Waterfall, Agile, or AI-accelerated: 70% of project success is decided in the front end, 20% after go-live, and only 10% in the build — the part that still gets nearly all the attention.
The requirements phase is now arguably longer than the build it precedes. That inversion is new; the consequence of skipping it is not.
“Your build got 20x faster. Your requirements didn't.” Show the phase weighting and let the imbalance make the argument.
How much of your last AI project's rework traced back to a requirement nobody wrote down?
Lens: Field note on delivery and AI executive readiness. The 70/10/20 weighting reflects where outcomes are consistently decided across engagements, not time spent. It is orientation for executive conversations, not a benchmark study.
Where the project is won or lost — before anyone touches a build. AI acceleration has not moved a single item on this list.
Done right, once — instead of discovered in production.
The requirements phase is now arguably longer than the build it precedes. That inversion is new; the consequence of skipping it is not.
“Your build got 20x faster. Your requirements didn't.” Show the phase weighting and let the imbalance make the argument.
How much of your last AI project's rework traced back to a requirement nobody wrote down?
C-suite engagement, and increasingly boardroom visibility, before the first requirement is written.
One thing hasn't gotten easier just because build got faster: you still need real sponsorship to make the front end stick. A name on a charter is not sponsorship.
“The one input AI can't compress.” Position sponsorship as the scarce resource in AI delivery.
Who at the executive table would personally lose something if this project quietly stalled?
Set before pressure sets them for you; assigned before day one, not after the first miss.
AI-compressed builds create the illusion that the whole project compressed. Timelines get set to the demo, then accountability gets assigned to whoever is standing closest when it slips.
“Demo speed is not delivery speed.” Contrast the pilot timeline with the adoption timeline.
Was your timeline derived from the work, or from the date someone already promised?
Invested by the people who will actually use it.
Stakeholder time is the cost line that never appears in the business case and always appears in the post-mortem.
“The unbudgeted line item.” Make stakeholder hours an explicit ask in the charter.
How many hours of your actual end users' time is budgeted into this project?
The part everyone still watches — the status meetings, the headcount, the budget scrutiny.
That's how much the build timeline has compressed.
Build has gone from roughly a third of the effort to something closer to a twentieth — and it still absorbs nearly all the executive attention and anxiety.
“We optimized the 10%.” One chart: the phase that shrank versus the phase that decides.
In your last steering committee, how much of the agenda was build status versus adoption readiness?
Where the deal actually gets closed. Go-live isn't the finish line.
Faster builds produce more code to validate, not less. Compressing QA to protect a go-live date moves the defect, it doesn't remove it.
“Speed moves defects downstream.” Where AI-accelerated delivery quietly relocates risk.
Who has authority to hold a go-live date on quality grounds?
The lesson every Agile team relearns, project after project.
Someone must own adoption, not just deployment. Front-end plus post-build now carry nine-tenths of the outcome.
“Deployed is not adopted.” Tie AI ROI claims to usage evidence rather than release notes.
Ninety days after go-live, what number tells you this was worth doing?
Board members are now being sued personally over AI risk their companies allegedly knew about, and D&O insurers are beginning to discuss coverage carve-outs for AI claims. Seven boardroom topics AI keeps raising, five objections that quietly stall AI initiatives, and what changed this week: board liability just became concrete.
AI Boardroom DiscussionsWhat board members are actually talking about when AI comes up — and how to bridge the conversation to it. Seven business topics boards raise themselves, five objections that quietly stall AI initiatives, and the one-line bridge for each. Drawn from client conversations and current board-governance research.
AI in the NewsPepsiCo reports a 20% throughput lift and a 10-15% capex cut from a system-level digital twin, reached in weeks, while the first wave of twin projects is still being written off as pilot purgatory. The difference is what got modeled, and it is the same failure mode most AI pilots are sitting in right now.
AI in the News, Legal Signal, Security & Compliance, and ROI briefs — written for executives in regulated industries. No spam, unsubscribe anytime.