Problem
- Spinning up a chat, agent, or skill takes minutes with no IT involvement. Building a production workflow requires developer time, IT clearance, and weeks of delay — 88% of AI POCs never reach production (IDC/CIO 2024). The implementation bottleneck, not the AI capability, is the barrier.
- Non-deterministic workflows require constant human review — 5–30× token overconsumption vs equivalent deterministic pipelines (Stanford; actual range 5×–1 000× depending on design).
- Most AI solution requests are deterministic workflows in disguise — users can't define or code them; IT can't govern what it can't see.
- Skills and agents are session-bound — autonomous scheduled execution requires either API-rate billing or 20–40 hours of custom scheduling infrastructure. A workflow that already works cannot run nightly without re-platforming.
→ Validated across all 4 groups (06). Hermes: 40–60% of dev effort on manual session-to-workflow conversion. OpenClaw: 100× cost reduction documented. LoB: 156-day procurement cycle. Enterprise IT: 95% of AI pilots fail to reach production.
→ Pain #4 validated (04-research-findings §9): Paperclip open-source scheduler — 43k GitHub stars in 6 weeks. Anthropic June 2026 billing split: automated use (claude -p, Agent SDK) carved into a separate metered credit pool, separate from interactive subscription. GitHub #28229: "Sessions are ephemeral — there is no native way to run an agent on a schedule." Workaround requires 400+ lines of custom infrastructure per agent.
Existing Alternatives
- Model downgrade (quality loss accepted)
- Manual LangGraph / n8n coding (high skill barrier)
- Lobster engine (OpenClaw community, Feb 2026) — deterministic runtime; 20–40%+ token reduction on routing-heavy pipelines; requires manual typed pipeline authoring
- Temporal.io — durable workflow orchestration; production-grade (SOC 2, HIPAA); requires full hand-authoring per project; Phase 3 integration opportunity — Workflowify output deployable as Temporal Activities
- Cost caps and kill switches (reactive)
- LangSmith / Langfuse / LiteLLM — surface or route costs; don't convert agent sessions
- Power Automate / Zapier / Make — business process automation; not agent-trace aware
- Abandonment — the 88% outcome
Solution
① Local (input)
- Read session / skill.md → strip to structural skeleton → send to SaaS build
② SaaS build (online)
- Extract steps · classify (det. vs non-det.)
- Route extraction + conditional branches
- Edge case & failure mode identification
- Translate + generate workflow — LangGraph v1; n8n in parallel
③ Local (deployment)
- Provision runtime · correct source skill
- Schedule — first-class output: wire up via launchd / cron / systemd timer / n8n scheduler / LangGraph cron — one CLI flag. No Claude session or subscription quota required at run time.
- Workflow runs locally — no runtime data to workflowify
④ SaaS services (ongoing)
- Template library · team sync · governance dashboard
- Tool update monitoring → re-generation offers
- Guardrails · analytics · audit logs (Vx)
⑤ Web service — v2/v3 (LoB path)
- LoB user describes process in plain English via web wizard — no CLI, no trace required
- Guided generation with human-in-the-loop validation → deployed running solution
- Hard cap on deployments per period; additional deployments purchasable as extra units
- Options (SSO, audit, self-host) available as add-ons — same modular model as Enterprise
- Entry point for non-technical buyers; qualifies leads for subscription upsell
✓ Architecture: CLI-first — maximise deterministic processing; LLM used only for classification (low-cost model) and generation. Pre-flight trace quality check guards against malformed input.
Key Metrics
- Workflows generated per week — primary adoption signal
- Cost reduction per workflow (tokens before vs after) — core value proof
- Skills corrected / replaced — closed-loop adoption
- Free → Pro conversion rate — PLG health signal
Unique Value Proposition
Turn any AI agent session into a faster, cheaper, auditable workflow — with one command.
Built for AI developers who know what a trace is. If you can describe what your agent does step by step, this tool is for you.
High Level Concept
The /skillify for production.
/skillify captures what the agent does. /workflowify makes it run at scale without burning tokens.
Made for AI builders.
