents surface for the right submissions. Norm Ai generates answers using the supporting documents selected for each submission, with citations back to source so reviewers can see exactly which documents and excerpts were used.
Think about leveraging Supporting Materials when content is too long, too detailed, or too frequently updated to live in the Answer Bank, like full policy documents or fund-specific offering materials.
Tip: Use Advanced Registry Search to search your team's repository of Supporting Materials in the Documents tab of the Review pane.
Answer Methodologies
Verbatim Answer Extraction
Norm Ai pulls an exact, unmodified response directly from a matched Answer Bank entry or Supporting Material document. Use this mode when the exact language needs to be preserved, like a standard response that has already been reviewed and approved or wording your team has settled on for recurring questions.
AI-Generated Answers
Norm Ai drafts a response by synthesizing content from the Answer Bank and Supporting Materials selected for the submission. The engine adapts tone, structure, and length to fit the question, and every answer includes citations back to the underlying source. Use this mode when the question requires composition across multiple sources or when no single approved answer fully covers it.
When to think about using Answer Bank / Supporting Materials and Verbatim / AI-Generated Answer Generation
| Answer Bank | Supporting Materials |
|---|---|---|
Verbatim | Strong match to an approved Firmwide or team-tagged Q&A pair. Norm Ai will pull the approved answer as written, with tagging. | Exact passage in a source document directly answers the question. Norm Ai will extract the language as written, with citation. |
AI Generated | Related entries exist in the Answer Bank but no exact match. Norm Ai agent synthesizes across approved Q&A pairs with responses marked for verification where required | No Answer Bank coverage, but relevant source documents are available. Norm Ai agent synthesizes a response from those documents with citations. |
Note: Verbatim matches from the Answer Bank or supporting materials are the default DDQ response. AI-generated answers are preserved as the last entry in the alternative answer selection carousel, so teams can preview and apply these answers when needed.
Answer Generation Strategies
Answer All
You can answer questions individually, but Answer All provides the ability to answer all questions at once. When you run Answer All, you choose which Supporting Materials to use and whether to include the Answer Bank. Norm Ai then attempts to answer every question, prioritizing what you indicate. Users will also be prompted to indicate whether they would like the agent to generate AI Generated responses.
Norm Ai then attempts to answer every question, pulling verbatim matches from the Answer Bank first, and generating cited answers from Supporting Materials where needed.
Tip: Answer All works best when used with a targeted set of supporting materials relevant to the specific submission, or for a submission where you know it will heavily leverage the Answer Bank. Running Answer All against your entire supporting materials library may reduce the quality of generated responses as the scope of what the agent is looking at is too broad.
Answer Logic Engine
The Answer Logic Engine lets users control (i.e. "prompt" the agent) which answers from the Answer Bank Norm should prioritize when generating responses for a DDQ submission. Rather than searching the entire answer bank every time, you can define priority levels to focus Norm on exactly the right content for each submission.
Levels are numbered in priority order. You can add as many as you need, drag to reorder them, and configure each one to either include or exclude content based on tags (like a fund name, strategy, or business unit).
- Use Includes to tell Norm which content to look at. If you add multiple tags, Norm will consider any answer that matches at least one of them.
- Example: A Bain user completing a DDQ configures three levels:
- Level 1 includes Private Equity Europe Fund VII (fund-specific)
- Level 2 includes Private Equity - Europe (sub-strategy-specific)
- Level 3 includes Private Equity (BU/Strategy-specific)
Norm searches all three levels and surfaces the most relevant answer. If responses across all four levels are deemed equally relevant, the Europe Fund VII answer will surface first because it sits at a higher priority level. In the Review pane, the Europe Fund VII answer will be shown first and the ensuing PE - Europe and PE responses will appear as alternatives.
- Use Excludes to remove content from consideration. Any answer that matches even one excluded tag will be filtered out.
- Example: A Bain user wants to consider only Sub-Strategy level responses and exclude any fund-specific ones.
The Answer Logic Engine will not manipulate the integrity of underlying answer. It only controls which Answer Bank answers are considered.
Answer Statuses
What if the response Norm Ai retrieved is not the exact perfect response I was thinking of? What if I want to manipulate the response to better match the contextual considerations of the question?
See ensuing section "Working With Answers"
Answer Statuses
- Answered - The agent found sufficient supporting material to answer the question, and/or used an answer located in the Answer Bank.
- Answer Bank - Answered verbatim from an approved Answer Bank entry
- Verify - The agent found information that could only partially answer the question, or the agent found conflicting information amongst the source materials; thus, this answer type requires user review.
Note: Answers that are AI-generated will be marked as verify for the user's review and confirmation.
- Unanswered - The agent did not find supporting information that could partially or fully answer the question.
- Approved - Marking an answer as Approved confirms that a response has been reviewed. This action moves the answer out of your Answered queue and into the Approved filter, so you can focus on questions that require further review.
Citations
Every answer includes citations that link directly to the source material used to generate the response. Click a citation to view the original source text, including highlighted excerpts, so you can quickly verify the answer's accuracy and relevance.
In addition to the source text, each citation may include:
- Notes – Reviewer annotations, guidance, or additional context related to the source.
- Expiration Date – The answer's "valid until" date, when one has been set.
- Object Catalog Tags – Metadata used to categorize the answer, such as DDQ Team or Business Unit.
- Last Used & Submission Count – Indicates how recently the answer was used and the number of submissions in which it has appeared.
- Associated Image – A link to any image associated with the answer, when applicable.
- Variables - Identification of which variables are included in the answer.
How Norm Ai improves over time
Every DDQ makes the next one faster and stronger:
- Promote approved edits into the Answer Bank so reviewed language becomes the new default.
- Expand coverage with each submission as new Q&A pairs and source documents enter the system.
- Sharpen matching as the Answer Bank grows, raising verbatim hit rates and reducing AI-generated drafts that need heavy review.
- Tighten reviewer focus as standardized answers cut repeat edits, leaving attention for the genuinely new.
- Compound time savings as repeat questions resolve faster across funds, teams, and reporting cycles.
Note: Norm Ai does not "learn" from individual user interactions. Editing an answer, answering a question, or approving a response will not change how the agent answers future questions. To influence future answer generation, update the content in your Answer Bank and Supporting Materials.