The boldest bet is Shadow Company: watch how a firm really works, then manufacture its agents
Asked (summary)
This product sits in a weak, crowded space with serval.com — what else could this team build, especially given long-horizon agents and strong LLMs?
Asked (follow-up)
“give me 5 more ideas.”
Five horizontal product theses, each with a credible $200K+ initial ACV and a path past $1M per account, ranked on a qualitative 10-point matrix and backed by 27 source-linked evidence rows across 25 hosts — much of it vendor or secondary material, so economics and outcome claims here are directional. These are high-conviction hypotheses, not validated businesses.
The Shadow Company loop
Instead of selling another agent to one department, Action becomes the system that discovers an organization's real work and manufactures validated agents from observed trajectories. Hover or tap a stage.
The loop runs continuously: every pass deepens the company-specific work graph and trajectory corpus — the moat. Buyer: COO / CIO / Chief Transformation Officer. $500K–$1M for several functions; $2M–$10M enterprise-wide.
Work discovery already pays at enterprise scale: Orange France reports €300M in business value over four years and 500 days of labor saved through process mining — Shadow Company goes further, from maps to manufactured agents. uipath.com
Company War Room attacks a $374.67B (2025) management-consulting budget pool projected toward $761.54B by 2034 — but measured predictive validity of simulation is weak, so it must ship scored forecasts, not theater. maximizemarketresearch.com
Experimentation at scale is worth real money — PayPal reports roughly $180M of measured impact in 2025, readouts cut from four weeks to 24 hours (secondary, directional). podtail.com
Enterprise intelligence subscriptions already run $80K–$500K a year, and custom role-play simulations reach $100K–$1M per project — budget exists for Research Network and Simulation Academy to reprice. marketintel.co.in · virti.com
How the five bets score
Qualitative 10-point matrix. Shadow Company leads on fit, horizontality and ceiling; its weakness is validation speed — hence the 90-day plan below.
The five bets
Rank 1 · strongest recommendation
Shadow Company
What
A living computational model of how a company actually works: observe consenting workflows, reconstruct the work graph, generate supervised agents from it.
Buyer
COO, CIO, Chief Transformation Officer
Price
$500K–$1M for several functions; $2M–$10M enterprise-wide
Wedge
One 100–500-person ops org with 50+ repeated cross-system processes; a 90-day “Work Graph” SKU mapping the top 20 workflows plus three shadow-mode agents.
Why huge
The buyer funds continuous transformation, not ticket automation; expansion per function and per production agent.
Moat
Company-specific work graph, trajectory corpus, human/agent agreement data.
Fatal flaw
Surveillance optics, privacy and works councils; must be opt-in, role-transparent and workflow-focused — never employee scoring.
Rank 2
Autonomous Research Network
What
A managed network of research agents continuously recruiting and interviewing customers, ex-customers, employees, suppliers and experts; verifying claims; maintaining a queryable, source-linked market model.
Buyer
CSO, CPO, CMO, Corp Dev, PE operating partner
Price
$250K–$500K minimum; $1M–$3M for global coverage
Wedge
Companies needing 500–2,000 customer conversations a year; a query like “why are German mid-market customers churning?” returns a decision packet within days.
Why huge
Replaces episodic projects, expert-network calls and stale intelligence subscriptions with continuous primary research.
Recruiting economics and interview quality; consent, expert-network compliance, hallucinated synthesis.
Rank 3 · proof burden
Company War Room
What
A dynamic decision simulator: a causal company model with agents playing customers, competitors, employees and regulators; war-games pricing, market entry, layoffs, M&A, bundling.
Buyer
CEO, CFO, CSO, PE operating partner
Price
$500K–$1.5M minimum; $2M–$5M global / PE portfolio
Wedge
Annual planning and one board-level decision at PE-backed or public software/services companies; budget sits with strategy consultants today.
Why huge
Surfaces assumptions, second-order effects and leading indicators, then updates as reality arrives — no prediction promise.
Moat
Calibrated company-specific causal model plus decision/outcome history.
Fatal flaw
Becoming impressive theater: measured predictive validity is weak, so it must backtest, state uncertainty and score forecasts.
Rank 4 · cheaper competitors
Simulation Academy
What
A continuously generated flight simulator for every high-stakes role, built from real policies, products, incidents and rubrics; agents play customers, employees and adversaries; the system certifies readiness.
Buyer
CHRO, CRO, COO, Chief Learning Officer
Price
$300K–$750K minimum; $1M–$5M global workforces
Wedge
Enterprise sales and frontline-manager certification, tied to ramp time, quota, deal size, escalation rate, retention.
Why huge
Custom enterprise simulations already reach six-figure to million-dollar project bands (directional); this generates them from live company reality.
Moat
Scenario generation from the company's own reality; performance-based certification, not content libraries.
Fatal flaw
Fragmented training budgets, gameable scoring, and a field of cheap AI role-play vendors; must own certification across roles.
Rank 5 · causal complexity
Business Experiment Foundry
What
A managed OS for controlled experiments outside the product UI — pricing, packaging, sales motion, incentives, policy — from hypothesis to approved protocol, guardrails, rollback and causal readout.
Buyer
CPO, COO, Chief Growth Officer, CFO
Price
$300K–$750K plus a share of measured upside; $1M–$5M at high experiment volume
Wedge
Pricing and packaging tests for B2B software, then sales and service operations.
Why huge
Makes “test it safely” possible for business changes that today take months of meetings and custom engineering; measured impact at experimenters like PayPal is large (directional).
Moat
Cross-system experiment execution, causal measurement, intervention library, outcome history.
Three design partners at $100K+ each; instrument one consenting 50–100-person operating group per company across at most five systems.
Build
Reconstruct 20 workflows with 80%+ human validation of steps and dependencies; generate three executable agents in shadow mode.
Measure
85%+ agreement with expert human decisions on bounded cases; show 30%+ cycle-time reduction potential.
Convert
Two partners to $500K+ annual contracts.
Kill if
Observation data is too incomplete, employees reject it, every workflow needs bespoke consulting, shadow performance cannot be objectively measured, or gross margin cannot reach 70%+.
Evidence
All 27 source-linked rows. Scores are retrieval relevance (0–1). Vendor and secondary claims are labeled by type; treat economics as directional.
Type
Entity
Key claim / metric
Source
Method: 27 curated source-linked evidence rows across six categories (Action capability, work discovery, company simulation, enterprise research, simulation training, business experimentation), drawn from 25 hosts; compiled 2026. Scores are retrieval relevance on a 0–1 scale; matrix scores are qualitative judgments on a 10-point scale, not measured data. Pricing and outcome figures are vendor or secondary claims and directional. Long findings and metrics truncated for space; links give the full source.