Tacit knowledge
does not walk out the door.
When a senior expert resigns, the routines they wrote down leave with the file. The judgment, the workarounds and the relationship map (everything that makes the role actually work) leave with them. Torch is the AI agent that conducts the exit interview: it interrogates departing experts across the long tail of context, until tacit knowledge is rendered explicit, validated and structured for handover.
No monthly commitment. Per-extraction pricing. GDPR-compatible.
One engine, three lenses on departing knowledge
Torch does not transcribe an exit. It interrogates the expert, renders the tacit logic into structured assets, and reads the cohort-level signal behind the departure.
Interrogate, do not transcribe
An interviewer with the persistence of a senior analyst. It drills into vague answers, flags imprecision in flight, and does not stop until the underlying logic is articulated and the Completeness Score clears the gate.
Tacit knowledge, rendered explicit
The validated transcript is compiled into a structured Standard Operating Procedure and reusable knowledge assets: PDF for stakeholders, JSON for downstream systems, Confluence-ready HTML for direct ingestion.
Read the signal behind the resignation
Across exit transcripts, Torch surfaces the cultural friction and structural drivers behind attrition, reframing anecdotal feedback into a prioritised retention agenda the senior team can act on.
Interrogate until the gate clears
Torch drills into vague answers in real time. A validation agent scores the transcript as the conversation runs, and nothing is packaged until the Completeness Score clears the threshold. What you get is a structured, validated knowledge asset, not a recording.
- Active inquiry. It probes vague answers until the tacit logic is articulated.
- Completeness gate. Below 0.70, the interview re-opens with targeted probes.
- Structured asset. Compiled to SOP, PDF, JSON and Confluence-ready HTML.
There's a VIP list... it's on my laptop.Static documentation is obsolete the moment it is written.
Asking an expert to "write it all down" runs straight into the Polanyi Paradox: we know more than we can tell. The expert documents the routine they can articulate; the edge cases, the workarounds and the judgments that actually keep production stable remain in their head, and depart with them.
The cost of replacing a senior performer when their knowledge departs with them — well-documented in the HR research literature.
Static Wikis
Wikis decay into a knowledge graveyard. No context on why a decision was made, no upkeep cadence, no engaged readership.
Screen Recorders
Recordings capture sequence, not rationale. Knowing where to click is operationally useless without knowing why.
Torch Active Inquiry
An interviewer with the persistence of a senior analyst: it does not transcribe, it interrogates — until vague answers become specific and tacit logic surfaces.
Surface the soft context that documentation cannot.
Torch modulates its line of questioning based on the role and seniority. An engineer leaves behind code; a sales director leaves behind a relationship map; an operations lead leaves behind exception-handling judgment.
Engineering and DevOps
Surfaces the rationale behind legacy systems, the undocumented workarounds and the load-bearing improvisations that hold the production infrastructure together.
Commercial and leadership roles
Maps real decision-makers within client accounts, the unwritten conventions and the political topology. Torch interrogates the relationship map, not just the deal pipeline.
Operations and back office
Codifies exception-handling judgment: how experienced staff manage errors, vendor friction and procedural failure when the manual no longer applies.
The same engine runs before the resignation.
The Active Inquiry engine is not constrained to offboarding. It is equally effective at documenting how top performers actually work while they remain in role.
Codify top-performer practice
How does the top sales rep actually handle the objections that derail others? How does the lead architect choose the boundary between two services? Torch extracts the tacit method, structures it, and makes it teachable.
Internal handovers and transitions
Moving a project from engineering to operations, or back from an external agency, requires the rationale — not just the artefacts. A Torch session captures the setup logic and operational invariants so the transition does not regress.
Strategy: Bypass Legal bottleneck via CTO.
Asset: Pre-cleared Security Annex (Loc: /Sales/Docs/v2).
Standard offboarding captures the surface, not the substrate.
Standard offboarding extracts explicit knowledge — files, credentials, contact lists. The operationally decisive 80% sits in tacit knowledge: the situational judgment, the political context and the unwritten rules under which the role functions in practice.
Explicit Knowledge
What an employee can record on a standard handover form. Cheap to collect, low strategic information density.
- File locations & Folder paths
- Login credentials
- Official organizational charts
Tacit Knowledge
The 'how' and the 'why' derived from years of operating context. Difficult to verbalise without prompting, infeasible to reconstruct after the fact.
- Context & Judgment "Why did we choose this architecture over the standard one?"
- Political Awareness "Who are the real decision makers in the client's office?"
- Exception Handling "What do we do when the documented process fails?"
