After the AI Deck: A 100-Day Plan to Make AI Change the Number

The Signal

March Approved the Strategy. July Can’t Name the Number.

Picture a familiar, composite pattern. The deck was good. Forty slides, a clear market narrative, a capability map, a three-year ambition and a budget line the board approved without much argument in March. It is now July. The consultants have rolled off. There are four pilots running, two vendor contracts in legal review and a hiring requisition for a head of AI that nobody can quite agree on.

Ask the leadership team a simple question: which number moves, by how much, by when, and whose name is next to it? You get silence, then three different answers.

This is not an unusual story. It is the median story. McKinsey’s data says 81% of organisations see no meaningful bottom-line gains despite near-universal experimentation. Gartner’s forecast says more than 40% of agentic projects will be cancelled by the end of 2027, due to escalating costs, unclear business value and inadequate risk controls. Those two figures describe the same failure from different ends: activity without a named outcome, and spend without a kill switch.

For a media operator, that failure shows up as a churn model nobody in retention actually uses. For a PE owner, it shows up as a value-creation line in the investment memo that quietly disappears from the next board pack. Either way, the strategy was not wrong. The execution simply never had a shape.

The Frame

Strategy Picks the Destination. Sequence Gets You There.

Most AI strategies answer the question “where should we play?” Very few answer “what happens on day one, day 30 and day 100?” That second question is where value is created or lost, and private equity learned this long before generative AI arrived.

Bain’s private equity value-creation practice treats the structured 100-day plan after a deal closes as decisive, and reports that funds with active post-acquisition value-creation support achieve a 2.9x average MOIC and 30% higher returns than the industry average. AlixPartners’ work on the first 100 days makes the same point about any strategic decision: sort actions into no-regret quick wins, shorter-term value moves and longer-term structural investments, and stand up a governance office, or execution risk rises sharply.

An AI board approval is a strategic decision. It deserves the same discipline as a close.

The rest of this piece rests on two plain rules. First, the order matters: diagnose, then govern, then decide (build, buy or partner), then integrate, then measure and decide to kill or scale. Each step produces an artefact the next step depends on. Second, the sourcing decision sits in the middle, not at the start. Choose the tool before you have named the workflow, and every later step bends around the tool.

That second rule is the counterintuitive one. The most expensive mistake in the first 100 days is not a bad model or a slow vendor. It is committing to a platform, or to an AI hiring plan, before anyone has named the one number, the one workflow and the one accountable owner.

The Evidence

Four Places AI Programmes Leak Value

The research is consistent about where the money goes missing. It is rarely the technology.

The gap is organisational before it is technical

McKinsey’s research finds 86% of leaders feel their organisations are not prepared to adopt AI in day-to-day operations, and one in six organisations has no clear C-level owner for AI adoption at all. McKinsey’s prescription is a “double transformation,” technical and organisational together, rather than piecemeal adoption. Read that plainly: a tool with no owner is an expense.

Workflow redesign beats tooling

Of 25 adoption practices McKinsey tested, the redesign of workflows had the biggest effect on EBIT impact. It is also the hardest practice to act on, because it requires changing how people work rather than what software they log into.

Governance falls off a cliff at go-live

The Economist Enterprise and Databricks “Making AI Deliver” survey of 1,221 executives at large enterprises found about three in five review AI systems during development and before deployment, but fewer than two in five keep that oversight going after launch. One in eight waits until something breaks. And about three in five firms take seven to 12 months to ship a project, which means most AI programmes never reach a measured result inside a single budget year.

The vendor market is noisier than it looks

Gartner estimates that only about 130 of the thousands of vendors marketing agentic AI have genuine agentic capability, a pattern it calls “agent-washing.” Meanwhile the buyers are changing. The DPP’s research across 28 major European media organisations found more than 80% say AI-assisted coding tools have significantly or moderately increased their capacity to write code, and 11% now report a strategic shift to build, where a year earlier none did. Cheaper code tempts teams to build. It does not make building wise.

In Depth

The 100-Day Sequence, Gate by Gate

Here is the sequence, gate by gate. The day markers are gates, not calendar suggestions. You do not pass a gate without its artefact.

The Landscape

Build, Buy or Partner: What Each Path Really Costs

The sourcing debate in media usually collapses into a false binary. There are three paths. Most operators should land on the third.

The DPP data shows the build instinct growing as AI coding tools make development cheaper. Dalet’s counterargument, drawn from the media asset management era, is sharper: don’t rebuild the foundation unless that foundation is truly your differentiator. Buy the core platform. Customise the last mile: business rules, workflows, brand-specific experience. Foundation Capital, writing from the investor and vendor side of the market, observes why pure buying tends to disappoint: as AI capabilities commoditise, proofs of concept demand data ingestion, orchestration and workflow-specific customisation that traditional SaaS pilots never needed. It calls this a “cost of sale crisis.” That implementation effort has to be absorbed by someone: vendor, specialist or buyer.

The pattern is clear. For most media firms, buy the technology and integrate it with a specialist, with knowledge transfer and exit terms written into the contract. Build only where proprietary data or process creates a moat a competitor genuinely cannot buy.

The Implication

Two Readers, One Test

Different chairs. Same test: can someone name the number?

What to Watch

Six Signals That Will Decide the Next Two Quarters

Watch these six signals through the first half of 2027, in your own business and across the market.

The Operator’s View

The Deck Was Never the Hard Part

AI does not fail in the model. It fails in the gap between an approved slide and a named owner with a frozen baseline.

I have sat on both sides of this table, in media and telecom operating roles and alongside PE owners, and the pattern repeats. Programmes that change the number are rarely the most ambitious. They are the ones that picked one workflow, put human approval and measurement in place before scaling, bought the foundation, integrated the last mile properly, and then had the nerve to kill what did not work. Closing that gap between strategy and governed, measured execution is the work we spend our time on at nGülam.

So run the question against your own plan. If your board asked tomorrow which single number your AI programme will move by day 100, would your leadership team give one answer or three?