01 The role changes when the unit of work changes
By 2030, leading finance organizations could be organized around important enterprise decisions, supported by reusable data products and bounded agent execution. The CFO’s mandate remains to protect, create and orchestrate enterprise value. Trusted Finance makes that mandate credible: decisions rest on dependable information, execution has valid authority, and results withstand scrutiny. A smaller reporting team is only one possible consequence.
This is Koko’s conditional working future, not a prediction that every company will reach the same operating model by 2030. It assumes sufficient data reliability, viable economics, accountable ownership and employee adoption. Organizations can progress unevenly across processes.
PwC describes an agentic office of the CFO; BCG emphasizes the CFO’s role in turning AI investment into value; Deloitte connects technology with operating model, people and governance. The direction is well represented. The difficult work is translating it into who decides, who acts and who verifies the result. PwC ↗ BCG ↗ Deloitte ↗
02 Trusted Finance earns the right to delegate
By 2030, leading finance teams could provide more timely insight while preserving the integrity of the ledger, reporting, cash management and control environment. Faster information is useful only when its status is clear: an estimate is distinguishable from an approved forecast, a reconciled balance and a recorded transaction.
Consider a cash-allocation recommendation. The team should know which balances are current, which cash is restricted, which obligations are firm, which assumptions are uncertain and who can authorize the resulting transfer. An agent may assemble alternatives and evidence. Delegation ends where its approved authority ends; the financial system records the completed action.
Trusted Finance therefore needs named owners for financial meaning, reconciliations, access, approvals, exceptions and recovery. Management owns the controls; internal audit provides independent assessment. The operating model must preserve that distinction even when agents perform parts of the work. Hackett’s guidance on segregation of duties and ongoing model and control monitoring reinforces this requirement. The Hackett Group ↗
The move now: trace one material decision from source information through approval and execution to reporting and outcome evidence. Resolve the points where nobody can establish what is authoritative, who is accountable or whether the action worked. Increase autonomy only within the boundary that this evidence supports.
03 Finance becomes a designer of decisions
Consider customer profitability. Today’s improvement project might automate the margin report. The 2030 opportunity is larger: connect actual cost to serve, contractual promises, service events, payment behavior and capacity constraints so commercial leaders can make better pricing and service decisions.
The finance team helps design a reusable decision capability. It specifies what margin means, which costs belong in the decision, how uncertainty is presented and which outcomes would justify a change. A business owner decides what to do. A bounded agent may gather evidence, model alternatives and prepare approved actions. The transaction system retains the authoritative record.
The same customer data product can serve collections, credit review, renewal planning and pricing. Shared definitions reduce repeated reconciliation; context helps explain why a seemingly profitable sale may consume scarce capacity or worsen cash risk. The economic case depends on reuse and better action, not simply having more complete data. Domain ownership and data-product thinking have an established foundation in the data-mesh literature; adopting those principles does not require a wholesale architecture replacement. Zhamak Dehghani ↗
The provocation: the next finance transformation should be able to name the enterprise decisions it will change before it names the applications it will install.
What would weaken the thesis: decision redesign repeatedly costs more than it improves outcomes, or most benefits are achievable through simpler process fixes. Retain that comparator in the investment case.
04 Capital allocation becomes more continuous
Annual commitments still matter for capacity, control and accountability. But technology options may change between budget cycles. The CFO needs a way to redirect the next tranche of capital without renegotiating the entire strategy each month.
Maintain four funding purposes: essential integrity and resilience; reusable capabilities; proven applications; and strategic options. They have different evidence requirements. A mandatory security improvement should not compete on the same short-term ROI score as a customer-pricing product. A quantum experiment should buy relevant learning or access, with a bounded commitment, rather than depend on an unsupported forecast of broad commercial advantage.
Kearney includes capital allocation and scenario planning in its CFO framework. FTI emphasizes aligning inherited investment portfolios to strategy. Koko’s proposal makes those choices repeatable: a quarterly portfolio decision supported by monthly outcome evidence, with authority to stop work and transfer resources. Kearney ↗ FTI Consulting ↗
A CFO should be able to explain: which assumption changed; which commitment remains reversible; which investment now offers the better use of the same scarce people and capital; and what capability survives if the preferred technology changes.
05 Knowledge becomes an operating responsibility
Policies, definitions, contracts and past decisions contain knowledge that may never appear in a clean table. An agent can retrieve a document and still misunderstand which rule applies to this customer, entity, period or transaction.
