Enterprise Future Guide: Capabilities, leadership and value in 2030 and 2035

Koko: Welcome to the Enterprise Future Guide. I'm Koko. Joining me are Sam, our constructive skeptic, and Max, who tests whether any of this actually works at operating scale. Here's where we start: by 2030, the capability may become the real unit of enterprise investment, not the function.

Sam: I'm Sam. That's a hypothesis worth pressure-testing. A capability map can just recreate a matrix with new labels.

Max: Max here. My concern is coordination cost. Who actually owns the budget and the outcome when sales, ops and finance each have a stake?

Koko: That's exactly the operating contract the capability needs: a named outcome owner, an economic baseline and explicit delegated authority. BCG describes customer-journey teams combining people, agents and data, though those are ambitions, not confirmed returns.

Sam: Bain's reorganization research is the counterweight. AI-focused reorgs reported weaker performance. Changing structure before changing the actual work tends to add overhead.

Max: So the COO and CFO need to test the work first. Pick one customer promise, map the handoffs, set the baseline. If the pilot doesn't improve outcomes net of all coordination costs, you stop.

Koko: Right. And professional accountability stays distinct. Finance maintains economic integrity. Legal and risk keep independent challenge. The capability owner coordinates commitments, not acquires every professional responsibility.

Sam: What changes the hypothesis? If a well-run functional model with better integration outperforms the capability design after honest accounting, narrow the change and keep looking.

Koko: My hypothesis here is that the C-suite has to become a system of accountable partnerships, not a row of adjacent functions. When an agent touches credit, inventory and legal exposure in one transaction, no single title contains that decision.

Sam: But shared ownership is where accountability goes to die, Koko. A CAIO box on the org chart does not settle who funds the fix when something goes wrong.

Max: One possible failure mode is a new title without clear authority. The CFO is asked to sign a business case, while nobody has named who can actually stop the agent mid-run.

Koko: Which is why the hypothesis is a charter, not a title. Russell Reynolds' guidance makes the point that the design should match the transformation need. Spencer Stuart's executive alignment work shows distinct CDO and CIO accountabilities sitting side by side. One named outcome owner, with escalation rights written down.

Sam: The IIA's Three Lines Model is worth saying plainly here. Internal audit must stay outside the delivery line entirely. The moment assurance becomes a delivery responsibility, you have lost the check.

Max: And Forrester's CDO abstract separates data meaning and lineage from the CIO's execution environment. Those accountabilities can conflict. The charter has to establish how disagreements are resolved before the agent is live.

Koko: The test I'd give this is simple. Ask two executives to describe the same exception. If they name different owners, the charter is not working. That is also the decision gate before any expansion of agent authority.

Sam: So the CEO sponsors the charter, the CFO tests the capital case, and audit stays independent. That means the next question is what fills that charter: the knowledge the decision actually runs on.

Koko: So here is the third hypothesis: maintained knowledge becomes shared operating infrastructure. The claim is that reusable definitions, decision context and policy may matter more than the agents themselves. Andreessen Horowitz frames it as agent performance depending on maintained business context, though that is an investor thesis, not a proven return.

Sam: And Foundation Capital adds that exception records and approval history could become valuable systems of record. That raises a question about what the enterprise chooses to preserve.

Sam: The concern is that capturing decisions can also preserve mistakes. Context graphs are compelling on paper, but maintenance costs accumulate fast. HFS noted that context needs an accountable owner, though their survey had only fourteen responses, so treat that carefully.

Max: Operationally the question is who actually pays for upkeep. If the CDO owns quality and the CIO owns the infrastructure, but no business unit owns the outcome, the knowledge product quietly drifts.

Koko: That is exactly Forrester's distinction, even at abstract level: the CDO holds semantics and lineage, the CIO holds the environment. They are separate accountabilities, and conflating them is where drift starts.

Sam: The alternative scenario worth testing: incumbent systems already hold the data and permitted actions. Andreessen Horowitz makes that counter-case themselves. A well-governed ERP with clean definitions may outperform an expensive context layer.

Max: So the signal that changes the view is practical: fewer recurring exceptions across more than one workflow consumer. If the same interpretation fight recurs monthly, the knowledge product is not working.

Koko: The action today is narrow: CDO, CIO and one business owner fund a single maintained knowledge product. CFO and procurement agree on evidence export and supplier exit before the contract is signed. Ninety days, then test it against a hard exception case, not clean inputs.

