AI Search Public Beta 1.1
Release Date: Friday, June 6, 2025
We’re excited to announce the release of Advanced Employee Search (AES) — a major step forward in how Synthesis answers questions about people at your firm. This update will be automatically enabled for all current AI Search beta users at the close of business on Friday, June 6.
What’s Included in This Release
✅ Better Answers for Employee Queries
AES delivers significantly improved results for:
Skills and expertise searches (e.g., “Who knows InDesign?”)
Licenses and certifications (e.g., “Who’s a licensed architect in Texas?”)
“Go-to” person questions (e.g., “Who’s our BIM lead?”)
Role and title lookups (e.g., “Who are the project managers in SF?”)
Employee searches by office or location
✅ Expert Suggestions in Knowledge Searches
When you run a general knowledge query (e.g., “Revit standards”), Synthesis may now suggest a relevant person to contact — if there’s strong signal in the content.
✅ Improved Handling of Former Employees
Synthesis is smarter about when to include former employees:
It avoids recommending them for queries focused on current expertise
It includes them (with a clear label) for historical questions like “Who led Project X?”
It still responds accurately if you search for a former employee by name
✅ Faster, Smarter Answers for All Queries
This release also includes an upgrade to OpenAI’s new 4.1-mini model, which powers faster, more accurate responses across all AI Search queries — including employee and knowledge searches.
How Synthesis Finds Employees
When Synthesis AI Search answers questions about people (for example, “Who has experience with performance venues?” or “Who are our market sector leaders for healthcare?”), it prioritizes employee information in the following order:
-
Structured employee directory fields (most reliable)
Structured fields like role, title, office, certifications, and centrally managed skills are the most reliable source for employee queries. -
Employee blocks on intranet pages
Employee blocks placed on pages—especially when they are clearly titled and consistently used—are the next most trusted source. -
Unstructured references (least reliable)
Mentions in free-form text such as Personal and Professional Bio fields, “I’m a resource for” fields, pages, documents, videos, and posts are considered, but weighed lower than structured directory data and employee blocks.
This tiered approach helps Synthesis return answers based on the most authoritative and repeatable data available.
Structured vs. Unstructured Expertise: Understanding the Tradeoff
Not all expertise should be treated the same way.
- Structured fields (dropdowns, controlled lists, centrally managed fields) are best for expertise you want to find reliably and consistently, every time.
- Unstructured fields (free-form text entered by employees) are useful for capturing the long tail of knowledge—unique experiences, niche tools, or one-off expertise that would be difficult to manage centrally.
However, there is an important tradeoff:
The more unstructured and inconsistent the data is, the less complete and repeatable employee search results will be.
Synthesis can often surface some relevant people from unstructured data, but it cannot guarantee that it will find all of them when the same expertise is described differently or stored in different places.
A Helpful Rule of Thumb
The more reliable and repeatable you want an employee query to be, the more structured and consistent the data should be.
Synthesis AI Search works best when your data strategy reflects how important it is to find every relevant person versus some relevant people.
Once you understand the tradeoffs between structured and unstructured employee data, the following tips will help you configure your directory and pages for the outcomes you want.
Pro Tips
Pro Tip: Pick a Lane - Consistency Matters More Than Location
For any given type of employee query, the most important rule is consistency.
If you want reliable results, make sure that answers to a specific kind of question live in the same place and are entered in the same way.
For example:
- If you want to answer “Who are our market sector leaders?”, choose one primary source of truth (such as employee blocks with a consistent title) and use it across all market sectors.
- Avoid splitting the same information across multiple locations (for example, some people listed in employee directory fields, others in employee blocks, and others mentioned only in text).
When the same type of expertise is scattered across different sources, Synthesis may return partial results—even though all of the information exists somewhere.
Pro Tip: Guidance for Common Scenarios
Use structured fields when:
- The expertise is strategically important.
- You expect users to filter using these fields in the Employee Directory or search for them in AI Search on a regular basis.
- You want repeatable, predictable results.
Use employee blocks when:
- You want to associate people with topics, markets, or services which evolve rapidly and would therefore be too cumbersome to manage as structured fields.
- You want to represent the employee's relationship to a topic, market, or service in a way that is unique to each person and would therefore be too cumbersome to manage as structured fields. For example, "Healthcare Market Sector Leader".
- You want visibility and context alongside related content and thereby associating employees with individual processes or specific expertise in AI Search.
Use free-form text fields when:
- The expertise is niche or uncommon.
- Central management would be impractical.
- Users can find what they need with partial rather than exhaustive search results.
Pro Tip: Choosing the Right Field Types for Your Employee Directory and AI Search
When setting up your employee directory fields to optimize results in AI Search, follow these guidelines to determine the most effective field type:
- Preferred Field Types: AI Search works best with "Short Text," "Dropdown – Select Single Value," and "Dropdown – Select Multiple Values" fields. These are ideal for accurate filtering and querying.
- Use "Dropdown – Select Single Value" for Brief, Defined Values: Select "Dropdown – Select Single Value" fields for simple, single-value entries such as "Office" or "Practice Group."
- Use "Dropdown – Select Multiple Values" for Multiple Entries: For fields where employees might have multiple values (e.g., languages spoken), use "Dropdown – Select Multiple Values" rather than text-based fields. This ensures AI Search accurately filters data.
- Handling Less-Reliable Data: If you have similar fields, one authoritative and the other less precise, consider setting the less precise field as a "Long Text" field. This reduces the likelihood of AI Search using less structured, user-generated data for filtering. For instance, employee-generated lists of additional certificates or skills should be Long Text fields, while HR-controlled fields should remain structured for reliable AI filtering.
Pro Tip: How Employee Block Titles Impact AI Search
Descriptive block titles help AI Search identify and return the right people in Advanced Employee Search. When used well, titles provide essential context that improves search accuracy — especially when you're showcasing specific teams, committees, roles, or areas of expertise.
That said, a clear and specific page title can also provide strong context. For example, a page titled “Revit Container Files” with a generic employee block labeled “Contact” works well — the page title helps clarify the employee’s role.
However, in other cases — like listing multiple committees or workgroups on a single page — a more descriptive block title (e.g., “Design Committee Members”) is important to avoid confusion and help AI Search distinguish between groups.
Best Practice:
Use clear, concise block titles when a page includes multiple groups or when roles could be ambiguous.
On pages with a single, clear purpose, a strong page title may be sufficient, and a generic block title like “Contact” can work well.
NOTE: Hidden block titles are not indexed by AI Search. If you hide a title, you're removing a valuable signal from search.
Help Us Keep Improving
Please encourage your users to continue rating their search results — it’s one of the most helpful ways to improve the quality of Advanced Employee Search over time.
Additional Questions
Please email us at to support@knowledge-architecture.com.