Finding documented project experience inside your company
A project manager is looking for someone who has led requirements workshops. Similar work appears under different names in profiles and project descriptions, so they end up asking several colleagues again. AI can help find specific passages showing what a person did. A useful result gives the manager a starting point for a conversation about that experience.
Syntalith
From a profile keyword to a record of the work
Suppose an approved internal project record says, “Led stakeholder sessions to gather and confirm requirements.” Another profile says, “Attended a course on running workshops.” Both passages concern the topic, but only the first describes doing the work the manager is asking about.
A useful search tool shows the first passage with the project it belongs to and a link to the source. The manager can see why it is worth speaking with that person about their involvement in a similar task. If the course entry also appears, it remains a record of training attendance. It does not become evidence of leading a client workshop.
After reading the result, the manager confirms the experience with the person and checks availability separately. No matching passage means only that the search did not find that record in the material reviewed. It does not mean an employee lacks the skill. The tool retrieves documented work without ranking people or assigning them to a project automatically.
Maintained profiles may already provide the answer
Start with profile updates and the terms used in the company. If someone led requirements discussions but their profile says only “analysis,” clarifying the entry may help the next manager. Shared terms and synonyms can improve search without training a separate model.
Existing systems can hold this information. Microsoft Project Operations describes recording skills, education, certificates, and project experience. Keeping those profiles current remains part of the team’s work. Before commissioning another application, check whether the current fields and search can surface the description you need.
A model becomes relevant when the information exists but uses different wording from the manager’s question. “Led stakeholder sessions” might not appear in an exact search for “requirements workshops.” Search that accounts for meaning can surface the passage while preserving what it actually says about the work.
When to adapt the search
First compare ordinary search, agreed synonyms, and an available model. If the manager can readily find the relevant passages, further adaptation may be unnecessary. Opening the source also matters. A list of names without descriptions leaves another round of questions for the manager to ask.
An adaptation trial becomes relevant when the model repeatedly confuses the meaning of descriptions despite a clearly defined task. It might treat attending training as equivalent to documented experience leading sessions. People familiar with the projects can identify passages that answer the question and explain what distinguishes them from superficially similar entries.
Google describes adapting selected embedding models to a task or domain using labeled examples. These models can support retrieval of texts related in meaning. In this project, the assessment concerns the relevance of retrieved passages.
A result can also suggest linking a description to the company’s name for that type of experience. The profile owner reviews the suggestion before it is saved. If a correction remains only in the manager’s one-off results, the next search may encounter the same unclear entry. Connect result review with a way to request profile additions or corrections.
Check results from other projects
The manager and someone familiar with the records review results from projects that were not used to adapt the model. They check whether a passage actually describes the work being sought and whether its link opens the right source. They record irrelevant suggestions and cases where a known, relevant description was missed.
The comparison with the current method should establish whether the manager reaches a useful description sooner. If they still have to open every profile from the beginning, a fluent summary changes little.
If the manager needs the material for preparing a workshop, see our article on Company Brain for professional services firms. It covers finding reusable guides and explanations of when to use them.
Search project experience with links to its sources
Syntalith proposes an application for finding documented project experience. The manager describes a task, and the tool retrieves passages from profiles and approved project descriptions, showing their sources. Reviewing results also includes requesting corrections so useful additions reach the records the company maintains.
Your team identifies material intended for this search, explains experience categories, and resolves unclear descriptions of the work. We assess the existing profiles and search tool. Where recurring wording errors remain, we compare model adaptation with those options and connect the application to the places where project information is updated.
Describe one task that led a manager to ask colleagues about previous experience even though a relevant record existed. Tell us where it was eventually found. That is enough to begin discussing the tool you need; engagement information is on Syntalith’s pricing page.
Match a model to the task you need it to perform
Describe where your current AI falls short. We will compare model customization options, data requirements and the cost of running the resulting system.
Private LLMs and fine-tuning