What GRIPPify does

GRIPPify turns a large set of interview transcripts and company documents into a GRIPP process model. Every source is read in full; the five GRIPP layers (Processes, Actors, Systems, Information, Products & Services) are extracted, deduplicated across all sources and cross-validated. Each process is then modelled step by step (DPD), linked to the other processes, checked for red flags and gaps, and exported in the formats the GRIPP skills prescribe.

Every element carries one of three validation values. It is Confirmed when an interview and a document agree, or when two or more interviews do. Otherwise it is Interview only or Document only — unconfirmed. Nothing is assumed: missing information is marked [GAP] and routed to the gap register.

1. Password & Session

Enter the shared GRIPPify password (ask the team for it); your browser remembers it until it changes. Each engagement lives in its own session, which you manage from the session strip in the header. That's where you Use an existing ID, Copy it, create a New ID (optionally password-protected), or Save and Open it under a project name. Share the session ID with colleagues to work on the same engagement.

2. Upload

• Engagement: enter the client company name.
• Interviews: interview transcripts or notes, one file per interview. Every element they mention gets Source type Interview.
• Company documents: SOPs, process descriptions, org charts, policies, system overviews, BPMN/XML exports. Source type is Document.
• Sector framework (optional): a reference model such as eTOM, APQC or a sector process house. It is used for domain naming and the framework-alignment check. It is never counted as a source.

Drop any number of files, or a .zip of a whole folder. Uploads go in batches and start immediately. Supported formats are PDF, DOCX, PPTX, XLSX, CSV, TXT, MD and XML/BPMN; scanned PDFs go through OCR. The category decides the Source type, so put each file in the right box. Delete a misplaced file with ✕ and upload it again in the right category.

Voice interviews (ElevenLabs): instead of uploading a transcript, you can let the AI interviewer do the interview. Enter the interviewee's name, role and department and click + Create link; the personal link is copied to your clipboard. If you also fill in their email, the button becomes + Create link & email and a standard invitation (purpose, 20–25 minutes, what they need, recording notice and the personal link) opens in your own mail client, ready to send from your own address. Set Sign invitation emails as once; it is remembered in your browser. The link then shows ✉ Invited, and ✉ on any link opens the invitation again, e.g. as a reminder. GRIPPify only opens the email for you; it is sent when you click Send in Outlook. The interviewee opens it, agrees to recording and talks to the interviewer in the browser (microphone needed, no account). During the call they see a live transcript and can click ✎ Fix on any of their own lines to correct a mis-heard name, system or term; the interviewer is told straight away. When the call ends the transcript, with a short summary, appears under Interviews in this session as ElevenLabs_<name>_<id>.txt. The link then shows ✓ Done; click ↻ Refresh to update. A link only ever delivers into the session it was created in. Corrected lines are marked (corrected by interviewee; transcribed as: "…") so you can see what was changed. Withdraw a link with ✕; transcripts already received are kept.

3. Run

Press Run GRIPP analysis. The pipeline runs these stages:

1. Ingest: reads all files and assigns source IDs (INT-###, DOC-###).
2. Extract: the five GRIPP layers from every source, run in parallel.
3. Consolidate: merges duplicates and aliases across all sources, assigns IDs (PRC/ACT/SYS/INF/PRD), computes validation and confidence, and builds the domain map.
4. Model processes: the Process Builder writes the step-by-step DPD, the HLPD content, conflicts and red flags for each process.
5. Analyze: cross-process links, coverage matrices, readiness thresholds and the gap register (A–D).
6. Generate: Excel, HLPD and gap report documents, charts, and the anonymized Sia copy.
7. Validate: structural, naming, coverage, data-quality, completeness, anonymization and consistency checks. Leaks found in the Sia copy are fixed automatically.

Large engagements (hundreds of sources) take a while. You can leave the page; reopening the session reconnects to a running job. Re-running after adding sources reuses earlier results for unchanged files and processes, and process IDs never get renumbered.

4. Results

The menu on the left of the Results tab jumps straight to each section and highlights where you are as you scroll. BPMN for Adonis opens the list of BPMN files in the Client deliverables.

• Readiness: the READY / READY WITH CONDITIONS / NOT READY decision, each threshold, the validator result and the executive summary.
• Process charts: an interactive viewer. The Domain overview shows each domain card with its processes. Each process chart is a swimlane per responsible actor, with BPMN shapes and validation colours; document-only steps have a dashed border, red-flag badges sit on their steps, and loops show as dashed arrows. Click a step to see its full DPD row. Toggle between the Client and the anonymized Sia copy.
• Red flags, Gap register and Domains: sortable summaries of the deliverables.
• Downloads: the complete Client and Sia packages as zip files, or individual files. The HLPD documents and the BPMN files for Adonis are listed in their own expandable sections.

Validation & confidence

Value	Meaning
Confirmed	Interview + document, or ≥ 2 independent interviews
Interview only	One interview, no document support
Document only — unconfirmed	Only documented, not confirmed in practice. These steps get a "Confirm via interview…" open question and a Category B gap.

Confidence follows the number of sources: 1 = low, 2 = medium, 3+ = high. A DPD step takes the least-confirmed status among its sources and the elements it references.

Readiness thresholds

• Domain coverage ≥ 80%: share of domains with at least one interview.
• Layer coverage ≥ 60%: share of processes that reference actors, systems, information and products.
• Cross-validation ≥ 50%: share of processes that are Confirmed.
• Critical conflicts = 0.
• Red flags ready for design ≥ 70%.
• Orphan elements < 5%.

The result is NOT READY if domain coverage, layer coverage or critical conflicts fail. It is READY only when every threshold is met; anything in between is READY WITH CONDITIONS.

Deliverables

• …_process_model.xlsx: the Adonis import workbook. It has MV, BPO, a DPF and a DPD tab per process, Red_Flags, the gap register, coverage, and the element registers.
• HLPD/HLPD_PRC-###_….docx: one High-Level Process Description per process.
• BPMN/PRC-###_….bpmn: one BPMN 2.0 XML file per process, for import into Adonis (BPMN 2.0 import). Each file has one pool with a swimlane per responsible actor, the events, tasks, gateways and sequence flows, and the diagram layout, so Adonis draws it as in the process charts. The DPD details of each step (RACI, systems, data, validation, open questions) are in the step's documentation field. Several files can be imported in one go.
• …_gap_report.docx: the Final Gap Report, with the readiness decision, gap register A–D and deep-dive plans.
• …_process_charts.html: the interactive charts. The file is self-contained, so you can open it offline or share it.
• validation_report.md: the Validator result (Client package only).
• mapping_table.json: maps every anonymized ID back to the client value. It is in the Client package only; never share it with the Sia copy.

The Sia copy replaces persons, the company, departments, partners, systems and document names with IDs (Actor-001, Company-A, ERP-001, Doc-001, …), generalizes amounts, years and contact details, and drops the evidence quotes. Process IDs and validation values stay identical, so both copies can be compared.

Tips

• Transcripts with speaker names and roles give the best actor and swimlane results.
• Upload documents together with the interviews that cover the same area. This is what turns Interview only into Confirmed.
• Run early with the first batch of interviews, then use the gap register to plan the next interviews.
• Delete the session data when the engagement is finished (the Delete session data button appears in the banner once a session is 7 days old).