Most AI business cases that reach an FMCG board fail for the same reason regardless of industry vertical, use case, or vendor — they lead with model accuracy and technology capability instead of a specific, measurable business outcome tied to a dollar figure the CFO can trace through the P&L. A board approving capital for a new packaging line asks about payback period, capacity impact, and downside scenario without hesitation, yet the same board frequently receives AI proposals structured around confidence intervals and precision metrics that answer a question nobody on the board actually asked. The business cases that get approved are structured differently from the start, staged in outcome-tied phases with explicit downside protection built in before the first dollar is spent. If your last AI proposal got tabled for "more information," book a demo to see a board-ready business case structure built for FMCG.
Your Board Doesn't Need to Understand the Model. They Need to Trust the Number.
The AI business cases that get approved by FMCG boards are structured around outcome-tied use cases, staged value delivery, and explicit downside protection — not model architecture or accuracy metrics.
Why Technically Sound AI Proposals Still Get Rejected
A board does not reject an AI investment because the underlying technology is weak — most rejections happen because the proposal was built to satisfy the data science team's sense of rigor rather than the board's actual decision-making framework. Boards evaluate capital requests using a consistent mental model regardless of the underlying technology, and AI proposals that ignore this model create unnecessary friction.
Leading with model precision or F1 scores answers a technical question the board isn't equipped to evaluate, when what they need is the specific business metric the model is expected to move.
A single large capital request for full-scale deployment, with no intermediate proof point, asks the board to underwrite the entire risk before any evidence exists that the approach works.
A proposal that only models the upside scenario leaves the board to imagine the failure mode themselves, and an unaddressed risk is a reason to say no rather than yes.
When no specific operating leader is named as accountable for the outcome, the board reasonably questions whether anyone will actually be responsible if the projected value doesn't materialize.
The Three-Stage Value Structure Boards Actually Approve
Structuring the investment ask around proof stages rather than a single upfront commitment lets the board approve a small, bounded first step while reserving judgment on the larger commitment until real data exists.
Prove — Single Line or Category
A bounded pilot on one line or product category, with a specific measurable outcome defined in advance and a hard deadline for evaluation, typically 60 to 90 days.
Expand — Validated Use Case, Wider Scope
Once the pilot's outcome is confirmed against its target, expansion to additional lines or categories with the same measurement discipline, refining the model against a larger dataset.
Scale — Enterprise Deployment
Full deployment across the network, justified by two rounds of validated evidence rather than a single pilot, with the business case refined using real production data at each stage.
Build a Business Case Your Specific Board Will Approve
Bring your board's typical capital approval criteria and we'll help structure the use case, staging, and downside protection to match how your board actually evaluates investment.
What Each Board Member Is Actually Evaluating
A single business case document has to satisfy several different evaluation lenses simultaneously. Understanding what each role is specifically looking for prevents the proposal from over-indexing on one perspective at the expense of the others.
| Board Perspective | Primary Question | What the Proposal Must Show |
|---|---|---|
| CFO | What is the payback period and what happens if it underdelivers? | Staged investment, modeled downside, clear cost-to-value ratio |
| COO | Does this disrupt current operations or add operational risk? | Pilot scope, rollback plan, operational ownership assigned |
| CEO | Does this align with the stated strategic priorities for the year? | Explicit connection to a named strategic initiative or board priority |
| Independent Directors | Has this kind of investment been validated elsewhere in the industry? | Comparable case evidence, realistic benchmark ranges, not vendor claims alone |
Building the Downside Case the Board Won't Have to Ask For
Every proposal that only shows an upside projection implicitly asks the board to trust that nothing will go wrong. A proposal that proactively models the downside and shows how it's contained earns more trust than one that avoids the topic entirely.
Stage one investment is sized specifically so a full failure at that stage represents a bounded, board-acceptable loss rather than a material risk to the business.
Explicit, pre-agreed metrics that trigger a decision to stop rather than continue investing, removing ambiguity about what "not working" actually means.
A clear plan for reverting to the prior process if the pilot underdelivers, so operations are never left in a worse position than before the pilot began.
A measurement methodology, ideally reviewed by finance rather than solely the team proposing the investment, so results are credible rather than self-reported.
A Business Case That Got Approved on the Second Attempt
A mid-size snack manufacturer's operations team first pitched an AI-driven predictive maintenance program to the board with a full-network deployment ask, supported by vendor-provided industry benchmark figures and a technical architecture overview. The board tabled the proposal, citing insufficient evidence the results would hold for their specific plant configuration and no clear answer to what would happen if the projected downtime reduction didn't materialize.
