Nextpoint Intelligence
The clearer and more specific your search, the better your results.
Intelligent Search reads plain language, so describe the documents you need the way you would brief a colleague.
Intelligent Search combines two kinds of search in one search bar. Keyword search finds exact terms, names, and Boolean strings. Intelligent Search finds documents by meaning, even when they never use your words. Switching between them takes one click.
This guide covers how to structure a search, patterns that work well, how to narrow results, and how to check what comes back.
When to use Intelligent Search
Intelligent Search works best during review on a refined set of documents, not on a raw data dump. Cull first with deduplication, filters, and Boolean searches, then use Intelligent Search on what's left.
Think of it as choosing the right filing cabinet, not the whole file room.
How to structure a search
Strong searches combine two or three of these parts: what kind of document, who is involved, what it's about, and when.
| Part | What to include | Example |
|---|---|---|
| What kind of document | Communications, correspondence, contracts, financial statements, documents authored by | "Correspondence…" |
| Who | People by name, or parties by role | "…between the project manager and the general contractor…" |
| About what | The substance, described the way you'd explain it to a colleague | "…about delays to the foundation pour…" |
| When | A date range in plain language | "…between March and June 2024" |
Start with one or two parts. Add more only if your results are too broad. A search that combines every possible condition at once is harder to troubleshoot than one you build up step by step.
Recommended search patterns
These patterns work well for common review tasks. Replace the bracketed parts with details from your matter.
| When you want to… | Try this pattern | Example |
|---|---|---|
| Find what two people discussed | Communications between [person] and [person] about [topic] | Communications between the CFO and outside auditors about revenue recognition |
| Prepare for a deposition | [Person]'s correspondence about [topic] | Jane Doe's correspondence about the vendor contract renewal |
| Find documents by author or custodian | Documents authored by [person] about [topic] | Documents authored by the plant manager about safety inspections |
| Focus on a time period | Communications about [topic] between [month/year] and [month/year] | Communications about the product recall between August and October 2023 |
| Find a concept, not a keyword | Describe the conduct or situation | Pressure on employees to approve expenses without review |
| Find a specific kind of document | [Document types] for [entity] for [period] | Balance sheets and income statements for Acme Corp for fiscal years 2022 to 2024 |
| Identify sensitive content | Documents containing [category of information] |
Documents containing personal contact information of non-parties Note: Auto-redactions in NextGen are best-suited for redacting PII across document pages. |
| Map outside parties | [Role] involved in [subject] | Subcontractors involved in the east wing renovation |
When you need an exact phrase, a Bates number, a defined term, or a negotiated search term list, switch to keyword search.
Narrowing your results
If your results are too broad, refine the search you have rather than starting over.
- Build on your search in the chat panel. Copy your search into the chat panel beside your results and add a condition, such as "only from 2023" or "only final versions."
- Name the exact document types. A search for "financial statements" can return anything related to money. "Balance sheets and income statements for [company] for [years]" is much more precise.
- Find what you don't need, then set it aside. Search for documents to exclude, such as newsletters, marketing emails, or invoices. Tag or folder them in bulk so you can leave them out of your review.
- Use keyword search for precise filtering. Keyword search can filter on any document field. Use it when you need exact terms or need to work within specific folders.
- Combine both methods. For important review work, run a keyword search and an Intelligent Search on the same question and compare the results.
When keyword search is still the right tool
Keyword search is deterministic: the same search returns the same results every time. Intelligent Search can make mistakes, so match the tool to the task.
- Weed out with keyword search. Use it to remove what you know you don't need before you search by meaning.
- Find what keywords miss with Intelligent Search. Use it to surface hot documents about a deponent, a topic, or an area of the business.
- Rely on keyword search for privilege review, exhaustive responsiveness review, and redaction. These tasks need deterministic results.
- Use Intelligent Search as a quality-control check. It can help answer "did I miss anything?" but shouldn't be relied on alone. Trust, but verify.
Checking your results
Intelligent Search shows you why each result appeared, so you can confirm it before relying on it.
- Why this matched. Click "Why this matched" on any result to see a short explanation of how it relates to your search. The explanation is AI-generated and meant as a review aid. Always confirm it against the document itself.
