Governance Latency: Why Decision Speed Belongs on the Board's KPI List

Key Takeaways

• Governance latency is the elapsed time between a defined trigger point and a recorded decision. It runs on three separate clocks: board, management and operational oversight. Each has a different owner and a different failure mode.
• Latency migrates between the clocks. A board can appear fast because escalation arrived late, which is why the three measures only mean something when reported together.
• Deloitte's 2026 survey of 3,235 leaders found 21% of organisations have a mature governance model for agentic AI, while 74% expect to be using AI agents at least moderately by 2027.
• McKinsey survey work found that organisations making decisions quickly are twice as likely to make high-quality ones, so speed and deliberation length are not straightforwardly traded against each other.
• The link between governance latency and capital outcomes is associative. Timeliness metrics measure throughput, not judgement, so they require paired outcome measures such as decision reversal and rework rates.
• Reversibility is a gradient rather than a binary, and the latency target should vary along it.

Boards in the Energy, Minerals and Resources (EMR) sector measure cost, schedule, production and safety. Decision speed is rarely on the same reporting line. Governance is instead assessed structurally: whether the committee exists, whether the terms of reference are current, whether the papers were circulated in advance.

That framing held while the pace of consequence was set by human activity on site. It is now under pressure from two directions. Capital programmes have grown more concurrent and more interdependent, so a deferred decision propagates further. And operational decision making is beginning to happen at machine speed, in systems that do not wait for the next scheduled cycle.

This post sets out what governance latency is, why it runs on three separate clocks, what the evidence does and does not establish, and how a board might instrument it without rewarding throughput at the expense of judgement.

What Is Governance Latency, and Whose Clock Is It?

Governance latency is the elapsed time between a defined trigger point and the moment the accountable body records a decision. The definition depends entirely on the trigger being defined in advance, which is the part most organisations skip.

Without pre-agreed triggers, latency is measured retrospectively and therefore politically. A board looks fast because management escalated late. A decision looks slow because the board correctly declined to proceed until a piece of assurance arrived. The clock must start at a documented event specified in the delegation of authority: a risk appetite threshold breached, a variance exceeding defined tolerance, a gate dossier assessed as complete against published criteria. It is recorded by the party raising the matter, not the party deciding it.

The second requirement is separating clocks that are routinely conflated:

• Board latency. Strategic, capital, risk appetite and irreversible decisions. Owner: the board. Failure mode: material decisions queued to fit a quarterly calendar.
• Management latency. Gate decisions, escalations, change approvals, contract variations. Owner: the executive and the programme governance chain. Failure mode: decisions circulating between committees without a defined owner.
• Operational oversight latency. The interval between an automated or delegated action and human review. Owner: the accountable line manager, supported by guardrails and intervention thresholds. Failure mode: actions accumulating faster than review capacity.

These are different clocks with different owners, and latency migrates between them. Compressing one while ignoring the others simply relocates the delay. This is one of the distinctions between a PMO that reports and a PMO that governs.

Why Does Machine Speed Change the Calculation?

Gartner forecasts that at least 15% of day-to-day work decisions will be made autonomously through agentic AI by 2028, up from effectively none in 2024. Its Senior Director Analyst Shiva Varma states the consequence directly: when agents operate autonomously, actions are executed at a scale and speed that can outpace human oversight.

The oversight capability has not kept pace with deployment intent. Deloitte's State of AI in the Enterprise 2026 edition, surveying 3,235 business and technology leaders across 24 countries and fielded in late 2025, found that 21% of respondents report a mature governance model for agentic AI, while 74% expect their organisations to be using AI agents at least moderately by 2027. Around 80% lack what Deloitte describes as clear boundaries defining which decisions agents can make independently and which require human approval. In Singapore the maturity figure sits at 14%.

The board is not the control here, and should not attempt to be. No governing body can review an action taken in seconds. The board's role is upstream and periodic: approving the risk appetite and delegation boundaries within which automated decisions operate, setting the maximum tolerable interval between an action and its human review, obtaining assurance that those operational controls function as designed, and deciding whether the AI-enabled operating model remains acceptable.

What machine speed changes is therefore not board cadence but board tolerance. An operational oversight gap that was merely inefficient at human pace becomes a control failure when part of the operation no longer runs at human pace, and the board is accountable for having set the boundary. We have written previously on how AI entering the capital programme shifts the governance question from tooling to accountability.

What Does the Evidence Actually Establish?

The case for measuring decision timeliness predates agentic AI, and it should be stated at the strength the evidence supports rather than above it.

McKinsey research published in 2019, drawing on a survey of more than 1,200 managers, found that 61% of executives say at least half the time they spend making decisions is ineffective, and that fewer than half say decisions are timely. Only 20% of respondents said their organisations excel at decision making. The opportunity cost was estimated at around 530,000 days of managers' time a year at a typical Fortune 500 company, equivalent to some $250 million in wages. On the speed and quality question, the finding is that organisations making decisions quickly are twice as likely to make high-quality decisions compared with slow decision makers. Bain research across more than 1,000 companies found decision effectiveness and financial results correlate at a 95% confidence level or higher, with top-quintile decision makers returning close to six percentage points more to shareholders.

Both findings measure decision effectiveness, of which timeliness is one component alongside quality, alignment and execution. Neither isolates speed as an independent lever. The plausible reading is that the same disciplines producing speed, namely clear decision rights, defined information thresholds and empowered owners, also produce quality. Speed is as likely to be a symptom of good decision architecture as a cause of it.

