Scope vs Excel, Boondmanager,
Furious, ChatGPT.

Boondmanager and Furious are staffing ERPs (mission management, consultant planning, invoicing). They operate downstream of the project cycle and don't cover upstream scoping — hence the ❌ or 📋 marks depending on the criterion. Several tier-1 European IT services firms run both tool families in parallel (Scope for scoping, Boond/Furious for staffing).

1. 01Is it just a GPT wrapper?

   Pipeline structured in four distinct stages: sourced extraction, interactive clarification, scoping with a mandatory PM wizard, deterministic estimation.

   The final estimation compute uses no LLM — it's a PM-configurable formula with documented multipliers (complexity, maturity, coupling, etc.). The LLM only feeds the semantic structuring of the perimeter, and every entity remains traceable back to the original brief through a [Brief: ref-id] badge visible in the UI and in the exported PDF.

2. 02What happens if the AI hallucinates?

   Five guardrail layers are shipped in production:

   • Mandatory PM wizards on scoping (RACI + planning + risks validated before any LLM call) and on estimation (assumptions validated before any compute).
   • Enforced clarify: ≥ 5/8 fill rate is mandatory and high-severity contradictions block the rest of the pipeline.
   • Per-item provenance badges: every line carries a visible marker — [Brief: ref-id] (from the brief), [Entered] (edited by PM), [Suggested by Scope] (AI-inferred), [Inferred] (extrapolated without a direct source).
   • Transparency counter visible on every UI page + PDF (X% Brief, Y% Entered, Z% Suggested, W% Inferred).
   • Audit log + post-validate phantom refs: every mutation tracked, orphan references detected and flagged.

   For technical details, see getscope.dev/trust. No estimation number is invented by an LLM — the compute is deterministic from the assumptions the PM entered.

3. 03GDPR / DORA / AI Act — where do you stand?

   Durable application persistence is EU-only: Supabase (Frankfurt), Gladia STT (Paris GDPR-native), GlitchTip (Frankfurt), Vercel Paris cdg1. LLM inferences go through OpenRouter (stateless US proxy, `data_collection=deny` + EU SCCs) to OpenAI and Anthropic (United States) — Enterprise EU-resident routing option (Bedrock Paris / Mistral FR) under roadmap validation. Postgres RLS on 60 tables. Append-only audit logs with 13-month CNIL retention + 10-year accounting retention. Full GDPR export + delete cascade are working.

   DORA Art. 28(3) + Implementing Regulation 2024/2956 compliance documented in the public Trust Center. For procurement deep-dives: nine downloadable artefacts on getscope.dev/trust (DPA template, SLO, Incident Response, Exit Plan, RTO/RPO, Security Questionnaire, Subprocessors, Architecture data-flow, PoC Contract).

4. 04How is Scope different from Furious / Boondmanager?

   Furious and Boondmanager are staffing ERPs: mission management, consultant planning, invoicing, time tracking. Scope operates upstream of the cycle: client brief → structured scoping + defensible estimation, delivered as a portable PDF/DOCX.

   No direct overlap — chain complementarity: Scope produces the scoping and the defensible quote, Boondmanager/Furious staffs and runs the mission once approved. Several tier-1 European IT services firms use both tool families in parallel (Scope for scoping + Boond/Furious for delivery + invoicing).

5. 05How is Scope different from Excel + Word + a PM brain dump?

   Excel + Word will remain the universal tool — and that's fine. Scope automates the most expensive part of the process: the brief → semantic structuring transformation (actors, needs, constraints, assumptions, risks) with end-to-end traceability in a portable PDF.

   On a practice that produces 60 projects per year, that's the equivalent of 240+ days of manual scoping that can be automated (conservative estimate). The PM keeps editorial control at every step — Scope is a copilot, not a replacement. Manual edits are preserved across later LLM regenerations (immutable [Entered] badge).

6. 06What about ChatGPT/Claude directly in the browser?

   A free ChatGPT/Claude session gives you generated text, but not:

   • Line-by-line sourcing back to the original brief (every entity traced to its source)
   • The mandatory PM wizard that forces active validation of assumptions
   • The decision journal carried into the PDF/DOCX
   • Deterministic compute on the estimation (no LLM inventing person-days)
   • The multi-user, GDPR/DORA-compliant audit log
   • Cross-piece consistency (multi-brief / large RFPs split into pieces with namespaced ref_id)

   Scope ships all of that traceability and governance infrastructure. ChatGPT/Claude still shines for free brainstorming — not for a client deliverable that has to defend itself in procurement.

7. 07How do I try it?

   Three entry doors depending on your maturity:

   • 20-min guided demo: book directly with the founder via getscope.dev/contact
   • Free 1-month pilot on 3-5 real projects with a bi-weekly (~30 min) live feedback session and product iteration during the month
   • Commercial PoC contract after the pilot if value is validated (terms co-built per volume)
   • Public demo page with a sample PDF you can download without signing up: getscope.dev/demo