- Match composition. The bar above your results shows how many came from keyword matches, AI matches, or both.
- Document viewer. When you open a result, the explanation appears in the side panel, and exact keyword matches are highlighted in the document.
- Saved searches. Save searches that work well so you can run them again. You can keep a saved search private or share it with everyone on the case.
Good to know
- Intelligent Search returns your top 1,000 matches by meaning. The most relevant documents appear first, and relevance decreases further down the list. A more specific search makes better use of those 1,000 results.
- Intelligent Search covers everything indexed in your database. It can't yet be limited to a single folder or run within a previous set of results. For folder-specific work, use keyword search. Intelligent Search is available to Standard, Advanced, Power and Basic users, who have access to every folder. It isn't available yet to Reviewer or View Only users.
- Intelligent Search reads document text. It covers emails, PDFs (including scanned PDFs with OCR text), Word, Excel, and PowerPoint files, and attachments. Images without readable text and audio or video without transcripts are not included.
- It doesn't learn from your activity. Your searches and review decisions are not used to train the model or build a profile of your behavior, and your case data stays within Nextpoint's private cloud environment.
- You remain the reviewer. Intelligent Search speeds up finding documents, but it doesn't replace attorney review. Confirm results before using them in productions, privilege decisions, or filings.
Help and feedback
If a search isn't returning what you expect, contact Nextpoint Support. Include the search you ran and what you were looking for, and we'll help you refine it.
Your feedback shapes what we build next. Tell us which searches worked well, which didn't, and what would make Intelligent Search more useful in your review.
Getting Better Results from Nextpoint's Intelligent Search
Overview
Intelligent Search lets you find documents by describing what you need in plain English. It combines keyword matching and meaning-based (semantic) matching in one ranked result set, inside Nextpoint.
Most AI search tools only match on meaning. Intelligent Search also keeps the precision of keyword search, so exact names, terms, and phrases still count. Every top semantic match can explain why it appeared, so you can verify it before you rely on it.
- Find what matters. Describe your search in plain English instead of building a Boolean string.
- Save time. Hours of query-crafting condensed into seconds.
- Catch what keywords miss. Code names, nicknames, synonyms, and phrasing a search term would never uncover.
- Keep it simple. Any team member can search in plain language, regardless of search expertise or case context.
- Defensible by design. Top results explain why they matched, so attorneys can verify before they rely on them.
- Built for legal workflows. AI that speeds up your review instead of distracting from it.
- Securely contained. No exporting, copying and pasting, or switching between AI tools. Your data never leaves Nextpoint, and no models train on it.
How it finds documents
Every Intelligent Search runs your query two ways at once, then merges the results into one list.
- Keyword matching finds documents that contain the words, names, and terms in your search. This is the precision you know from keyword search.
- Semantic matching finds documents by meaning, even when they never use your words. It recognizes synonyms, code names, nicknames, roles, and related phrasing. A search for "pressure on employees to approve expenses" can surface emails that never use the word "pressure."
- Hybrid matches are documents found both ways.
Intelligent Search reads extracted document text, including OCR text, plus document metadata such as date, author, custodian, and document type. Results are only as good as the text layer: a poor scan or a document without OCR limits what any search can find.
Each search runs within a single database. Cross-database search is not supported at this time.
How results are ranked
Results are ranked by a single hybrid relevance score that blends the keyword match and the semantic match for each document. The most relevant documents appear first.
- Relevance score. Each result shows a score from 1 to 5 bars in the Relevance column. More bars mean a stronger match to your search.
- Up to 1,000 results. Intelligent Search returns your top 1,000 matches across keyword and semantic combined. Accuracy drops significantly past that point, so the list stops there.
- Relevance decreases down the list. A more specific search makes better use of those 1,000 results.
If you need every document that contains a term, with no cap, use keyword search.
Why this matched
The top 50 semantic matches in each result set can include a plain-language explanation of why the document appeared. Each explanation is specific to that document and points to the passage that supports the match.
