The frontier moves.
Value must compound.

01 The through-line

The next technology breakthrough will not arrive on the enterprise’s planning calendar. Neither will the next competitive threat. A CFO who waits for annual budgeting to reconsider the business’s technology choices risks allocating capital against assumptions that have already expired.

Koko’s point of view: the CFO must continuously protect, create and orchestrate enterprise value—connecting capital, data, knowledge, people and delegated action, then redirecting resources as the evidence changes. Better finance operations are part of that mandate. The larger prize is better enterprise decisions: what to sell, what to build, what to buy, where to deploy assets and which risks to carry.

This extends the argument in Enterprise AI 2026. The capabilities around intelligence determine what the business can do with it. The Frontier adds a harder question: which capabilities should the CFO fund when the technology, economics and adoption path keep changing?

KPMG’s Frontiers work and McKinsey’s technology outlook describe developments across computing, physical systems and other domains. They support a broader field of view. They do not establish that every technology will take off together in the next two to three years. KPMG ↗ McKinsey ↗

The acceleration that matters is uneven. Technical capability can advance while integration, economics, infrastructure or customer adoption stalls. That is why this series pairs a technology map with conditional futures and investment gates.

02 Trusted Finance is the foundation of the mandate

Protect. Create. Orchestrate. These responsibilities reinforce one another. Protect the integrity of information, cash, assets and stakeholder confidence. Create better commercial outcomes and strategic options. Orchestrate the capital, knowledge, people and execution that make those outcomes repeatable.

Koko uses Trusted Finance to describe the sustained ability to rely on financial information, exercise accountable judgment, authorize action and verify consequences. It includes reporting integrity, liquidity discipline, appropriate controls, resilience and credible communication. As AI expands, trust must extend through the full decision and execution process.

The strategic argument is that trust can expand the amount of consequential work the enterprise can delegate responsibly. If every answer requires an expert to reconstruct its basis, apparent speed hides a growing review burden. If an action cannot be traced to valid authority, automation scales exposure alongside activity. Trusted data, maintained policy, tested controls and independent challenge make greater autonomy more usable. Hackett’s finance perspective connects existing financial controls with AI-specific controls and continuing oversight; NIST’s framework treats trustworthiness as a lifecycle responsibility. The Hackett Group ↗ NIST ↗

The countercase: demanding certainty or identical controls for every task can destroy value. Set requirements according to consequence, materiality and reversibility. Trusted Finance supports informed risk-taking, with uncertainty visible and accountability retained.

The test for every frontier investment is threefold: what value could it create, what must remain protected, and what needs to be orchestrated for both to hold?

03 What the evidence actually says

The literature already makes a strong case for the CFO as a value creator. KPMG connects intelligent finance with enterprise value. EY exposes the gap between the role CFOs aspire to and the investment decisions they lead. Alvarez & Marsal describes finance as a capability in continuous evolution. Koko should build on those arguments, not claim to have invented them. KPMG ↗ EY ↗ Alvarez & Marsal ↗

Three findings sharpen the problem. EY’s 2026 survey reports that 60% of respondents believe CFOs should shape enterprise value, while 25% lead investment decisions with uncertain, indirect or long-term returns. BearingPoint’s public study summary identifies data, governance and skills as barriers to moving beyond pilots. Gartner’s March 2026 finance-leader survey finds AI investment more often oriented toward productivity than decision quality. These are different instruments and populations, not percentages in one adoption funnel. EY ↗ BearingPoint ↗ Gartner ↗

The implication is practical: a company can become much better at producing financial work without becoming much better at making money, deploying capital or adapting. The CFO has to make that conversion explicit.

04 Six propositions worth arguing about

A faster finance function can still be steering the wrong business

A close completed sooner has value. But it does not answer whether pricing is wrong, inventory is misplaced or the investment portfolio is overexposed to yesterday’s economics. Put decision quality and realized enterprise outcomes beside cycle time. McKinsey’s work on agent-supported FP&A points toward earlier intervention; Koko’s additional requirement is to record what changed because of the intervention. McKinsey ↗

The countercase: reliable reporting and control are prerequisites. Protect that foundation while insisting that the next layer changes a business decision.

The scarce asset may become enterprise judgment that can be reused

More data alone does not explain why a customer received an exception, which policy applied or what an experienced operator knew at the time. Knowledge engineering makes business meaning, policies and decision records usable by people and machines. Foundation Capital’s context-graph thesis and SAP’s architecture direction make this an emerging strategic issue. A decision record is evidence to evaluate; an old exception is not automatically a new rule. Foundation Capital ↗ SAP ↗

The countercase: knowledge that is costly to maintain and rarely reused is an operating burden. Start with decisions where ambiguity creates measurable loss.

