Home / Services / Portfolio Allocation Review
Stop picking AI projects one at a time. In 3 to 4 weeks I classify every AI bet you have across H1 core, H2 adjacent, and H3 transform reach. You walk out with a board-ready roadmap and the language to defend it.
A module for execs defending an AI budget. See where it fits in the three-step path.
I borrow McKinsey's Three-Horizon model from 1999. It was built for growth portfolios, and it works on AI with one change: I measure the horizons by reach instead of time. AI moves too fast for time windows to hold. How far a bet sits from your current operating model stays put.
Each horizon gets its own protected budget, its own measurement standard, and a named failure mode.
H1 · About 70% of budget
Workflows your business already runs. Document processing, lead routing, internal copilots. Efficiency plays on existing operations.
Failure mode: pilot purgatory. Wins that never reach production because the innovation team owns them instead of operations.
H2 · About 20% of budget
New agents on existing problems and new capabilities on the same operating model. Custom multi-agent workflows, AI-native features, vertical models.
Failure mode: stalled handoff. Builds that work in a demo and never connect to the core business.
H3 · About 10% of budget
A new operating model, business model, or product line. Org redesign, orchestrated agents replacing whole workflows, AI-first product categories.
Failure mode: frozen bet. Ideas talked about for two years and never funded, because no one will defend them.
Each one is built to survive executive scrutiny.
Every active bet, paused pilot, and backlogged idea, classified across H1, H2, and H3. Each one tagged with status, owner, and expected payback.
Why the current mix is wrong, what to move, and the language to defend it to your board. Every bet gets a baseline before it gets a budget, scored on cost, revenue, time, and risk. Named failure modes per horizon, so you flag your own risks first.
What ships at the core this year, what gets seeded one step out, and what gets a discovery budget at the transform horizon. Quarterly milestones, named owners, measurable outcomes.
Pilots to retire, vendors to consolidate, and teams to redirect. The goal is to save more here than the engagement costs.
Me, your CEO or COO, and one finance partner. Four working sessions plus interviews with 8 to 12 people on your team, timed so the artifact lands in your next board cycle.
Week 1 · Inventory
I interview operators, engineers, and executive sponsors, and surface the pilots nobody wants to admit failed.
Week 2 · Classification
Each bet goes on the H1, H2, H3 grid, scored by distance from the core, capital required, strategic fit, and risk. The gaps show up here.
Week 3 · Allocation
Reallocate across horizons, draw up the kill list, and stress-test the result against the named failure modes.
Week 4 · Board package
An executive deck or memo. I walk your team through the defense and hand over the artifact your CFO will quote in next quarter's planning.
My main path is three steps: the plan, the build, and keep me on. The Portfolio Review sits beside it for an exec who has to defend an AI budget. It usually follows the plan, once the workflow map shows what is in flight and the board wants an allocation. When the question on the table is where the budget goes, it can replace the plan. Either way, the next step is the build.
Two weeks. The workflow map, the 90-day sequence, and the kill list.
Where the main path starts. Credited in full toward the build or a retainer within 30 days.
This page. Every AI bet classified by reach, with a board-ready roadmap.
Usually follows the plan. For a budget question, it can replace it.
The H1 wins, shipped into the stack your team already runs.
From $7,500 a workflow, two-workflow minimum.
AI System Continuity keeps it running. The Fractional retainer owns the program and the next builds.
Fractional is 1 to 3 days a week, 6-month minimum.
You can check this guarantee by reading the review. There is no savings estimate to argue about, only whether every bet carries the four things your board needs to decide on it.
Full refund
By the end of week 4 you get the review. Every AI bet in it has a named owner, its current spend from your own invoices, a horizon, and a call: fund, hold, or kill. If the readout does not happen by the end of week 4, or any bet is missing any of the four, you get the fee back. Assumes you give me the invoices and contracts on the kickoff list.
Exit
Either party can end the review inside the first 14 days. Days worked are billed at the agreed day rate, with no further commitment.
A 3 to 4 week module that classifies every active AI bet, every paused pilot, and every backlogged idea across three horizons: H1 Core (about 70% of budget), H2 Adjacent (about 20%), and H3 Transform (about 10%). It adapts McKinsey's 1999 Three-Horizon growth model. You get a board-ready roadmap with named failure modes per horizon and the language to defend the budget.
It runs 3 to 4 weeks and is scoped against the number of active AI bets and the number of stakeholders involved. I price it in writing within 48 hours of the discovery call.
The plan is where the main path starts. The review usually follows it, when the workflow map has shown what is in flight and the board wants an allocation. For an exec whose question is where the AI budget goes, it can stand in for the plan. Either way, the next step is the build.
AI moves too fast for time windows to hold. A 12-month roadmap is stale by month four. Reach, meaning how far the bet sits from your current operating model, stays anchored across model changes. H1 Core is the work your business already runs. H2 Adjacent is one step out. H3 Transform is a new operating model.
Yes, especially. The review is worth more before the portfolio gets messy. Setting allocation rules is easier than unwinding bad ones.
Generic AI strategy ends in a deck. This ends in an allocation defense your CFO will sign off on, and I write the language for the board memo.
No. This is the portfolio review. Shipping the H1 wins is the Scoped AI Build, priced per workflow. The review tells you which ones to ship.
Good. That is the engagement. The framework forces the disagreement into the open so it stops stalling the portfolio.
Three diagnostics at a time. Builds and retainers are capped at four client commitments at a time.
Per workflow, against a baseline taken before anything is built. If you cannot say you were at X and now you are at 3X, the project failed. The math is the full cost of running the workflow, tooling plus the human review it still needs, against what it costs a person to do the work today, then the breakeven point. Measure business outcomes, not AI usage. A UK government trial put Copilot in 1,000 hands and found 1.14 actions per user per day with no solid productivity evidence, while 72% of users said they were satisfied. Usage felt good and the numbers did not move.
Tell me what the board is asking. I will tell you whether you need the review, the plan, or both.
Book 30 minutes with EdwinNot ready to talk? Score your AI program first. Twelve questions, about four minutes, scored breakdown by email.