- An explanation appears only when the passage backs it up. Intelligent Search checks each top semantic match for a passage that clearly or partly supports your search. If it finds one, you see a reason. If a document is close in meaning but no passage supports the match, it gets no reason rather than a made-up one.
- So you may see fewer than 50 explanations. For example, a search with 55 semantic matches might show 15 explanations. The other 50 documents are still ranked results; they just don't have a supporting passage to cite.
- Keyword matches don't need an explanation. Your exact search terms are highlighted when you open the document.
- Sort to work explained matches first. The Why this matched column is sortable, so you can bring explained documents to the top.
Explanations are AI-generated and meant as a review aid. Always confirm them against the document itself.
The search experience
-
Search bar. Run a Boolean keyword search when you know exactly what you're looking for, or search in plain English when you don't.
-
Intelligent Search toggle. Switch between keyword search and Intelligent Search in one click.
-
Chat rail. Turning on Intelligent Search opens a chat rail on the right. Start a search, then continue to refine it in the search bar or with follow-up searches. You can narrow, broaden, or change direction without losing what you already searched for. The rail is resizable.
- Composition bar. Above your results, a bar shows what share of matches came from keyword, semantic, or both.
-
Results in grid view. Click View Results to see matches in the standard grid, with two added columns: Relevance (1 to 5 bars) and Why this matched. Semantic matches are marked with the sparkling wand icon to the left of the document title. Keyword matches have no icon.
-
Document viewer. Click into a result to see the match in context. The Why this matched explanation appears in the search panel on the left for intelligent search searches, and keyword hits are highlighted in the document.
-
Saved searches. Save a search from the chat rail to run it again later. Saved searches are private to you and specific to each database.
You can code, tag, or folder documents from an Intelligent Search result set, just as you would in any grid view. Use Bulk Actions to move through a review set quickly.
Turning on Intelligent Search for a database
Intelligent Search is enabled one database at a time, through a short consultation with your Nextpoint team. Before a database can be searchable, its document text has to be indexed.
- Tell your Client Success team you're interested. They'll set up a short consultation call.
- Confirm the fit. Together you'll review the matter and the data set. Intelligent Search works best on a culled review set, so if the data needs more culling first, Nextpoint Services can help.
- Size the database. Nextpoint estimates how much Intelligence Capacity the database will use. Capacity tracks text volume, so long, text-heavy documents use more than short emails. Included Intelligence Capacity varies by client.
- Indexing runs. Once approved, indexing usually starts within a few hours. It takes anywhere from a few hours to more than a day, depending on text volume.
- Start searching. Your Client Success team confirms when Intelligent Search is live on the database.
Culling first with deduplication, filters, and search terms keeps results focused and capacity use predictable.
Security and Data Handling
- Intelligent Search runs entirely inside the Nextpoint platform, with the same security controls as the rest of Nextpoint.
- Your data stays in Nextpoint. No exporting documents to outside AI tools, and no copying and pasting between systems.
- No model training. Your documents, searches, and review decisions are not used to train, fine-tune, or evaluate the underlying models.
- Isolated to your account. Searches are isolated to your account and are not retained beyond Nextpoint's standard operational window.
- Contained environment. Intelligent Search runs in Nextpoint's private cloud environment on AWS.
Intelligent Search is an internal review tool. It is not an ESI-protocol collection methodology. Keyword search and Search Hit Reports remain the deterministic method of record.
Good to know
- Who can use it. Intelligent Search is available in NextGen Discovery and Litigation databases to Standard, Advanced, Power, and Basic users. It isn't available yet to Reviewer or View Only users.
- What it reads. Emails, PDFs (including scanned PDFs with OCR text), Word, Excel, and PowerPoint files, and attachments. Images without readable text and audio or video without transcripts are not included.
- What it doesn't do. It does not generate answers or content, code documents, predict coding decisions, or certify that nothing was missed. It finds relevant documents faster; you remain the reviewer.
- Coming soon. Filters by match type (keyword, semantic, or both).
Help and feedback
In the chat rail, you can note whether a response was helpful or not by clicking Yes or No. If a search isn't returning what you expect, contact Nextpoint Support. Include the search you ran and what you were looking for. For tips on writing searches, see Getting Better Results from Intelligent Search.