The capital project evidence should be read with the same restraint. Bent Flyvbjerg's Oxford dataset of over 16,000 projects finds that a one-year extension of the implementation phase correlates with a 4.64 percentage point increase in cost overrun. EY's 2021 study of 192 mining and metals projects valued above US$1bn found 64% ran over budget or schedule, with average cost overrun of 39%. IPA's research base of more than 25,000 projects identifies front-end loading completeness as the strongest single predictor of cost, schedule and operability outcomes, and names governance among the root causes when that definition decays, a pattern examined in the death of FEL by a thousand compromises.

None of that demonstrates that slow board turnaround causes cost overrun. Many delays are symptoms of the same upstream conditions that drive cost growth: weak front-end definition, late scope change, financing constraints, permitting. Shortening a committee's response time would not by itself shorten an implementation phase. What the evidence supports is narrower and still useful: elapsed time is priced into capital outcomes, decision architecture is measurable, and organisations that decide well tend to decide promptly. Governance latency is a legitimate diagnostic and a plausible lever. It is not a proven cause, and the article for it should not be argued as though it were.

Where Does Decision Timing Already Sit in the Management System?

A board does not need a new framework. The clock is already implicit in standards the sector runs against.

• IOGP Report 510. Element 10 expects managers to formally review the effectiveness and fitness-for-purpose of the operating management system. The Risk Management fundamental states that decisions requiring acceptance of higher levels of residual risk are escalated as necessary, including to the company's highest governing body. Element 8 expects intervention to happen in a timely manner when required. Report 510 is guidance rather than a mandate, but each expectation carries a duration that is currently unmeasured.
• ISO/IEC 42001. The first certifiable AI management system standard requires human oversight, defined decision boundaries, escalation mechanisms and management review across the AI lifecycle. Because it shares the harmonised Annex SL structure, it can run inside the management system already in place rather than alongside it, an argument set out in one management system, not two.
• EU AI Act, Article 14. High-risk systems must be designed so that humans can effectively oversee them, including the ability to decide not to use a system or to stop its operation. The timeline moved in 2026: the Digital Omnibus on AI, Regulation (EU) 2026/1744, was published in the Official Journal on 24 July 2026 and entered into force on 27 July 2026, deferring high-risk obligations for stand-alone Annex III systems to 2 December 2027, and to 2 August 2028 for AI embedded in Annex I regulated products. The oversight obligations themselves were not amended. Only the application dates moved, and Article 50 transparency duties applied from 2 August 2026 as originally scheduled.

When Is Slower Governance the Correct Answer?

Deliberation is sometimes a deliberate brake, and a board metric should be built to survive that objection rather than dismiss it. Peer-reviewed work by Shepherd and colleagues in the European Management Review (2021) finds that in resource-scarce environments cautious and analytical approaches may be the most effective, and that acting quickly without deliberation risks wasting scarce resources. Perlow, Okhuysen and Repenning's 19-month ethnography, published in the Academy of Management Journal in 2002, documented a decision-making body that produced worse outcomes because an artificial sense of urgency made speed more important than accuracy.

The answer is calibration, not acceleration. Reversibility in a capital programme is a gradient rather than a binary, and a four-band classification survives contact with real portfolios better than a two-band one:

1. Fully reversible. The cost of being wrong is the cost of changing course. Target short cycle times.
2. Reversible at cost. Recoverable, with a quantifiable penalty. Target moderate cycle times, and record the estimated reversal cost at the point of decision.
3. Staged or contingent. Optionality preserved through sequenced commitment. Measure each stage separately rather than treating the sequence as one decision.
4. Irreversible. Final investment decision, asset disposal, market exit. Apply a protected minimum deliberation window rather than a speed target.

Classification must be set when the decision enters the governance chain, not when it is closed. Otherwise the classification itself becomes the thing that is gamed.

How Should Boards Instrument Governance Latency?

Timeliness measures on their own reward throughput. A body under pressure to clear a backlog can close decisions by narrowing scope, delegating aggressively, accepting thinner assurance, or reclassifying an irreversible decision as reversible. The metric improves while governance degrades. Instrumentation therefore comes in two halves.

Timeliness measures, reported by clock:

1. Time from trigger to recorded decision, reported separately for board, management and operational oversight, and segmented by reversibility band.
2. Gate review turnaround, measured from the point a dossier is assessed complete against published criteria to a documented go or no-go.
3. Decision backlog, being the count and age of open decisions past their needed-by date, where that date is set by the party raising the matter at the point of escalation against the defined trigger, not derived from the committee calendar.

Outcome measures, reported alongside them:

• Decision reversal and rework rate.
• Decisions reopened because information was inadequate at the point of decision.
• Post-decision assurance findings attributable to the decision process rather than to execution.
• Lag from decision to execution, and from execution to observed outcome.

Read together, the two halves answer a question neither answers alone: is this organisation deciding promptly, and are those decisions holding? Selecting metrics that change behaviour rather than populate a pack is the harder half of the exercise, as discussed in what portfolio KPIs actually drive better decisions.

Measuring the Governing Body, Not Only the Project

Assurance in the EMR sector is conventionally pointed downward, at the project, the contractor and the schedule. Governance latency points the instrument the other way, at the speed and reliability of the governance chain itself, including the board that sits at the top of it. That is uncomfortable, which is part of the reason it goes unmeasured.

The case does not rest on AI. Decision cycle time was worth measuring before any system held a delegated decision right, and the strongest version of the argument treats the two strands separately. Capital programme governance has always carried a latency cost. What automation changes is the tolerance for latency in the operational oversight layer, and therefore the precision required of the boundaries a board sets. The questions boards should be asking about their capital projects now reasonably include one directed inward: what starts the clock here, who stops it, and is anyone measuring the interval?

PDAS provides independent assessment of governance and operating management system effectiveness for Energy, Minerals and Resources organisations, including how decision authority, escalation triggers and review cadence perform in practice rather than on paper.

Book a discovery call with our team.