The signal behind
the resignation.
Operational extraction is only one axis. Torch also applies psychological pattern matching across exit transcripts to surface the cultural friction points and structural drivers that the senior team cannot read directly.
Dynamic Context Adaptation
The line of questioning adapts in real time to the interviewee's role, seniority and criticality. No static template — a model-driven interview.
Cultural Diagnostics
Surfaces toxic micro-cultures and silent disengagement patterns at the cohort level — observable in exit data before they migrate into engagement metrics.
Retention Intelligence
Reframes anecdotal exit feedback into a prioritised retention agenda. Addresses the upstream drivers of attrition, not the symptoms reported at exit.
Strategic_Exit_Brief.pdf
Confidential AnalysisCore Drivers of Turnover
"Convergent signal across Logistics exits: senior performers depart due to authorisation latency, not compensation. Recommend a fast-track operating privilege for the cohort."
Three phases, one Completeness Score.
Three-phase pipeline with a quantitative gate. Extractions below threshold are returned to the interview phase before any asset is generated.
Active Interview
The departing expert interacts with Torch by voice or text. The agent maintains a dynamic context graph of the interview, flags imprecise answers in flight, and drills down recursively until the underlying logic is articulated.
Validation & Scoring
A back-end validation agent scores the transcript in real time. When the Completeness Score falls below 0.7, Torch re-enters the interview phase with targeted clarification probes before proceeding.
Asset Generation
The validated transcript is compiled into a structured Standard Operating Procedure — PDF for stakeholders, JSON for downstream systems, and Confluence-ready HTML for direct ingestion into the internal knowledge surface.
Integrates with your stack
Real-time notifications and user sync. Where your team works, with the identity you already use.
Slack
Microsoft Teams
Google Chat
SCIM 2.0
Microsoft Entra ID
Okta
Google Workspace
Pay per extraction.
Per-extraction pricing — no committed monthly spend, no per-seat tax. Credits do not expire.
Pilot extraction
Single-extraction engagement to validate output quality before committing volume.
Includes:
- 1 Full Torch Interview Session
- PDF, JSON & Confluence-ready HTML Deliverables
- Multi-topic Extraction Support
- 60 days Report hosting
Team engagement
For organizations running a structured extraction programme.
Everything in Pilot, plus:
- 5 Torch Sessions (Save €50)
- Up to 5 chatbots with your Knowledge Assets
- 6 Months Knowledge Hosting
- Enrich Knowledge Bases with your own documents
Operating at higher volume, or extending Torch into pre-departure retention work? Schedule a scoping conversation for an enterprise engagement.
Frequently asked questions
How much time does it take for the expert?
What are the addons included in the Team plan?
- Knowledge Enrichment: Upload your existing documents (PDFs, DOCX, PPTX, Manuals) to the Knowledge Base. Chatbots will combine the expert's wisdom with your official documentation.
- Multi-Agent Deployment: Configure up to 5 distinct Chatbots (e.g., one for "Logistics", one for "Sales"), assigning only relevant knowledge to each.
- Active Hosting (6 Months): The Team plan keeps your Knowledge Base and Chatbots active online for 6 months, allowing your team to query them daily without downloading anything.
What happens to the data? Is it secure?
Can I use it for things other than offboarding?
What output formats are available?
- PDF: Professional, structured reports ready for sharing and archiving.
- JSON: Machine-readable format for integration with your systems and tools.
- Confluence HTML: Pre-formatted HTML files ready for direct import into your Confluence workspace. Simply download and paste into Confluence's HTML macro for instant documentation.
How does Torch adapt exit interview questions to different roles and seniority levels?
- For Technical Roles (Developers, DevOps, Engineers): Focuses on system architecture, undocumented workarounds, technical debt, and infrastructure dependencies. Questions dive deep into code logic, deployment processes, and troubleshooting procedures.
- For Sales & Leadership Roles: Emphasizes relationship mapping, client decision-making processes, unwritten agreements, and strategic insights. Questions explore key contacts, negotiation tactics, and market intelligence.
- For Senior Executives: Addresses strategic vision, organizational culture, high-level decision-making processes, and critical business relationships. Questions focus on long-term implications and strategic knowledge transfer.
- For Mid-Level Managers: Balances operational knowledge with team dynamics, process improvements, and cross-functional collaboration patterns.
- For Junior Staff: Concentrates on day-to-day procedures, tool usage, and practical workflows while identifying knowledge gaps that need to be addressed.
What if the output isn't good enough?
The next resignation is already in motion.
Capture the tacit knowledge before the calendar runs out.
Run the extraction