The next software negotiation is about who controls the action

Headless software exposes capabilities beyond its own screens. Agents can coordinate those capabilities through an orchestration layer. Salesforce documents headless agent access today. That is evidence of an architectural direction, not proof that all enterprise applications become disposable. The CFO’s questions are about economics and authority: who can act, which system records the transaction, what each outcome costs and whether the enterprise can change providers. Salesforce ↗ Oliver Wyman ↗

The countercase: a well-integrated incumbent suite may be cheaper and safer than a custom layer. Preserve choice without assuming ownership of every component is valuable.

Governance capacity determines how much autonomy the enterprise can absorb

Approving a pilot is an event. Sustaining accountable autonomy is an operating capability. Every new agent, tool permission, model update and counterparty changes the conditions under which an action is acceptable. The Hackett Group addresses continuing controls and segregation of duties in AI-enabled finance; NIST’s agent initiative shows that standards are still developing. The Hackett Group ↗ NIST ↗

The countercase: indiscriminate review destroys the economics of low-risk automation. Match oversight to the consequence and reversibility of the action.

Every productivity dividend needs a destination

Hours saved can fund better customer decisions, faster product launches, stronger controls or lower expense. They can also disappear into unused capacity and additional review. A workforce plan has to name the destination, prepare people for it and verify that the value arrived. Forvis Mazars’ CFO commentary highlights the need for sustained investment in people and integration. Forvis Mazars ↗

The countercase: some work should disappear. Preserve the development of judgment and institutional knowledge when redesigning roles.

The cheapest experiment can create the most expensive dependency

A pilot with attractive marginal model cost may create integration obligations, recurring control work and dependence on a proprietary context layer. The real option is valuable only if the enterprise can stop, switch or expand on acceptable terms. Measure the next commitment and its reversibility, as well as the first invoice. McKinsey ↗

The countercase: excessively small pilots can underfund the changes needed to prove a business model. Fund a coherent value stream in stages, rather than scattering money across disconnected demonstrations.

05 The frontier map

The following horizons are Koko’s planning judgments. 2026–2029 is a window for action and evidence, not a blanket forecast of commercialization. CFO 2030 is a working design for leading organizations. CFO 2035 explores different futures.

Technology anchors: KPMG, McKinsey, WEF, NIST, IEA and BIS. Institutional work supplies useful constraints alongside consultant and investor forecasts. Quantum migration readiness, for example, does not depend on claiming that commercial quantum advantage is imminent. BIS’s Project Agorá work is a prototype, not proof of universal treasury infrastructure. KPMG ↗ McKinsey ↗ World Economic Forum ↗ NIST ↗ International Energy Agency ↗ BIS Innovation Hub ↗

06 The enterprise value cycle

Koko’s proposed operating discipline is a repeating cycle:

Signal → evidence → value hypothesis → scenarios → capital decision → authorized action → verified outcome → reallocation.

The CFO convenes the cycle. Business leaders own commercial and operating results. Technology leaders own delivery and architecture. Data owners maintain meaning and quality. Control owners set and test boundaries. HR leaders build the workforce. Internal audit provides independent assurance rather than operating the controls it later assesses.

The cycle has to cover six kinds of value: growth and customer economics; operating margin; cash and capital productivity; risk and resilience; reusable enterprise capability; and strategic options. Those categories help organize judgment. They are not additive accounting balances, and the same benefit must not appear in several totals.

07 The first 90 days

Choose two enterprise decisions with material value at stake—such as customer profitability and inventory deployment—and one unavoidable foundation or security priority. Give each a business owner, a baseline and a decision date. Map the data, knowledge, skills and permissions needed to act. Fund the smallest coherent version that can generate meaningful evidence.

By day 90, the executive team should have a decision-ready portfolio: what to scale, what to redesign, what to stop and which options to preserve. A larger pilot list is not the intended output. The capital and workforce agenda specifies deliverables, owners, economics and gates.

08 Keep the argument open

This edition brings together 90 source records and a 30-firm comparison. It includes professional services, industry analysts, venture investors, institutions and technical providers. Public summaries, gated material and historical sources are identified in the evidence library.

The point of view should evolve when evidence changes an important decision—not merely when a new report appears. Track each thesis, its current support, the strongest counterargument, the signal that would change the judgment and the capital affected. Revisit operational outcomes monthly, the portfolio quarterly and strategic scenarios after a material technical, economic or regulatory change. These are recommended management practices; an automated research monitor has not been activated.