Asynchronous Intelligence

Imagine asking a capable analyst to review an entire property portfolio, then pulling up a chair beside her. You wait as she opens reports, compares figures, and checks calculations; whenever she finishes one step, you tell her to begin the next.

No manager would deliberately organise work this way, yet conversations with AI can easily follow this pattern.

With a chatbot, you exchange messages with an AI assistant: you ask, it answers, and you guide the next step. You remain responsible for keeping the work moving.

My first article looked at how intelligence is organised, following an analyst from a phone conversation into a team of specialised agents. This second one follows that team through time: what can it accomplish while the person directing it is elsewhere?

Léa leaves the desk

Return to Léa, our portfolio analyst. On Monday, she needs to prepare a review for Thursday's investment committee: what changed across the buildings, what deserves attention, and why?

Her team can gather reports, compare figures, and check the findings. But if Léa must prompt each next step, the whole arrangement still runs on her availability.

She needs an assignment that holds beyond the conversation:

“Review the portfolio for Thursday's committee, identify important changes, and prepare the evidence. Ask me before changing any valuation assumptions.”

The team confirms the scope, the deadline, and the decisions reserved for her. Léa leaves for a tenant negotiation, and the review continues.

Here, asynchronous has a simple meaning: the work can progress without the person who requested it remaining present.

The office closes; the work continues

On Monday evening, Léa goes home while the AI agents continue checking reports, tracing unusual figures to their sources, and revisiting calculations that do not agree.

There are now two clocks: the time the assignment takes, and the time Léa must spend on it.

Work that would stretch across several office days can advance through evenings and weekends. Deeper investigations can run for days while Léa attends meetings and visits properties, without occupying the same stretch of her attention.

On Tuesday morning, two reports with different valuation assumptions need her decision. That part has paused while the other checks continued; once she answers, it resumes.

Human checkpoints depend on the profession and assignment: calculations may proceed autonomously, while changing assumptions or committing expenditure requires approval.

A team that checks its own work

While Léa is away, one agent prepares the analysis, another reviews its calculations and sources, and the coordinator tracks what is complete, needs correcting, or requires a human decision.

Suppose the analyst reports a sharp rise in vacancy. The reviewer finds that the reports cover different periods and sends the comparison back for correction. If the discrepancy remains unresolved, the coordinator brings Léa the question, the evidence, and the work already attempted.

These responsibilities must be built into the process, with access to documents, a shared record of progress, and clear decision rules. Each important finding stays connected to its source, checks, and corrections. Review checks the work as it progresses; this record lets a colleague audit how a conclusion was reached afterwards. Errors remain possible, but become easier to trace.

If a document is unavailable or a step fails overnight, completed work is preserved. The affected part can be retried or paused while independent work continues; repeated failure brings the issue back to the responsible person.

The organisation from the first article now supports work over time through its specialists, reviews, and rules. Léa can follow its progress and redirect or stop it.

One person, several assignments

Once Léa can rely on this arrangement, she can give it more work.

Alongside the portfolio review, another team investigates an acquisition and a third examines lease renewals. Each assignment has an objective and an agent responsible for carrying it forward. Those agents can be human or AI, according to the work.

Léa sets priorities and sees which assignments are progressing, waiting, or need a decision. AI agents can run around the clock, while human colleagues contribute where their expertise or authority is needed.

This is the multiplier: several long processes advance in parallel under one person's direction, through days and nights, with human attention at the points the work requires.

Léa returns for the decision

By Thursday, Léa can bring the committee a portfolio review, an acquisition assessment, and a clearer picture of lease exposure. With the evidence and unresolved questions ready, she can question conclusions, request further checks, and decide what follows.

In the first article, we gave the analyst tools and a team. Now that team can carry several assignments through days of work, with Léa setting their direction and returning at the points her responsibilities require.

She leaves the committee with new questions, gives the teams their next assignments, and carries on with her day.