PromenautUse case · Investment decisions
Business Case Evaluation
Large organisations evaluate dozens of major investment proposals every portfolio cycle. Each case runs to hundreds of pages. Senior analysts spend weeks reading, scoring, and writing committee memos, and the quality is inconsistent because it depends on individual judgement and how much time anyone has.
From raw input to signed-off output.
$ promenaut run business-case-evaluation
Evaluated against a live $377M investment case
Extract & validate
Parse the investment case PDF, extract sponsors, IDs, cost lines, benefits, OKRs, assumptions and dependencies into a structured object. A 15-point quality checklist surfaces missing fields before scoring begins.
Score in parallel
Five agents run simultaneously: financial ROI scorer (payback, IRR, NPV, benefit density), CEO strategic alignment, GCIO strategic alignment, architectural fit against the enterprise’s FSA matrix, and cultural fit. Each produces a score with cited evidence.
Risk & synthesis
A risk agent stress-tests assumptions and dependencies. The verdict agent weighs all lens scores and produces a Proceed / Rework / Reject / Defer call with cited reasoning and remediation asks.
Human review
The structured verdict goes to a human reviewer before it reaches any committee. Every claim is traced to its source.
How it actually works.
- Nine agents across three sequential stages: extraction, parallel scoring, synthesis.
- Strategic context (goals, architectural frameworks, leadership principles) is seeded once into the knowledge graph and reused across every evaluation run.
- The portfolio-level unlock: once multiple cases are evaluated, cross-case queries become possible: find every case aligned to a given strategic goal, sum costs, surface dependency overlaps.
- Decision-support by design: the workflow produces the verdict; humans decide.
What it produced.
Evaluated against a live $377M investment case
Minutes
To evaluate a case that took weeks to score manually
Five lenses
Financial, strategic, architectural, risk, and cultural, all at once
Cited verdict
Every conclusion traced to its source, with no black-box scoring
Consistent
Same framework applied to every case in the portfolio
As more cases are evaluated, portfolio-level analysis becomes available: find overlapping dependencies, surface strategic gaps, rank cases by weighted score across lenses.
The parts that break naive tools.
Manual scoring is inconsistent
Two analysts will score the same case differently. A structured framework applied consistently means the portfolio committee is comparing like with like.
Strategic context is tacit
Scoring against the CEO’s goals requires knowing what those goals are. The knowledge graph holds the strategic context; every run uses the same reference.
Volume caps coverage
A senior analyst can only read so many cases deeply. Scaling the reading and scoring lets the whole portfolio be covered, while the decisions stay human.
The verdict must be defensible
Every score, gap, and recommendation is cited to the source case content or knowledge graph entry, so a committee can trace "why Rework?" to the specific failing.
See it on a workflow you own.
Bring a real process. We'll show what your first agentic workflow looks like on the platform.
