The New Operating Model for Portfolio and Asset Teams

Agent Systems Need Managers, Not Just Prompts

Much of the public conversation around AI still revolves around prompts. That focus makes sense for experimentation, but it is the wrong abstraction for scale. MIT Sloan Management Review's recent discussion of managing 'a superhuman workforce' makes the deeper point: once organizations deploy agentic systems, the central challenge becomes management, not engineering.

The prompt-engineering narrative has dominated AI adoption for two years. It suggests that the key skill is knowing how to talk to the machine. For individual tasks, that is partially true. For organizational deployment, it is fundamentally insufficient.

MIT SMR's work on managing agentic systems argues that once AI agents operate with autonomy, pursuing multi step goals, making intermediate decisions, and interacting with data and other systems, the management challenge becomes central. Agents need objectives, scopes, decision rights, escalation paths, quality standards, and failure responses.

That looks much more like organizational design than prompt engineering. And it requires leaders who think in systems, not in chat windows.

Prompts define how a human interacts with a single AI instance. Management defines how an organization deploys, governs, and evolves AI across workflows. These are different problems operating at different scales.

A well-prompted agent that operates without clear boundaries, escalation logic, or accountability is still a risk. A moderately prompted agent with strong governance, clear scope, and embedded review points is far more likely to create durable value.

The firms that focus primarily on prompt quality will find that their AI projects produce impressive demos but struggle to become operational. The firms that focus on management quality will scale.

The New Operating Model for Investment Teams: Humans, Agents, and Structured Workflow Memory

Most firms still operate on fragmented workflow memory. Facts in one system, assumptions in another, conversations in inboxes, reasoning in presentations. McKinsey's work on the 'agentic organization' argues that durable value comes not only from agents themselves, but from the operating model, governance, and data foundation around them.

Every CRE firm has an operating model, even if it was never deliberately designed. It is the sum of how information flows, who does what, where decisions get made, and how institutional knowledge gets preserved (or lost).

In most firms, this operating model is fragmented by default. Deal data lives in spreadsheets. Communication lives in email. Analysis lives in presentations. Decisions live in meeting minutes that nobody retrieves. And institutional memory, the accumulated knowledge of what worked, what failed, and why, is stored in people's heads.

McKinsey's work on agentic organizations makes the argument that durable competitive advantage in the AI era will come from firms that build deliberate operating models where humans, agents, and structured data work together, rather than alongside each other by accident.

Fragmented memory means that every new workflow starts from partial context. A new deal requires reassembling information that already existed but was not stored in a reusable form. A new asset review requires rebuilding analyses that were done before but were not captured structurally.

This creates an invisible tax on every decision. Not a financial tax, a quality tax. Decisions are made with less context than the firm actually possesses, because the context is scattered, outdated, or inaccessible.

Agent systems amplify this problem. An agent operating on fragmented data produces fragmented output. An agent operating on structured, reusable workflow memory can reference prior patterns, compare live cases against historical ones, and maintain traceability across the entire lifecycle.

How Agentic AI Will Change Asset Management

Asset management is often imagined as a reporting-heavy discipline. In practice, it is an intervention discipline. McKinsey's recent work on agentic AI in operations argues that these systems are especially valuable where judgment, process sequencing, and action all need to happen together, which describes asset management precisely.

The core job of asset management is not to produce numbers. It is to decide what must happen next: what needs attention, what is drifting, what is underperforming, and who should act. That makes it one of the disciplines most likely to be transformed by agentic AI, not because the reporting gets faster, but because the entire operating logic can shift.

McKinsey's work on agentic AI in operations makes the case that these systems create the most value where work requires judgment, sequencing, and coordinated action across multiple steps. Asset management fits that description almost exactly.

For asset managers, the shift means moving away from manually assembling operating views and toward supervising continuous detection, escalation, and intervention.

Most asset managers spend a disproportionate share of their time on activities that do not require their expertise: pulling data from different systems, reconciling numbers, formatting reports, and chasing updates from property managers or leasing teams.

The actual analytical work, identifying rent leakage, spotting lease events before they become problems, evaluating capex priorities, or deciding when to intervene in an underperforming asset, often gets compressed into whatever time is left after the data preparation is done.

This is not a technology gap. It is an operating model problem. The data exists, but it is fragmented across systems, formats, and communication channels. The asset manager becomes a manual integration layer, and that is precisely the kind of work that agentic systems can transform.

What the Job of an Investment Manager Looks Like in the Age of AI

The classic investment manager job has been built around gathering, synthesizing, and translating information. In an agentic environment, much of that translation layer becomes automatable, which means the role itself starts to move up the value chain. McKinsey's work on agentic workforces and MIT SMR's research on choice architecture both point to the same conclusion: the human contribution shifts toward framing, judgment, and outcome ownership.

Emails become spreadsheets. PDFs become summaries. Rent rolls become assumptions. Internal discussions become memos. The investment manager's traditional job is, at its core, a translation function: converting unstructured information into structured decision context, then shepherding that context through the firm's internal process.

McKinsey's recent work on the future of agentic work suggests that human roles will increasingly shift toward orchestration, judgment, and outcome ownership rather than repetitive knowledge processing. In investment management, that shift is already beginning.

That does not make the investment manager less important. It makes the role more exposed. When the machine can produce summaries, options, and first pass recommendations, the human contribution becomes easier to see, and easier to evaluate.

The traditional investment manager spends a large share of time on intake, synthesis, and coordination. Gathering documents, normalizing data, preparing analyses, formatting outputs, coordinating feedback, and pushing materials through internal approval chains. These tasks are important, but they are increasingly agent ready.

The new job moves upstream. The investment manager of the near future will need to be excellent at framing questions, setting constraints, identifying what matters in an edge case, and knowing when a seemingly clean output should not be trusted.

MIT Sloan Management Review's work on intelligent choice architectures supports this shift: the quality of a decision depends heavily on the quality of the options presented. Investment managers who can design better question frames and evaluation structures will produce better outcomes, not because they work harder, but because they work on the right layer.

This is not a marginal skill adjustment. It is a role redesign.

Portfolio Teams in the Age of AI: From Reporting Owners to Decision Architects

Many portfolio teams still spend a disproportionate amount of time packaging information rather than shaping decisions. MIT Sloan Management Review's recent argument is instructive: better decisions often begin with better choices, and better choices require environments designed to surface relevant options and trade offs.

Portfolio management in most CRE firms is still organized around reporting cycles. Teams spend weeks assembling data, reconciling numbers across assets, formatting board materials, and producing recurring packages that satisfy compliance requirements but rarely change strategic direction.

The irony is that the people best positioned to shape portfolio strategy, those who see across assets, markets, and cycles, are often the most consumed by the mechanics of information assembly.

MIT Sloan Management Review's research on choice architectures offers a useful reframe: the quality of decisions depends heavily on the quality of available options. If portfolio teams can shift from assembling reports to designing better choices, the impact on decision quality could be significant.

A report describes what has happened. A choice architecture shapes what happens next. The distinction matters because most portfolio teams are structured around the former while the organization needs the latter.

Shaping choices means defining which comparisons matter, which thresholds deserve attention, which anomalies are noise, and which should trigger management discussion. It means designing portfolio views that make trade offs visible, not just presenting data in a standard format.

When AI can increasingly generate recurring views, summaries, and alerts, the human role shifts toward asking the right portfolio level questions. Which assets require intervention? Where is conviction weakening? Which exposures deserve a different strategy?