Why enterprise AI value increasingly depends on redesigning decision rights, organisational boundaries and the way work moves between people and systems.
Most organisations do not have a shortage of decisions.
They have a shortage of decisions being made efficiently.
Information moves between departments. Recommendations await approval. Managers request further analysis. Teams reconcile different interpretations of the same data. Meetings are arranged to resolve matters that should already have an accountable owner.
The cost is rarely visible in a single budget line.
It appears as delayed execution, duplicated effort, additional coordination, slower customer responses and management time consumed by moving work through the organisation.
AI is beginning to accelerate many of the activities within these processes.
The difficulty is that making individual activities faster does not necessarily make the organisation faster.
New research suggests this distinction is becoming central to enterprise AI value realisation.
The productivity gap is becoming an operating-model problem
McKinsey’s October 2026 analysis reports that 80% of respondents have experienced improvements in individual productivity from AI, yet only 37% report positive enterprise-level EBIT impact. Just 6% qualify as AI high performers.
These are survey findings rather than audited financial results, and the figures should not be interpreted as a direct conversion rate from productivity to profit.
Nevertheless, they expose a critical management problem. Employees can become more productive while the organisation’s operating economics remain largely unchanged.
An analyst might complete a report in two hours instead of five. A customer service employee might retrieve information in seconds rather than minutes. A finance team might automate reconciliation activities that previously required substantial manual effort.
Each improvement has operational potential. However, if the resulting work still waits for the same approvals, crosses the same organisational boundaries or encounters the same downstream constraints, much of that potential remains unrealised.
The technology has accelerated the activity. It has not necessarily accelerated the outcome.
The hidden cost of organisational hand-offs
Consider a customer-facing process involving sales, operations, finance and compliance.
A request enters the organisation. Information is collected, validated and passed between functions. Each department performs its assigned activities, with approvals and exceptions handled according to established responsibilities.
AI might substantially reduce the time required for data collection, analysis and document preparation. Yet the request may still move through four departments and multiple approval stages.
The economic benefit depends on what happens across the entire process, not simply within the automated activities.
If a workflow takes ten working days, reducing one activity from four hours to one hour does not necessarily reduce the ten-day elapsed time.
The bottleneck may be waiting for a decision. It may be an unnecessary hand-off. It may be unclear accountability. It may be a policy requiring approval at a level disproportionate to the underlying risk.
This is where enterprise AI investment can become trapped inside an unchanged operating model.
The organisation becomes more efficient at producing work that still waits to be acted upon.
McKinsey’s latest research suggests organisations further along in AI reinvention are more likely to organise cross-functional teams around end-to-end products, customer journeys and business processes. Decisions and accountability are positioned closer to where value is created.
The important insight is not that every organisation should abandon functional structures. It is that enterprise performance increasingly depends on how effectively those structures coordinate work.
Decision rights are becoming part of AI architecture
Traditional technology transformation often concentrates on systems, integration, data and applications. Enterprise AI requires another layer of design.
Who has authority to make a decision? What evidence is required? Which decisions genuinely need human approval? Which can be delegated within defined limits? Who owns the outcome when an AI-enabled process crosses organisational boundaries?
These questions cannot be answered by selecting a more capable model. They concern the organisation’s distribution of authority, responsibility and risk.
Agentic AI makes this particularly important. An agent might be technically capable of completing a transaction, updating a record, requesting information or initiating a workflow. That does not mean it should have unrestricted authority to do so.
The appropriate degree of autonomy depends on business context, financial exposure, regulatory requirements, reversibility and the consequences of an incorrect action.
A low-risk administrative activity might operate autonomously within established controls. A financially material decision might require explicit approval. An exceptional case might need escalation to an accountable specialist.
The objective is not to remove human judgement. It is to ensure human judgement is applied where it creates value, rather than being consumed by unnecessary coordination.
Governance-by-design is therefore not simply about controlling AI. It is about making the organisation’s decision architecture explicit.
The CIO cannot redesign the enterprise alone
Another important signal comes from Thoughtworks’ October 2026 survey of 3,200 CIOs across ten countries.
The research finds that 89% agree they are now more responsible for redesigning workforce workflows and labour models than for managing core IT infrastructure. Meanwhile, 88% report that AI adoption is progressing faster than governance structures can adapt.
