TL;DR
- A CFO can deliver a credible 12-month AI portfolio plan in 90 days, with funding decisions and governance gates in place.
- The framework runs three workstreams in parallel: portfolio shaping, capability assessment, and governance scaffolding.
- The output is a board paper with three options, expected value ranges, and a recommendation. Not a strategy slide deck.
Why now
CFOs in Asia are increasingly the executive sponsor for AI strategy, not the CIO. PwC's 2025 Asia-Pacific CFO Survey found that 61% of mid-market CFOs in HK, SG, and JP now own at least co-sponsorship of the AI agenda, up from 28% in 2023.[^1] The reason is straightforward: AI spend is now material to the operating model, and material spend goes through finance.
If you are a CFO at a 200-1,000 person Asian enterprise, you do not need a 200-page strategy. You need a 90-day plan that produces a board-ready portfolio with funding gates. This is that plan.
The shape of the 90 days
The plan runs three workstreams in parallel.
Workstream A: Portfolio shaping. What are the candidate AI initiatives, what is each worth, what does each cost, what is the risk profile?
Workstream B: Capability assessment. What can your organisation actually deliver today, what would it take to close the gap, and where are the dependencies?
Workstream C: Governance scaffolding. Who decides, who approves, who monitors, and against what criteria?
By day 90 the three workstreams converge into a single board paper with three portfolio options, expected value ranges, and a recommendation.
Days 1-30: Discovery and shaping
Week 1. Constitute a steering group of 5-7 people. CFO chairs. Members include a business unit head, the head of IT or CIO, the head of HR, and the head of risk or legal. Avoid more than seven. Larger groups do not decide.
Week 2. Run a structured demand-side workshop. Each business function brings 2-4 candidate use cases with a one-page brief. Format: problem, current state, target state, indicative value, indicative complexity. You will end with 18-30 candidate initiatives.
Week 3. Filter to a longlist of 10-12. Filter criteria: clear owner, clear value mechanism, technical feasibility on commodity tools, and absence of regulatory blockers in your priority markets.
Week 4. Build the value-cost matrix. For each longlist initiative, expected annual value (range, with low/base/high), expected one-year cost (also a range), and expected payback period.
The key discipline in this month is to insist on ranges, not point estimates. A point estimate communicates false precision. McKinsey's State of AI 2024 found that AI value capture is more sensitive to operating-model design than to model selection, which means cost and value should both be modelled with confidence intervals.[^2]
Days 31-60: Capability and constraint mapping
Week 5-6. Capability assessment. Score the organisation across five dimensions: data foundation, talent, tooling, governance, and change capacity. Use a 1-5 scale per dimension. Anything below 3 in any dimension is a constraint on the portfolio.
Week 7. Constraint reconciliation. For each shortlisted initiative, check it against the capability map. An initiative requiring 4-out-of-5 governance maturity cannot run in an organisation scoring 2.
Week 8. Vendor and partner market scan. For each initiative, identify three viable delivery options: build, buy, partner. Document indicative costs and the decision logic. The Wardley Mapping framework (Wardley, 2018) is useful here for distinguishing strategic builds from commodity buys.
The most common mistake in this month is to fall in love with the candidate initiatives and skip the constraint check. The result is a portfolio you cannot deliver. Do not skip the constraint check.
Days 61-90: Governance and the board paper
Week 9. Governance design. Define the AI investment committee (often the steering group, formalised), the approval thresholds, the review cadence, and the kill criteria. Three approval thresholds typically work: under US$50,000 (delegated), US$50,000-US$250,000 (committee), over US$250,000 (board).
Week 10. Risk and assurance design. Map each initiative against the relevant regulatory regime (PDPA in Singapore, PIPL in China, APPI in Japan, PIPA in Korea, PCPD guidance in Hong Kong). Identify which initiatives need a Data Protection Impact Assessment.
Week 11. Construct three portfolio options. Conservative (3-4 initiatives, low risk, US$0.4M-US$0.8M total spend, 18-month payback). Balanced (6-7 initiatives, mixed risk, US$1.0M-US$1.8M total spend, 12-15 month payback). Ambitious (8-10 initiatives, mixed risk including one strategic build, US$2.2M-US$3.6M total spend, longer payback but higher ceiling).
Week 12. Board paper. 12 pages maximum. Three options, expected value ranges, capability gaps and how they will be closed, governance design, recommendation. Pre-brief the board chair and the audit chair before the meeting.
Implementation playbook (run this Monday)
- Day 1: Send the steering group invitation. Names, dates, the 90-day calendar.
- Day 3: Brief each business function head individually on what you need by week 2.
- Day 7: First steering meeting. Confirm scope, governance, and the week 2 workshop.
- Day 14: Demand workshop. Use a structured one-pager template. Reject submissions that do not fit the template.
- Day 21: Longlist agreed in steering meeting.
- Day 30: Value-cost matrix circulated.
- Day 60: Constraint check complete, vendor scan circulated.
- Day 90: Board paper submitted.
Counter-arguments
"Ninety days is too slow. The market is moving." It is not too slow if you do parallel workstreams. It is too slow if you run them serially. Most strategy efforts take six to nine months because they run serially.
"This is too structured. We need more experimentation." The Cynefin framework (Snowden, 1999) is right that AI is a Complex domain that requires probing. The 90-day framework is the structure around the probing. Without the structure the probing produces no portfolio decision.
"Our board does not want options. They want a recommendation." Give them the recommendation. The three options are scaffolding for them to interrogate. A recommendation without alternatives is a vendor pitch, not a portfolio decision.
Bottom line
A CFO can deliver a credible AI portfolio plan in 90 days if the work runs in three parallel workstreams and ends with a board paper, not a strategy deck. The deliverable is decisions: how much to invest, where, who is accountable, what gets killed if it underperforms.
Most AI strategy efforts in 2024-2025 produced strategy artefacts that no one used to make decisions. This framework produces decisions. The artefacts are by-products.
Next read
- What Asian Mid-Market AI Pilots Actually Cost in 2026
- Reading the AI Vendor Hype: 9 Questions to Ask Before You Sign
By Hyejin Lee, Director, CFO Advisory.
[^1]: PwC, 2025 Asia-Pacific CFO Survey, March 2025, p. 22. [^2]: McKinsey & Company, The State of AI in 2024, May 2024, p. 14.
Where this applies
How AIMenta turns these ideas into engagements — explore the relevant service lines, industries, and markets.
Beyond this insight
Cross-reference our practice depth.
If this article matches your stage of thinking, the underlying capabilities ship across all six pillars, ten verticals, and nine Asian markets.