Governing the Redeployment Dividend: Turning Saved Hours Into Value

Teams that deploy AI save the equivalent of five hours per person per week, according to research by Gartner, and most of that time is then spent on non-value-added tasks. The same research shows that productivity gains from AI are no higher than those from earlier technologies, and that only around one in three teams report high gains at all. The surprise is not that AI saves time. The surprise is how efficiently organisations waste the time it saves.

When saved time is reabsorbed by low-value work, the saving never reaches the results the Board sees. Preventing that requires clarity about who is accountable for what: AI owns the execution of the work, managers own the workflow it sits within, and the Board owns whether the freed capacity creates value. In January I argued that AI’s real prize is the Redeployment Dividend, releasing intellectual capital from undifferentiated work rather than cutting headcount. The evidence now shows what happens when that dividend is left ungoverned. Capturing it is a governance discipline, and it begins with redesigning the operating model around the capacity AI frees.

What the evidence actually shows

The primary study is Gartner’s survey of 724 respondents across business functions, conducted between June and August 2024 and reported by Gartner in March 2025. It found that around one in three AI-enabled teams reported high productivity gains: 37% of teams using traditional AI and 34% of teams using generative AI, a pattern Gartner reports as the ‘AI productivity paradox’, with gains visible at the level of individual teams but far harder to capture across the organisation as a whole. The paradox has not aged. In April 2026, research by PwC covering 1,217 senior executives found that around three quarters of AI’s economic gains are being captured by just 20% of companies, and that the companies pulling ahead are twice as likely to redesign workflows around AI rather than simply add AI tools.

The minority is where the business case lives. Among the high-productivity teams, 81% reported enterprise cost savings that were, on average, 27% higher than those achieved by lower-productivity teams, in the same Gartner analysis. Note the comparison carefully: this is a gap between high-productivity and lower-productivity teams, not an absolute saving that AI delivers by default. The same group reported outcomes beyond cost, with 71% creating more novel products and offerings and 68% reporting quality improvements. Where the recovered hours are put to work, they show up in results the Board actually cares about.

The same failure appears wherever anyone measures it. In HR, a July 2025 Gartner HR survey found that 55% of HR leaders want a freed hour redirected to higher-value special projects, while a parallel survey of 1,973 managers found that only 28% would prioritise that redirection. The gap between intent and action could hardly be more explicit. The aggregate consequence is now visible at the level of the firm: a 2026 study published by the National Bureau of Economic Research, fielded with research teams at the Federal Reserve Bank of Atlanta, the Bank of England, and the Deutsche Bundesbank, surveyed nearly 6,000 senior executives across the United States, the United Kingdom, Germany, and Australia. It found that while 69% of firms actively use AI, roughly nine in ten executives report no measurable impact of AI on productivity at their own firm over the past three years. Time is being saved almost everywhere; value is being recorded almost nowhere.

Individual teams, whole functions, entire economies, and the same pattern at every altitude. Capacity is freed, and then it is reabsorbed unless something deliberate intervenes. The saving is measured, celebrated, and lost. The timing matters because the scale is changing: as AI moves from isolated pilots into enterprise-wide deployment, the freed hours stop being a rounding error and start being a material share of the organisation’s capacity. Most of the Boards I meet can put a number on the hours their AI tools recover; far fewer can say where those hours went. The measurement problem is largely solved. Governing what happens to the measured time is the question the measurement has exposed, and it is the one most organisations have not yet answered. Recovered capacity is not redeployed capacity.

This maps directly to the AI Stages of Adoption. Organisations that can evidence time saved but not value created are stuck between Adopting and Optimising, having mistaken the deployment of tools for the redesign of the operating model. The tools are working. The organisation around them has not changed.

Why the dividend leaks away

The default behaviour of freed capacity is to fill itself with the nearest available work, and the nearest available work is usually the low-value work that was already there. An hour recovered from drafting, reconciliation, or reporting does not sit waiting for strategic direction. It absorbs into inbox management, meetings, and the residue of the old process. Some of that residue is work that should not survive at all: as I argued in Not Everything Needs AI, the question that comes before any AI decision is whether the work deserves to exist, and freed capacity has a habit of giving undeserving work a second life. Absent a deliberate redirection, saved time becomes slack rather than redeployed capacity, and slack is invisible on the results the Board reviews.

