Management capability matters more than the AI tool

A family office receives login credentials for three new AI tools in one week. An asset manager’s inbox fills with platform invites. A portfolio team gets procurement approval for software none of them chose. Property teams inherit AI tools daily, often from corporate IT, occasionally from a vendor relationship, rarely because someone in the room asked for them. The question is never whether the tool works. The question is whether the team has the management capability to decide what asset-grade work looks like, which prompts deserve thirty seconds, which outputs go straight in the bin, and when to override the machine entirely.

I train the management layer that makes those calls. Not the tools. The people who decide whether a generative model produces something fit for a planning committee on Thursday morning, or expensive noise that delays the pitch timeline another week.

Why does management capability matter more than the tool?

Because the tool does not know what a leaseholder expects in the corridor. It does not know the planning authority’s recent form. It does not know that the family office prefers understatement, or that this investor wants net lettable area called out on page two, not page nine. Management capability is the editorial layer between the output and the asset. Ethan Mollick at Wharton describes this as management becoming a superpower in the age of agents, because someone still has to decide what good looks like.

What does that look like in a commercial property context?

An associate runs a draft asset profile through a large language model to tighten the narrative before the quarterly report goes out. The portfolio director reads it, spots that the model has flattened every claim into generic optimism, and rewrites three paragraphs. That is management capability. A leasing agent uses AI to generate ten variations of hoarding copy. The asset manager picks the one that reflects tenant mix, location, and the planning condition they agreed eighteen months ago. That is management capability.

Among the Yorkshire developers and surveyors I work with, this gap appears most often when a brochure deadline arrives and nobody has checked whether the model understands RICS leasing standards or local authority presentation norms. The tool produces clean copy. Management decides whether it fits the brief.

What happens when that capability is missing?

Output volume rises. Quality does not. A partnership launch brand arrives in five versions, all coherent, none referencing the brief. Signage goes to print with American spelling because nobody caught it. A drone composite of the scheme looks sharp until someone realises the neighbouring site, currently under separate ownership, has been digitally erased. The tool did what it was told. Management did not intervene.

How do I approach this in training?

I design the handover for management capability from day one. If I build a brand system for a portfolio, I train the team that will inherit it: what the design principles mean in practice, how to commission photography without me, when the templates break and how to fix them, which decisions need a designer and which do not. If I set up a custom content management system, the next associate gets a walkthrough, file sources, fonts, and the content rules that keep the system consistent after I leave. Training and capability transfer sit inside the engagement, not bolted on at the end.

What does AI literacy look like for an asset management team in 2026?

It looks like knowing which tasks compress well and which do not. Summarising a seventy-page environmental impact assessment: compresses. Writing the two-hundred-word executive summary for a risk-averse investor: does not. Generating ten layout options for a sales brochure: compresses. Choosing the one that reflects the asset’s position and the audience’s expectations: does not. McKinsey’s recent work on AI agents in the workplace notes that tasks requiring contextual judgment remain human, even as automation handles pattern-heavy work.

How should a small property team start adopting AI this year?

Start with one repeatable task where quality is easy to judge. Summarising monthly reports. Drafting first-pass leasing descriptions. Reformatting asset data for different audiences. Pick something the team does weekly, where they already know what good looks like. Run the AI output against their existing standard. If it saves twenty minutes and holds quality, keep it. If it introduces errors or requires more editing than writing from scratch, bin it. Build management capability around that one task before adding another.

What documentation should a property team have before the agency leaves?

Everything required to reproduce the work without a phone call. File sources, not just PDFs. Font licences and install instructions. Brand guidelines that explain why, not just what. The content rhythm: what gets updated monthly, what stays static, who approves changes. Login credentials, hosting details, and the support contact for the content management system. A walkthrough video of the one task that always breaks. If the next associate can run the system alone, the handover worked.

Why do I engineer the handover from the start?

Because the asset outlives the engagement. A brand refresh runs six months. The asset trades or leases for years. If the brand system only works while I am in the room, it is not a system. I document as I go: the rationale behind type choices, the photography brief that produced the hero set, the signage specifications that survive a contractor change. My role is to arm the team, not create dependency. Management capability transfers, or the project has failed.

If your team is inheriting tools faster than it is building the capability to use them properly, or if you are planning a branding project and need the handover designed in from the start, send me a paragraph. I will write back.