Koko: So the hypothesis here is that autonomy is earned, not granted. My claim is that agent governance has to persist through deployment and change—identity, scoped authority, evidence, recovery and a named human owner. A policy document at launch is not enough.

Sam: The structure sounds right, Koko. My concern is that an inventory documents risk without controlling it. NIST's identity guidance and the WEF-Capgemini authorization profile are useful starting points, but neither makes a specific deployment safe by itself.

Max: And operationally, the cost of building that control layer is real. Spend limits, monitoring, rollback—those need ownership and budget before the agent touches a consequential action.

Koko: Agreed. The IIA's Three Lines model is clear that independent audit must test the evidence without operating the control. And on regulation, the EU AI Act framework is live but applicability needs jurisdiction-specific legal review, not a generic deadline.

Sam: What's the alternative scenario? If controls are too layered, you slow legitimate work. Automated controls can fail too—Gartner's abstract flags that accountability between technology and business leaders has to be explicit, not assumed from a shared dashboard.

Max: The decision gate I'd use: two executives should be able to name the same owner, explain the same exception and confirm who stops or reverses the action. If they can't, the delegation contract isn't real yet.

Koko: That test connects 2030 to 2035 directly. In the Compounding Enterprise scenario, authority scales because the evidence trail holds. In Stranded Intelligence, agents proliferate but the controls drift and the review burden absorbs the gains.

Sam: So the action is concrete: capability owner, CISO, CIO, legal and risk define one delegation contract now—spending limits, permitted actions, exception ownership, rollback. Independent audit reviews it. That's the next step before we talk about scaling.

Koko: My hypothesis here is that value creation and judgment development have to travel together. Agents can absorb more routine work, but if that work is what builds expert intuition, we may hollow out the bench we need by 2030.

Sam: The evidence is mixed though. METR's 2025 developer study found slower completion with early AI tools, but sixteen people on familiar repositories is too small to generalize. Their 2026 update flags further design problems. It doesn't settle what happens at enterprise scale.

Max: Operationally the pressure is to show savings now. Freed capacity looks good until the CFO asks whether it's cash or a projection.

Koko: That's the core problem, Max. Accenture's Pulse of Change found a persistent gap between AI investment ambition and sustained value, with leaders and employees describing it differently. Hours saved isn't value until you subtract review costs, rework and absorbed risk.

Sam: The alternative scenario is that AI-supported learning accelerates judgment if it's deliberately designed. Loss isn't inevitable; it's a design failure. What signal would change your view?

Koko: Demonstrated proficiency alongside automation metrics. If junior staff handle exceptions and succession strength holds, I'd revise upward. If the senior expert becomes the only person who can recover a failed workflow, that's Stranded intelligence in the talent layer.

Max: So people leadership, operating managers and the CFO redesign supervised practice together, measure accepted outcomes, and only move capital after net benefits and control costs are visible. Ninety-day gate: does the next cohort handle exceptions without a rescue call?

Koko: Let's land this. The four scenarios—Compounding enterprise, Stranded intelligence, Disciplined optionality, Fragmented progress—give you a stress-test, not a forecast. 2035 should change a decision you make this quarter, not sit in a strategy deck.

Sam: And resist picking a favourite scenario to invest against. McKinsey's Technology Trends Outlook is explicit: technical, commercial and organisational readiness are different things. One improving doesn't pull the others along.

Max: The question to ask is: which scenario does the current capital budget actually assume? If a team has not written that down, it cannot justify a pause trigger, let alone an exit.

Koko: It extends to physical AI and quantum too. The IEA shows efficiency per task can coexist with higher total consumption—enough to flip the investment case. And NIST is clear that post-quantum cryptography is a classical-system decision the CISO owns now, separate from whatever the CTO is exploring on quantum applications.

Sam: Two decisions, different owners, different timelines. Treating them as one quantum programme is how budget gets stranded.

Max: The concrete action: take one material capability investment, run it through all four conditions in a single session, record which scenario the current spend assumes, and set a numeric scale, pause or exit trigger before the next quarterly review.

Koko: CEO, CFO and COO in the room together. That's the whole series in one move: value, judgment and accountability travelling together. The CFO Future Guide is the companion read on capital and Trusted Finance.

Sam: That's Koko Knows: Enterprise Future Guide. Set the trigger. Don't wait for the destination.