The headcount-reduction framing makes this worse rather than better. When AI is justified primarily as a route to a smaller workforce, the organisation optimises for cost removal and never builds the mechanism for redirecting capacity toward differentiating work. There is nowhere for the freed hours to go because nobody designed a destination for them. It is also a framing the evidence does not support: Gartner’s own analysis, as distinct from its survey findings, attributes under 1% of the layoffs announced in the first half of 2025 to AI productivity gains, and forecasts that AI will produce a net increase in jobs from 2028. On the evidence so far, AI is not shrinking the workforce; it is changing the work, and the forecast points away from shrinkage rather than towards it. The argument I made in the Redeployment Dividend applies directly here: optimising for a smaller workforce forfeits the larger prize of a more valuable one.

Beneath both problems sits an accountability gap. AI owns the execution of the work, and that transfer is well under way. Managers own the workflow it sits within, and they decide what fills the working week. But the third element is missing: in most organisations, nobody owns whether the freed capacity creates value. Function heads measure adoption rates and hours saved because those are the numbers their tools report. The destination of those hours belongs to no one, which means the decision about what fills them is made by default, one diary at a time, by people whose incentives reward keeping the existing workflow running. The 55% of HR leaders who want redirection against the 28% of managers who would prioritise it is not an anomaly; it is the predictable result of intent that has no owner at the level where work is allocated.

The evidence on workflow redesign points the same way. As the research cited in the Redeployment Dividend shows, productivity improvements are substantially larger when AI adoption is accompanied by genuine redesign of the surrounding workflow rather than layered onto existing processes. Time saving without operating-model change produces exactly the leak the data describes. A redeployment dividend is not a by-product of adoption. It is the output of a designed system, and the evidence suggests few organisations have built one.

The Board’s redeployment discipline

What follows is not a list of five recommendations. It is a single governance discipline, an instance of Minimum Lovable Governance, whose one purpose is to ensure that every hour AI recovers is deliberately redirected toward strategic value. It is light enough to enforce and coherent enough to own, and each element depends on the others.

A governance system produces what it measures, so the discipline starts with the success metric. A Board that asks how many roles AI can remove will get cost removal and a leaking dividend. The better question, and the one I would put to management, is where freed capacity is being redirected and what value that redirection has produced. This is the Six Board Concerns applied in practice, Strategic Alignment and Financial and Operational Impact above all, since a time saving with no strategic destination fails the first test and no evidenced return fails the second.

The metric only functions if redeployment is owned rather than assumed. Someone must be answerable for whether recovered capacity creates value, and that ownership must sit at the level where work is actually designed and allocated. Without it, the intent-to-action gap between leaders and managers is not a risk but the default outcome, as the HR evidence demonstrates.

Ownership in turn only matters if there is something to own, and this is the largest claim in the discipline: freed capacity delivers value only when the work is redesigned to receive it. Approving an AI tool is not the same as redesigning the work around that tool. A tool approval changes what one task costs; a redesign changes what the role, the team, and the process exist to produce once that task no longer consumes them.

Redesign answers the questions an approval never asks. What does this role do with the five hours it gets back each week? Which decisions move faster because of them? What does the team now take on that it previously could not? Redesign is also where the accountability split stops being a diagram and becomes an operating model: AI owns execution, managers own the workflow, and the Board owns whether the freed capacity creates value, and the redesigned work is the place where those three ownerships meet.

A business case that claims a time saving without specifying the operating-model change that converts the saving into value is an incomplete business case, and I would treat it as one. This is a design responsibility the Board owns, not an operational afterthought it reviews.

With the design in place, the Board can measure what matters. Time saved is an input. The evidence to require is downstream: the cost, innovation, and quality outcomes that the high-productivity teams in Gartner’s study reported, produced by the redirected hours. Hours recovered are the beginning of the account, never the result.

Finally, the discipline extends to what the organisation deliberately stops doing. As I argued in the Redeployment Dividend, some capabilities should be allowed to fade as AI absorbs them and some must be preserved; the point here is only that this atrophy should be a governed choice rather than an accident of adoption.

Practised together, these elements are what moving from Adopting to Optimising looks like in reality. Not more AI, but an operating model redesigned around what AI has already made possible.

The dividend goes to the governed

An organisation that measures only the hours AI saves will mistake a leaking dividend for a captured one. The measure that matters is what the recovered capacity produced, and of the three ownerships, that one sits with the Board and nowhere else.

AI is creating capacity almost everywhere; the advantage will belong to the organisations whose Boards decide what that capacity becomes. That decision, more than any technology decision, is what will separate the organisations that scale AI from those that merely adopt it.