Project led by Víctor Muñoz, now an Akaen partner.
A global architecture consultancy received CVs continuously and built a valuable candidate archive, but could make limited use of it. HR still relied on manual searches, individual CV reading and whoever remembered a similar profile seen months earlier.
Executive summary
About the customer
A multinational architecture consultancy with recurring recruitment and assessment of technical specialists. It needed to reuse candidate knowledge quickly when a vacancy appeared, rather than merely storing CVs.
The challenge: a valuable archive that was difficult to use
The challenge was turning CVs into a searchable talent pool. Each CV contained useful information about qualifications, languages, previous roles, technologies, years of experience, location, availability and work permits, but it was trapped in unstructured documents.
This created a clear operational burden for HR: manually reading many documents, struggling to rediscover previous candidates and taking longer to build a useful shortlist.
The solution: digitised CVs searchable from Teams
We designed a workflow that turns each CV added to SharePoint into structured, searchable data. An LLM processed the extracted text into JSON containing relevant recruitment metadata, feeding both the client database and a vector index for semantic search.
A Microsoft Teams assistant let HR ask in natural language which profiles met specific criteria and receive an initial selection with a summary and direct link to each original CV in SharePoint.
Architect with 9 years of experience at international practices. Specialising in BIM, coordination of large-scale projects and team leadership. Work permits ES · EU · UK.
I found 3 profiles that fit:
BIM Manager · 9 years · English C1 · 1 month’s notice
View CV in SharePointBIM architect · 6 years · English B2 · 2 weeks’ notice
View CV in SharePointBIM Coordinator · 7 years · English C1 · Immediate
View CV in SharePointAutomated processes (4 key flows)
1. Centralised CV intake in SharePoint
The Problem: CVs were stored, but not ready for advanced search or efficient reuse.
Transformation: Each CV added to the corporate folder became a processable input in the digitisation workflow.
2. Structured extraction with AI
The Problem: Key recruitment data was buried in different formats, requiring people to read each document individually.
Transformation: An LLM processed each CV’s text into structured JSON with fields such as qualifications, languages, technologies, years of experience, previous roles, location, availability and work permits.
3. Dual storage for operational use
The Problem: It was not enough to save the document; it needed to be filtered and retrieved intelligently.
Transformation: structured data was stored in the client's database and content was semantically indexed in Azure AI Search with metadata ready to apply filters.
4. Conversational Search from Teams
The Problem: the HR team needed to access knowledge without changing tools or learning complex queries.
Transformation: A Teams assistant let HR request profiles using specific combinations of filters in natural language. It returned relevant candidates, a summary of each and a SharePoint CV link for final review.
Operational comparison
- • Initial manual reading of multiple CVs.
- • Difficult reuse of candidate history.
- • Slower shortlisting, dependent on recruiters’ knowledge.
- • Useful information enclosed within the original PDF or Word.
- • Automatic extraction of key recruitment metadata.
- • Semantic search with filters in Teams.
- • Initial screening without manually reviewing every CV.
- • Immediate access to the original document only when the profile already fits.
The Impact
Although no final operational metrics were retained after deployment, the design addressed clear bottlenecks in recruitment: less initial manual reading, better access to previous candidates and faster shortlisting.
- Faster shortlisting: HR moved from searching documents to finding already-filtered profiles.
- Reusing the candidate archive: past candidates became visible again when a similar need arose.
- Less initial manual screening: part of the screening happened before opening each CV.
- Better internal experience: queries were made in Teams, without requiring the team to use another tool.
Operational results
- Richer search: qualifications, languages, technologies, mobility and work permits could be combined in a single query.
- Actionable response: the assistant returned candidates, summary and direct access to the CV for final human validation.
- More scalable process: the growing repository no longer increased the team’s manual workload proportionally.
Conclusion
“Digitising CVs means turning a document repository into a searchable, reusable talent pool.”
Do you want to apply this approach to HR documentation or recruitment? .