Edge for the Fund
Portfolio AI fails when it runs as N independent projects. We start by identifying the work: every PortCo placed on one map, every workflow ranked by EBITDA impact, before anything is built. Then parallel pods prove the highest-conviction use cases, and what ships in one company transfers to the next.
The L1–L5 tier is one axis. The second is use-case clarity: does the company know what to build, or does it need discovery? Edge Assess places every PortCo on this map with a read-only service account, in days, not a quarter of workshops. The map drives sequencing: who moves first, who needs substrate, who waits.
Strong substrate, unranked work. Inventory first.
Clear use case, substrate in place. Pilot now.
Low maturity, unclear work. Scan and sequence.
Named use case, weak substrate. Build the floor.
This is the core discipline: AI value starts with work identification, not tool selection. Two value levers, nine use-case categories, each banded by EBITDA impact, time-to-value, and the confidence we’d defend in front of an IC.
| Use-case category | Representative work | Typical impact | Time-to-value | Confidence | Tier req. |
|---|---|---|---|---|---|
| Revenue growth | |||||
| B2B sales effectiveness | Predictive lead scoring, next-best-action, proposal/quote copilots, account prioritization | 5–15% sales productivity (10–50 bps) | 6–12 wks | Medium | L2–L3 |
| Pricing & margin optimization | Dynamic pricing, discount guardrails, quote margin guidance, renewal/contract pricing | 1–3% gross margin uplift (50–200 bps) | 8–16 wks | High | L2–L3 |
| Customer retention & aftermarket | Churn prediction/save flows, reorder prediction, spare-parts recommendation, warranty analytics | 1–2% revenue retention (10–75 bps) | 8–16 wks | Medium | L3 |
| AI-enabled differentiation | Customer self-service agents, technical documentation assistants, customer-facing reporting | Stickiness, strategic (25–100 bps) | 12–24 wks | Low | L3–L4 |
| Cost efficiency | |||||
| Procurement & sourcing | Supplier scouting, quote comparison, negotiation copilots, spend classification, supplier risk | 2–5% addressable spend (50–150 bps) | 8–20 wks | High | L2–L3 |
| Product & process development | AI-driven BOM, design-to-cost copilots, requirements processing, engineering copilots in CI | 1–3% material cost, select categories (25–150 bps) | 6–12 wks | Medium | L1–L3 |
| Manufacturing ops & quality | AI-powered inspection, predictive maintenance, production scheduling | 10–30% lower scrap/rework (25–150 bps) | 10–20 wks | Medium | L3 |
| Supply chain & logistics | Demand forecasting, inventory optimization, route optimization, warehouse labor planning | 5–10% logistics cost, 10–25% inventory (50–150 bps) | 10–20 wks | Medium | L3–L4 |
| Back-office & SG&A | Order intake automation, AP/AR automation, contract review, HR workflows, document processing | 10–30% lower SG&A, select areas (50–150 bps) | 6–16 wks | High | L2–L3 |
Bands from measured engagements and cited benchmarks. Ranges, not promises. Calibrated to $50–500M revenue companies.
Two named pods, two PortCos, one timeline. Rapid prototyping starts the moment a use case is concrete, and production implementation lands inside the 90-day window. Pods scale up or down at the fund’s discretion.
Scan sorts the cohort first: ready-to-build companies skip to design; discovery-track companies run the same spine with one extra stage. The five-stage spine and seven quality gates govern every build.
Work identification is a funnel, not a workshop: every PortCo’s workflows enter, and only the use cases that survive a written business case and a measured POC earn production capital. Two gates on the way, both signed.
Inside each build, the seven engagement-level quality gates still apply. G1 and G2 are the portfolio wrapper.
Three things compound across a portfolio: the workflows, the platform, and the playbooks. This is why the portfolio motion gets faster and cheaper with each company instead of scaling linearly with headcount.
Proven agent workflows packaged as skills: prompts, tool wiring, eval suite, runbook, versioned in source control. The variance-close agent built at PortCo A is the starting point at PortCo B, not a blank page.
One control plane running in production across the cohort: governance, audit, per-agent cost attribution, evals. Not a future platform slide; the runtime the agents already run on.
Codified plays: per-category implementation runbooks, gate criteria, eval baselines, and banded economics from every measured engagement feeding back into the inventory.
Platform-era reuse demanded standardized data structures and common systems before transfer. Agentic systems don’t. With the right context and a proven skill, a workflow moves between PortCos even when data is messy and processes differ; the agent adapts to the company, not the company to the platform. Transfer is hardest where the business process is atypical and data access is poor, exactly what the scan detects.
The traditional strategy-consulting motion sells the operating model and rents you the labor. Edge ships the operating model as a product: diagnostic, value engine, and orchestration plane.
| Traditional strategy-consulting motion | The Edge motion | |
|---|---|---|
| Discovery | 4–8 week paid workstream plus a multi-week data request | Read-only diagnostic: days, portfolio-wide |
| Prioritization | Workshop-driven hypotheses | Codified plays, banded EBITDA thesis |
| Delivery | Pods of consultants plus borrowed engineers | Named pod shipping production agents through signed gates |
| Production | "Evaluated after the pilot" | In scope from day one, inside the 90-day window |
| Reuse | Requires a common platform and standardized data | Skills and context transfer across PortCos as-is |
| Price | $500K+ before production is scoped | Published bands, fixed |
| Portfolio scale | Linear: more PortCos, more consultants | Compounding: every workflow shipped makes the next cheaper |
Published investment bands are on the framework page. Ranges, not promises.