Knowledge engineering makes those distinctions explicit. It connects business entities, definitions, relationships, policies, provenance and decision records. A context graph is one emerging way to connect decisions to their circumstances; it is not a synonym for all knowledge engineering, and a graph database is not a prerequisite for every use case. SAP’s knowledge-graph direction and Foundation Capital’s decision-context thesis show why this layer deserves attention. SAP ↗ Foundation Capital ↗
The controller may own accounting meaning and policy; operations owns process realities; commercial leaders own approved offers and customer terms; data teams maintain the products through which that knowledge is served. Disagreement is recorded and resolved by the appropriate owner. It should not be silently settled by whichever model generates the most fluent answer.
The move now: select one decision with frequent exceptions. Record the applicable policy, evidence, approver, scope, effective date and observed outcome. Test whether that context improves decisions. Retire stale records and restrict access to sensitive content.
06 The workforce changes through work
Finance does not become valuable by turning every accountant into an engineer. It needs a mix of accounting expertise, commercial judgment, data fluency, product ownership and control design. The Hackett Group’s finance perspective explicitly combines traditional expertise with new AI and organizational capabilities. The Hackett Group ↗
| Role | Valuable work to develop now | Evidence of proficiency |
|---|---|---|
| Controller and accounting teams | Define policy boundaries, evaluate exceptions and maintain authoritative evidence. | Resolve unfamiliar cases accurately and explain the basis to an independent reviewer. |
| FP&A and business partners | Frame choices, challenge causal assumptions and connect operating drivers to economics. | A recommendation changes a decision and its outcome can be evaluated. |
| Treasury | Model liquidity under different futures and test bounded execution with counterparties. | Scenario quality, exposure visibility and tested escalation and recovery. |
| Finance data-product owners | Maintain definitions, access, service expectations and consumer feedback. | Reliable reuse across more than one meaningful workflow. |
| Finance technology and control teams | Evaluate agents, manage changes and test permissions, costs and recovery. | Release evidence, controlled exceptions and successful rollback exercises. |
| Early-career finance professionals | Learn the business through supervised judgment and operational rotations. | Demonstrate why a recommendation is right, when to challenge it and when to escalate. |
Protect the apprenticeship. If automation removes the work through which new employees used to learn, redesign the learning path: reviewed cases, customer and operations exposure, scenario exercises and supervised exceptions. Otherwise the firm can harvest today’s efficiency while weakening tomorrow’s leadership pipeline. This is a Koko design recommendation, not a forecast of a particular employment outcome.
07 Authority becomes explicit and portable
By 2030, a business process may involve an enterprise application, a vendor agent, an independent orchestration layer and several human owners. A conversational interface can obscure that chain. It must not obscure accountability.
Separate the right to read, recommend, prepare, approve and execute. Tie delegated authority to the action, amount, counterparty, timing and risk context. Preserve segregation of duties across the combined workflow, including cases in which several agents share an underlying identity. Changes to models, tools or policies require renewed evaluation when they affect the decision boundary.
MCP and A2A provide mechanisms for connection and communication. They do not determine a company’s financial authority. An agent registry and a governance dashboard help with visibility; enforcement must reach the application and transaction boundary. Model Context Protocol ↗ A2A Protocol ↗ Workday ↗ ServiceNow ↗
The CFO should co-own these boundaries with relevant business, technology and risk executives. Internal audit should retain an independent assessment role. There is no requirement that the CFO operate every technical control personally.
08 The scorecard has to reach enterprise value
The 2030 board scorecard should show outcomes and the capability to sustain them:
| Dimension | What to measure | What not to assume |
|---|---|---|
| Trusted Finance and value protection | Material reporting corrections, reconciliations, liquidity resilience, control exceptions, evidence gaps and stakeholder concerns. | An error-free dashboard proves financial integrity or a hypothetical avoided loss is realized cash. |
| Growth and customer economics | Incremental contribution after delivery, service and acquisition costs. | More AI-generated activity means more profitable demand. |
| Cash and capital productivity | Cash conversion, avoidable inventory, asset utilization and marginal returns on deployed capital. | Faster reporting itself releases cash. |
| Decision quality | Calibration, material errors, accepted interventions and results versus a credible comparator. | A faster or more confident answer is better. |
| Reusable capability | Active consumers, quality, reuse effort and maintenance economics. | Every dataset or ontology is a valuable asset. |
| Workforce | Demonstrated proficiency, redesigned work and realized use of released capacity. | Training attendance or estimated hours equals economic value. |
| Governed execution | Policy breaches, exception severity, evidence completeness and recovery performance. | A policy document ensures compliance in production. |
Gartner’s productivity-versus-decision-quality findings reinforce the need to distinguish those value claims. IBM’s execution research supports attention to operating maturity, while remaining survey-based evidence rather than proof of causation. Gartner ↗ IBM Institute for Business Value ↗
The CFO’s recurring question becomes: what value did we protect and create, how do we know, and what must we orchestrate differently with the next dollar?