The team returned three months later with a restructured proposal: a 90-day pilot on a single packaging line, a specific target of reducing unplanned downtime by 15% against that line's own trailing twelve-month baseline, a capped investment representing less than 2% of the eventual full-scale ask, and an explicit exit criterion if the target wasn't met by day 90. The board approved the pilot in the same meeting it was presented. The pilot beat its target, and the subsequent expansion request — now backed by the plant's own validated data rather than industry benchmarks — was approved without extended debate.
Answering the Objections Before They're Raised
Experienced board presenters anticipate the specific pushback a capital committee is likely to raise and address it directly in the proposal rather than waiting to be asked. These are the objections that come up most consistently around AI investment specifically.
Address directly by proposing a pilot scoped to your own plant's data and baseline, rather than asking the board to trust an industry benchmark or vendor case study alone.
Address directly with a staged payment structure tied to milestone achievement, so vendor risk is shared rather than fully absorbed upfront by the company.
Address directly by naming the specific operating leader accountable for the outcome, with the metric tied to their existing performance review process.
Address directly with a specific cost-of-inaction figure — the ongoing cost of the current process gap the AI investment is meant to close.
What to Bring Into the Board Meeting
Beyond the slide deck, a well-prepared presenter arrives with specific supporting material ready in case a board member asks for it directly.
A one-page summary of the staged investment amounts, timeline, and specific success metric for each stage.
The current baseline data for the metric the pilot is expected to move, sourced from existing internal reporting.
A named accountable owner and their planned time allocation to the initiative.
A specific, pre-agreed exit criterion defining what "not working" means at each stage.
Two or three comparable industry examples with realistic, sourced outcome ranges rather than best-case vendor figures.
A rollback plan describing exactly how operations reverts to the current process if the pilot is discontinued.
Frequently Asked Questions
How do we choose which use case to lead with when proposing AI investment to the board for the first time?
The strongest first use case is one with a clearly measurable outcome, a short time horizon to results, and low operational disruption if it underperforms — predictive maintenance, quality defect detection, and demand forecasting accuracy improvements are common starting points in FMCG specifically because the baseline metric already exists and is already tracked by the business. Avoid leading with a use case that requires a new metric to be defined from scratch, since the board will reasonably question whether success can even be measured. Our team can help identify the strongest first use case for your specific operating context — book a demo to review your options.
How large should a stage-one pilot investment be relative to the eventual full-scale ask?
Most successful staged proposals size the initial pilot at somewhere between 2% and 8% of the eventual full-scale investment, large enough to produce a statistically credible result but small enough that a full loss at that stage is clearly not material to the business. This ratio varies by company size and risk appetite, but the core principle holds broadly — the number should be small enough that board approval feels low-risk relative to the information gained. For a specific sizing recommendation based on your company's typical capital approval thresholds, contact our support team.
What happens if the pilot succeeds but the board still hesitates to approve full-scale expansion?
This usually signals that the expansion ask wasn't structured with its own staging, or that the pilot's success wasn't independently verified in a way the board fully trusts. Rather than treating stage two as a single large leap from stage one, apply the same staging logic — expand to a meaningfully larger but still bounded scope, with the same rigorous measurement discipline, before asking for full enterprise commitment. Independent verification of the pilot's results, ideally by finance rather than solely the proposing team, also meaningfully increases board confidence at this stage. For guidance on structuring a credible stage-two proposal, schedule a session with our team.
Should the business case include comparisons to what competitors are doing with AI?
Competitive context can be useful supporting evidence, particularly for independent directors evaluating whether the investment reflects industry-standard practice, but it should never be the primary justification for the ask. A board is ultimately evaluating whether the specific investment will produce a specific outcome for their own company, and "competitors are doing this" without a company-specific value case tends to read as pressure rather than analysis. Use competitive benchmarks to add credibility to realistic outcome ranges, not as the central argument. For guidance on incorporating industry benchmark data appropriately, reach out to support.
Who should present the business case to the board — the operations team, IT, or an outside vendor?
The operating leader who will be accountable for the outcome should generally lead the presentation, since the board needs to see clear ownership rather than a proposal that could be perceived as coming from a vendor with an obvious interest in the sale. IT or a vendor partner can support with technical detail if specific questions arise, but the business case itself — the outcome, the staging, the downside protection — should be owned and presented by the person whose performance will actually be measured against it. For guidance on structuring the presentation and anticipating likely board questions, book a demo to prepare together.
Get Your Next AI Proposal Approved the First Time
See how iFactory's value framework structures FMCG AI investment cases around outcomes, staging, and downside protection your board will actually approve.