If you don't know what a trace is, use the web service (→ v2/v3). If you do, one command is all you need.
Unfair Advantage
- Founder is the target customer — daily Claude-based agent usage (cos-morning, job-scan). Firsthand pain = product instinct.
- First mover in confirmed white space — no production competitor exists.
- Community-first distribution in Hermes / OpenClaw ecosystem.
- Workflow template library → network effect over time.
⚠️ Weakest block. No hard moat yet. Speed to community adoption is the strategy — not a guarantee.
→ Differentiation vs Lobster engine (closest partial competitor): Workflowify generates from observed traces automatically. Lobster requires manual typed pipeline authoring. Target: Lobster users who want automation, not authoring.
→ Endorsement outreach: Jeffrey Quesnelle (NousResearch/Hermes) · Peter Steinberger (OpenClaw). Contacts confirmed. Contribute demo template first.
Channels
- Hermes / OpenClaw community (Reddit, Discord, GitHub)
- GitHub open-source CLI release
- Product Hunt launch
- LoB self-serve — Pro at $19.90/mo — well below $50/mo expense threshold; no IT approval needed
- Web service (v2/v3) — LoB users get a deployed workflow without CLI or trace knowledge; one-time fee or concierge tier
- Enterprise direct outreach (IT / LangGraph / n8n leads) — second-wave via LoB pull
→ CAC modelled: $300–800 (SMB) · $800–1,400 (mid-market) at $6–24k ACV (04-research-findings).
Customer Segments
Primary (revenue)
- Enterprise IT / CIO — cost governance, compliance, access control — Enterprise tier buyer
- Enterprise power users / builders — build skills, want production deployment
Secondary (adoption → entry)
- Solo AI developers on Hermes / OpenClaw
- LoB managers — buy at expensable Pro price; push LoB adoption; pull IT toward Enterprise governance
⚠️ B2B qualification (SSO, audit, DPA) required to convert LoB adoption to Enterprise contracts — target v2.
Early Adopters
Hermes / OpenClaw power users who have already built skills and hit the token cost or production wall.
Concentrated in community (Reddit, Discord, GitHub). Willing to try a CLI tool before a polished product exists.
Second wave: LoB managers expensing Pro independently — no IT clearance needed at launch. Web service path (v2/v3) will serve non-technical LoB buyers without CLI skills.
Cost Structure
- LLM API — classification (Haiku / GPT-4o Mini) + generation (Sonnet + caching + Batch API) — modelled at $0.02–0.05/generation optimised; <5% of Pro COGS at 10 gens/user/month
- SaaS infrastructure — API hosting, template library, team sync, governance dashboard
- Developer time (founder) — dominant cost in v1
- Community / GTM — low in open-source CLI model
→ Architecture principle: CLI-first; maximise deterministic processing before LLM calls. Prompt caching from day one. Free tier rate-limited to 5 gens/month to protect margins.
Revenue Streams
- Web service (v2/v3) — subscription with hard deployment cap; additional units purchasable; modular options (SSO, audit, self-host) as add-ons — LoB concierge path; leads into CLI subscription
- Free — Open-source CLI + rate-limited SaaS (hard cap, $0) — community adoption
- Pro — SaaS ($19.90/month · $199/yr) — hard generation cap; additional Pro unit purchasable when surpassed — expensable; well below typical $50/mo expense threshold
- Enterprise bare — SaaS ($49.90/month · $499/yr) — base tier; hard cap + additional units; core governance features
- Enterprise + options — modular add-ons (SSO · audit logs · self-hosting / air-gap · guardrails · advanced analytics · team sync) — each option priced separately; bundled tiers TBD or separate offer at much higher ACV
→ Pricing model: hard caps throughout; overage = buy another subscription unit (no bill shock). Annual = 10 months billed (≈17% discount).
❓ Define: option pricing per add-on vs bundled Enterprise Premium offer. Generation cap values per tier TBD before v1.
→ LTV modelled (04): Pro LTV thin at $19.90 — volume play. Enterprise + options is the revenue engine; ACV depends on option uptake.