Engineers spend weeks (6 weeks in the cited example) in discovery sessions trying to figure out what stakeholders actually need vs what they say they want, especially in large orgs like DOD.
Structured interview framework with AI analysis that ingests stakeholder conversations, documents, and meeting transcripts, then surfaces contradictions, ambiguities, and the likely real problem underneath the stated one. Outputs a clear problem statement and scoped requirements doc.
subscription
The pain is real, expensive, and widely acknowledged. A 6-week discovery at consultant rates ($200-400/hr) means $50K-200K burned per engagement just figuring out the problem. Misunderstood requirements are the #1 cause of project failure (per Standish Group data). Every senior consultant has war stories about this. The pain is acute but episodic — it hits hard during discovery phases, then fades. Losing one point because most people have learned to live with it as 'just how it works.'
The niche is real but narrow. Target personas (technical consultants, solutions architects running discovery engagements) number perhaps 50K-200K globally. At $100-300/month, that's a $60M-720M addressable market. Realistically capturable in year 1-3: maybe $5-20M. This is a solid indie/small SaaS business but not a VC-scale market unless you expand into broader product management or enterprise BA tooling. The government/defense sub-segment adds procurement complexity.
Consultants already bill $200-400/hr and would happily pay $200-500/month for something that cuts discovery time by even 30%. The ROI math is trivially obvious: save 2 weeks of a $300/hr consultant = $24K saved vs $3K/year subscription. Government contractors have budget for tools. Deduction: enterprise/gov procurement cycles are slow (6-18 months), and individual consultants may resist paying out-of-pocket for tools their clients should fund. Proof-of-willingness is moderate — people pay for Jama/DOORS but those are mandated, not chosen.
Very buildable with current AI. Core loop: ingest transcripts/docs → LLM analysis for contradictions, ambiguities, unstated assumptions → structured output. Modern LLMs (Claude, GPT-4) are excellent at this kind of analytical reasoning. Transcript ingestion is solved (Whisper API, existing transcription services). The hard part is the structured interview framework and domain-specific prompt engineering for defense/enterprise contexts — but that's product design, not technical risk. A strong solo dev could have an MVP in 6-8 weeks. Losing points for: handling sensitive/classified content requires careful architecture, and getting output quality consistently high across diverse domains is non-trivial.
This is the strongest signal. Nobody is solving upstream requirements DISCOVERY with AI. The entire market focuses on downstream requirements MANAGEMENT (Jama, DOORS, Visure). The current solution is literally 'hire expensive humans and hope they ask the right questions.' The DIY ChatGPT approach validates demand but has no productized competitor. The gap between 'managing well-formed requirements' and 'figuring out what the requirements should be' is enormous and unaddressed. This is genuinely greenfield.
Mixed. Consultants run multiple discovery engagements per year, so there IS recurring need. But usage is bursty — heavy during active engagements, dormant between them. Risk of high churn during slow periods. Annual contracts with government/enterprise clients help smooth this. Could improve with: (1) a requirements library that grows over time (switching cost), (2) team collaboration features, (3) ongoing requirements monitoring/evolution features. Without deliberate retention mechanics, natural churn could be 5-8% monthly.
- +Massive unserved gap: nobody productizes the upstream discovery process — this is genuinely greenfield
- +Trivially compelling ROI story: saving even 1 week of consultant time at $300/hr pays for years of subscription
- +AI is exceptionally well-suited to the core task (contradiction detection, ambiguity surfacing, synthesis across documents)
- +Target users (senior consultants, solutions architects) are high-intent buyers who make their own tool decisions
- +Defense/government market provides high-value anchor clients with large contracts and long retention
- !Government/defense procurement cycles are brutally slow (6-18 months) — you could run out of runway waiting for your first enterprise deal
- !The 'secret sauce' is largely prompt engineering + UX, which is replicable — Jama or IBM could bolt on similar AI features in 6-12 months if the category proves out
- !Handling sensitive/classified content (CUI, ITAR) requires FedRAMP or equivalent compliance, which is expensive and slow for a solo founder
- !Bursty usage pattern creates churn risk — users may cancel between engagements and re-subscribe, making revenue unpredictable
- !Narrow initial persona could cap growth unless you deliberately expand into adjacent use cases (product management, business analysis, sales engineering)
Enterprise requirements management platform with traceability, review workflows, and risk analysis. Used heavily in defense, automotive, and medical device industries. Supports structured requirements authoring and collaboration.
Legacy heavyweight requirements management system dominant in defense and aerospace. Manages large-scale requirements databases with full traceability.
AI meeting transcription and summarization tools. Many consultants currently use these to record stakeholder interviews, then manually analyze transcripts for requirements patterns.
The current DIY approach: consultants paste transcripts and notes into ChatGPT/Claude with custom prompts to extract requirements, identify gaps, and draft specs. Often combined with Notion or Confluence for documentation.
Requirements management tools that bolt onto Azure DevOps or standalone. Some have added AI features for requirements quality checking
Web app with three core features: (1) Upload transcripts/documents from discovery sessions and get AI-generated analysis highlighting contradictions, ambiguities, unstated assumptions, and the 'real problem' underneath stated requests. (2) A guided interview framework — suggested questions for each discovery phase that adapt based on what's been learned so far. (3) One-click export of a structured requirements document (problem statement, scoped requirements, open questions, risks). Skip team features, skip integrations, skip compliance. Target independent consultants first — they buy fast and give candid feedback. Use a waitlist from the Reddit thread's audience.
Free tier: analyze 1 transcript (lead gen, let people feel the magic) → Solo plan at $99/month: unlimited transcripts, full analysis, requirements export → Team plan at $249/month/seat: shared requirements library, collaborative editing, client-facing reports → Enterprise/Gov at $500+/seat/month: SSO, audit trails, compliance features, dedicated support, on-prem/GovCloud deployment. First revenue from Solo plan within 4-8 weeks of launch. Enterprise deals at month 6-12.
4-8 weeks to MVP and first paying users (targeting independent consultants via content marketing on Reddit/LinkedIn). 3-6 months to $5K MRR if you nail the positioning with the consulting community. 6-12 months to first enterprise/government pilot. The Reddit thread itself is a warm audience — 207 upvotes on the exact pain point you're solving.
- “I spent 6 weeks doing a discovery and framing for a branch of the DOD”
- “We spent 6 WEEKS literally just trying to figure out what their actual problem was”
- “understanding how to translate what executives think they want vs what they actually need”