Trillium Legal Ops, a fictional Series B vendor selling matter-management software to mid-market law firms. Seven AI bets in flight, no allocation defense, board pressure rising. Below is the artifact shape a real engagement produces: portfolio map, allocation table, named failure modes per horizon, kill list, and the language the CEO uses in the next board meeting.
Vertical SaaS for mid-market law firms. Matter management, billing, intake, conflicts checking. Series B closed 18 months ago at $280M post-money. Two GP investors have asked the CEO three quarters running to articulate the AI strategy. COO is the de facto AI sponsor. Head of product ships pilots. Head of engineering is quietly worried about eval coverage. Seven AI bets in flight, none with a documented horizon, reach, or kill criterion. This is the company the AI Portfolio Allocation Review is built for.
Three at Core, two Adjacent, two Transform. The classification surfaces what's overweight, what's missing, and which bets sit on the wrong horizon for their funding profile.
Status: In production, ~120 paying firms
Draft redlines inside the existing matter workflow. Owned by the legal-ops product team. Paying back inside the core on seat expansion. The clearest H1 win in the portfolio.
Status: Pilot, 8 firms
Generates the time-entry narratives associates currently write by hand. Pilot started 9 months ago, no production owner named. Sitting on the boundary between pilot purgatory and a clean H1 ship.
Status: Internal beta
Runs the conflicts query across the matter graph, surfaces ambiguous matches for human review. Built by one engineer in the gaps. No PM, no eval set. Real demand from firms.
Status: Build phase, 4 engineers
Coordinator agent that triages new client matters, routes to the right specialist agent (litigation, IP, transactional), and packages the case for the intake partner. New revenue line as a premium SKU. One step out from the current matter-management product.
Status: Research, 1 ML engineer
Fine-tuned on the firm's redline corpus, scoped to commercial real estate. Promises better redline quality than the general copilot. Demo-only today; no integration path to the production product.
Status: Concept, no funded team
From precedent database plus the firm's matter history, generate a litigation strategy memo a senior partner can edit. A new product line. Discussed in two board meetings, no roadmap.
Status: Concept, executive interest only
Reads case file, generates witness questions, drafts cross-examination outlines. The CEO mentions it on every customer call. Engineering has not been asked to size it.
Trillium's current implicit allocation is roughly 45% H1, 45% H2, 10% H3. The review's recommended mix moves H1 up to defend the core, holds H2 at 20% with a stricter handoff plan, and frees a real budget line for H3 instead of leaving it as concept slides.
H1 defends the core with three production bets at Core reach. H2 funds one capability build one step out with a named handoff plan. H3 carries one funded discovery sprint and one formally parked concept under quarterly review. The kill list (next section) removes the vertical-specific contract model.
Each horizon carries a named failure mode the review forces the team to flag honestly. Trillium is exposed on all three.
Billing-narrative drafter and conflicts-check automation have lived in pilot 9-plus months with no production owner. Fix is mechanical: name a product owner, set a production date, move the bet to ops's roadmap. Ship this quarter or kill.
Multi-agent matter intake is built by a research-style team with no production integration path. Review forces a named handoff plan to the matter-management product team before another dollar is spent.
Both H3 bets sit with no budget line and no named owner. Review unfreezes them: fund a discovery sprint (case-strategy generator) or formally park with a quarterly review date (deposition-prep agent). No more vague mentions in board decks.
The kill list usually saves more than the engagement costs. Trillium's three calls below free a head of ML engineer, retire a vendor contract, and absorb a stalled pilot into a shipping product.
A research project with no integration path that competes for budget with the production contract redline copilot. Rationale: the general copilot covers 80% of the value at a fraction of the maintenance cost. Reassign the ML engineer to the H2 matter-intake build.
An internal beta built by one engineer in the gaps. Rationale: this is an H1 ship, not a side project. Fold it under the product team that owns matter management, give it a named PM, set a production date, and ship.
A CEO-favored concept with no engineering sizing. Rationale: park it formally on the H3 roster with a quarterly review date. The CEO can still talk about it on customer calls, but the engineering team is no longer being asked to context-switch on it ad hoc.
The artifact ships with language the CEO can quote, not just charts the board has to interpret. Three sentences Trillium's CEO drops into the next board read-out. They neutralize "where is our AI strategy" without overpromising.
"We've split our AI spend across three horizons, measured by reach not time. Seventy percent defends our core matter-management revenue with three production bets."
"Twenty percent funds one Adjacent capability, multi-agent matter intake, with a named handoff plan to the product team so it ships instead of staying a demo."
"Ten percent is allocated to a single Transform discovery sprint on agentic case strategy, with the second transformation bet formally parked under a quarterly review. We will not fund H3 narrative without a discovery budget behind it."
A real Trillium engagement runs four weeks: interviews with 8 to 12 across product, engineering, finance, revenue. Four working sessions. Board-ready memo. Kill-list rationale your CFO scrutinizes line by line. The page above is the output shape. The engagement is the work that gets you there.
Three to four weeks. Me plus your CEO or COO plus one finance partner. Board-ready output. The kill list usually pays for the engagement on its own. I hold ~4 active client commitments across Fractional, Scoped Build, and Audit. Mix depends on what's already booked.