Direct the system, not just the tools

The asset manager preparing a pitch document doesn’t need to know how a large language model weights tokens. She needs to know what good looks like, how to describe what she wants, and when the output is ready to send or needs another pass. That gap between “I’ve inherited this system” and “I can direct it” is where most handovers fail. I design mine so management skills for AI sit at the centre, not the periphery.

The observation comes from Ethan Mollick’s recent work on agents and organisational structure. His framing is sharp: the skill that matters in a world of capable AI isn’t technical facility, it’s the ability to manage, direct, judge, and improve what gets produced. In property operations, that maps directly onto the work I do with asset teams, portfolio managers, and family offices who commission brand systems, campaigns, and digital platforms, then need to run them after I leave.

What management skills for AI matter most in property teams?

Clarity of brief, speed of feedback, and editorial judgement. A portfolio director who can write a tight two-line instruction for a hoarding concept will get better output than one who delegates the prompting itself. The ability to say “this tone is wrong for our investors” or “the hierarchy here doesn’t match the asset” is more valuable than knowing which model to use. Management skills for AI are the same ones that make any creative operation work: knowing what you want, recognising it when you see it, and guiding the next iteration.

Why does handover engineering start on day one?

Because capability transfer isn’t a training session at the end, it’s a structural decision at the beginning. When I build a content system or a brand framework for a scheme or a portfolio, I document as I go: file sources, typeface licences, tone rules, image treatment, signage specs, and the logic behind each choice. The custom content management system is built for the next associate to learn, not just for me to operate. Workflow templates, prompt libraries, and approval structures are designed to outlive the engagement. If the handover feels easy, it’s because it was engineered for that from the first week.

How should a small property team start adopting AI in 2025?

By choosing one repeating task where quality matters and speed helps. Not the most complex workflow, the most frequent one. Investor update drafts, pre-let descriptions, planning statement summaries, photography briefs for new lettings. Build a documented process: prompt structure, editorial checklist, approval gate. Run it for a month. Adjust the guardrails. Then expand to the next task. McKinsey’s research on AI adoption across sectors consistently shows that narrow, repeated, managed use cases deliver more value than broad experimentation. Property operations are no different.

What does on-site training for a property team actually cover?

The things people will need to do the Tuesday after I leave. How to update the brand guidelines when a new building joins the portfolio. How to commission photography and apply the visual system to it. How to use the documented workflows for copywriting, signage, hoarding, or digital content without needing to call me. Where the content calendar lives and who owns which part of it. What an investor report should sound like in this portfolio’s voice, and how to check that an AI draft matches it. The session isn’t about the technology, it’s about the operation.

Why do I say AI arms teams rather than replaces them?

Because in every property engagement I’ve run, the constraint is never talent, it’s capacity. The asset manager who needs three investor documents, two hoarding concepts, and a pre-let pack in the same week doesn’t need replacing, she needs a force multiplier. AI gives her that, but only if she can direct it well. The planning consultant preparing for a committee presentation doesn’t outsource his judgement to a model, he uses it to draft, test, and iterate faster. The framing matters because it sets expectation correctly: the team still drives, the tools expand what’s possible within the time available.

What documentation outlives the engagement?

Everything that lets the next person run the system without me. The documentation typically includes:

• Brand guidelines with live examples, not abstract rules
• CMS editing guides with annotated screenshots
• Prompt templates with notes on what to adjust and why
• Photography and drone shot lists with seasonal and lighting notes
• Signage specifications with supplier contacts and material notes
• File structures with naming conventions and archival logic
• Typeface licences and install instructions

The principle is simple: if it’s not written down with enough detail for a capable associate to execute it six months later, it’s not documentation, it’s a dependency.

I work with asset managers, portfolio teams, and family offices who commission fewer, longer engagements because the output needs to work after I leave. If your team is preparing for a letting campaign, a pitch roadshow, a launch, or a project that requires systems you’ll operate yourselves, send a paragraph to hello@goldspink.co. I’ll write back within a day.