These findings reflect executive perceptions, not necessarily formal changes in organisational authority.
That distinction matters. CIOs may increasingly be expected to deliver AI-enabled business transformation without possessing decision rights over the processes, structures and workforce arrangements that determine whether transformation succeeds.
Technology leaders can provide platforms, integration, architecture, security and technical capability. They cannot independently decide how a commercial function should allocate its capacity, how operations should redesign its approvals or how finance should recognise realised benefits.
Those decisions require business ownership. Effective transformation therefore needs a coordinated leadership model in which business executives remain accountable for outcomes, technology leaders enable delivery, and risk, HR, finance and operational stakeholders participate from the beginning.
The transformation cannot be delegated to IT and reviewed after deployment.
Workforce adoption must change the way work happens
AI adoption is often measured through access, licences, usage frequency or employee engagement. Those measures provide useful information. They do not demonstrate that the operating model has changed.
An employee may use AI daily while continuing to work within an unchanged process. A manager may encourage experimentation without adjusting performance expectations, responsibilities or approval arrangements.
A department may report productivity improvements without identifying where the released capacity should be redeployed.
Sustainable adoption requires employees to understand not only how to use the technology, but how their work, judgement and accountability are changing.
That requires practical enablement embedded within the workflow. Managers must be equipped to redistribute work, handle exceptions and measure outcomes. Employees need clear guidance on what can be delegated, what must be reviewed and how to challenge unreliable outputs.
Capability transfer matters because the organisation must continue improving after an implementation team or external adviser leaves.
TFx experience: transformation happens around the technology
The operating-model principles informing TFx Holdings are grounded in enterprise transformation experience where AI capability was integrated with governance, workflow redesign, enablement and measurement.
In an anonymised Fortune 500 environment supporting approximately 15,000 users, adoption increased from 15% to 68%.
Measured programme outcomes included 38% greater efficiency, a 73% reduction in resolution time and a 63% reduction in compliance effort, contributing to $10–12 million in annualised ROI.
These are engagement-specific outcomes, not universal benchmarks or guarantees of comparable results elsewhere.
Their significance lies in the relationship between technology and organisational execution. The programme required more than introducing AI capability. It involved aligning operating practices, embedding adoption, managing governance requirements and connecting improvements to measurable outcomes.
Technology enables a different way of working. Leadership must establish that way of working.
Measure decision performance, not just task performance
Organisations seeking to realise more value from AI should extend their measurement frameworks beyond individual activity improvements.
Task completion time remains useful. However, it should sit alongside measures of end-to-end performance.
These include elapsed process time, decision latency, avoidable hand-offs, rework, exception rates, approval turnaround, cost-to-serve and business outcomes.
The economic implications should be calculated against a defensible baseline.
For example, reducing the time required to approve a customer request may improve throughput or accelerate revenue recognition. That benefit is only financially meaningful when the organisation can establish the relationship between faster decisions and the resulting outcome.
Likewise, reducing management coordination time may create capacity, but that capacity becomes financial value only when it supports measurable additional activity, avoided expenditure or another demonstrable benefit.
The distinction between operational improvement and economic realisation must remain explicit. Otherwise, organisations risk counting the same benefit repeatedly across productivity, capacity and financial return.
The next transformation challenge is organisational
AI is expanding what individuals and teams can accomplish. That capability will continue to develop.
The more difficult challenge is determining how the enterprise should operate when information can be produced faster, routine activities can be automated and increasingly sophisticated systems can act within delegated boundaries.
Existing management structures were not necessarily designed for those conditions. Some approvals will remain essential. Some functional boundaries protect important specialist responsibilities. Human oversight will continue to matter, particularly where decisions carry material financial, legal or operational consequences.
The opportunity is not to remove those protections indiscriminately. It is to distinguish necessary control from inherited friction.
Organisations that do this well can redesign workflows around outcomes, position accountability closer to execution, embed governance into decisions and develop workforce capability that continues to improve over time.
The advantage will not simply come from deploying more AI. It will come from building an enterprise capable of acting intelligently on what AI makes possible.
The executive question
If your organisation could complete twice as much work tomorrow, would its decision-making structure allow twice as much value to reach the customer?
Leave a Reply