A BRIEFING FOR PROPERTY FUNDS

What separates the winners from the strugglers

Why most property funds are getting nothing back from their AI licences, and what the ones getting results are doing differently.

REGAN KIRK · IMPACT FIRST AI

Book 30 minutes INCLUDES A LIVE DEMO WATCH THE AGENTS AT WORK
01IS THIS YOU?

You bought Copilot licences for the whole company. Twelve months on, a few people use it to tidy up emails, someone tried it on a board paper once, and nobody can point to a single hour it has given back.

CHALLENGE 1Your Copilot licences aren’t delivering an ROI

Any of this sound familiar?

CHALLENGE 2Your systems don’t talk to each other

Every report, reconciliation and investor pack depends on someone moving numbers between Excel, MRI, SharePoint, Forbury and NetSuite. They check the data, resolve mismatches and repeat the process next month. Right now, your highest-paid people are acting like the glue between systems.

WHAT NEEDS TO CHANGE

These are two different problems requiring two different responses: help people get more from the AI tools they already have, and redesign the manual work that sits between your systems.

02KEY TRENDS

Three AI trends property funds should pay attention to

TREND 01 / 03

Today’s AI is far more capable than most people realise.

Most teams only use what sits above the waterline.Most teams use only a fraction of what their licence can already do.

HOW MOST TEAMS USE IT Tidying emails Summarising documents Writing first drafts WHAT AI CAN ALREADY DO Automated research agents Inbox triage Compliance monitoring Deal screening + 40 MORE WORKFLOWS
HOW MOST TEAMS USE AI 3 routine tasks
ROUTINE TASKSFULL CAPABILITY
UNTAPPED CAPABILITY
40+higher-value workflows
  • Automated research agents
  • Inbox triage
  • Compliance monitoring
  • Deal screening
ILLUSTRATIVE
TREND 02 / 03

A new AI model is released every 10 hours.

New capabilities arrive every week, and the pace is accelerating, not slowing.

Source: Alex Atallah, co-founder of OpenRouter

MODEL RELEASES 0 IN 18 MONTHS
ILLUSTRATIVE
TREND 03 / 03

The models improve every month. Your output doesn’t, unless you do something.

AI CAPABILITY YOUR OUTPUT THE GAP TODAY +24 MONTHS
03WHAT’S POSSIBLE

Five ways property teams can use AI today

Most use cases fall into one of these five categories, from finding an answer to completing an entire piece of work.

01

Answer questions

Find answers using general knowledge or information from your business.

02

Research and analyse

Read, compare and draw insights from research, transactions and reports.

03

Create documents

Turn source material into documents that follow your templates and standards.

04

Run processes

Carry out repeatable steps across systems, the same way every time.

05

Complete work

Take a defined task from the brief to a review-ready result.

Seven demos of real use cases

Animated walkthroughs of work we have built or shown to property funds. Each tab plays one.

TAP ONE TO WATCH IT RUN

RESEARCH AND ANALYSE · BUILD 01

The asset dashboard

  1. 1Give the AI the asset folder: leases, tenancy schedule, rent roll, valuationCOMPLETE
  2. 2It reads every document and pulls out the numbersCOMPLETE
  3. 3It cross-checks the signed leases against the schedule and flags every varianceCOMPLETE
  4. 4It builds the dashboard: key metrics, lease expiry profile, scenario modellerCOMPLETE
  5. 5Refine it in plain language: “model the anchor tenant leaving at expiry”HUMAN IN THE LOOPCOMPLETE
310 EDWARD STREET · ASSET DASHBOARDBUILT FROM 14 DOCUMENTS
Fourteen documents read. Three variances found.
4.2 yrsWALE
94%Occupancy
38%Expiring in 24 mo
3Variances
LEASE EXPIRY PROFILE · INCOME BY YEAR OF EXPIRY HOLDING FY27 FY28 FY29 FY30 FY31 · NONE FY32+
VARIANCE Level 9 rent does not match the signed deed of variation

The schedule still shows the old rent. The dashboard flags the difference, cites both documents, and shows the income at stake until the record is fixed.

SCENARIO What happens if the anchor tenant leaves at expiry?

Downtime, re-letting rent, incentive and cap rate are sliders. The income and value impact moves as you drag them.

Illustrative dashboard built from one folder of documents. Figures are examples.

A folder of documents becomes a working prototype in an afternoon: the metrics, the expiry profile, every difference between the signed leases and the schedule, and a scenario tool for the conversation you are dreading. It proves the idea fast. The version your team relies on every month runs on connected data, so it stays current without anyone re-uploading a folder.

Open this exact dashboard and click around. It’s real, and the figures are illustrative.
KEY LEARNING

None of these seven started with the question “where can we use AI”. Each one started with a piece of work that was expensive, and worked backwards. Before we get to how you build that habit into your company, try one for yourself.

Watch the agents do this work, one short demo each.
▶TRY IT · A LIVE SCRIPTED BUILD

Ask the research agent yourself

This is the research library build you just saw, wired up so you can press the buttons. The two questions are preset, and the answers are reproduced from a real run on our own research library. Ask in plain language and get a sourced answer in seconds. On your library, it runs on your research.

RESEARCH AGENT · SCRIPTED DEMO

Pick a question to run the agent.

Preset questions. Answers reproduced from a real run on our research library, not live data.
Get this tool at my company
04WINNING MENTALITY

What the winners do differently: two tracks in parallel

They build AI fluency across the company while redesigning the workflows that cost them most. One helps people get more from off-the-shelf tools. The other puts AI to work behind the scenes.

TRACK 1

Build AI fluency

Help people use off-the-shelf AI tools in their everyday work.

TRACK 2

Redesign costly workflows

Embed AI into the process to remove manual steps and speed up decisions.

Most companies blur these together. Buying licences does not redesign a workflow, and building automation does not make people fluent. Track one improves how people work with AI. Track two changes how the work itself gets done. You need both.

Still judge AI by the first version they tried

Regularly test what today’s tools can do on real work

Treat AI as a one-off tool rollout

Build fluency and redesign the costliest workflows

Expect AI to compensate for fragmented, unreliable data

Create trusted, accessible sources for critical data

Assume AI is the answer to everything

Audit the problem, then choose the best solution

Run ad hoc pilots with no clear owner

Follow a structured process, assign an owner, validate quickly

05TRACK ONE

Track one: build AI fluency across the fund

AI fluency is knowing when to use AI, how to get a strong result and how to check the output. A licence provides access. These five elements build the capability, and three practices do most of the work.

THE ELEMENTS OF AI FLUENCY
1Give everyone access
2Train by role
3Preload the knowledge
4Reward the early users
5Support the first 90 days
THE 80:20 OF USING AI TOOLS
5.1

Create a knowledge base for each project or role

Add the documents the work relies on, then ask AI questions in plain language. For asset management, that might include tenancy schedules, leases and reporting templates. For finance, it might include the chart of accounts, reporting calendar and company definitions.

KNOWLEDGE BASE · IMPACT FIRST
PROJECTS
310 Edward Street
Head of Legal
Quarterly Report
Market Research
What is our WALE at 310 Edward Street?
ANSWERED FROM THIS PROJECT From the tenancy schedule and lease abstracts in this folder: the WALE is 4.2 years across the office and retail tenancies, with the two anchor leases expiring inside the term.
Tenancy scheduleLease abstractsValuation summary
ILLUSTRATIVE
5.2

Turn repeated documents into reusable AI skills

Give AI the source documents, company template and rules to follow. Save the process once, then use it to create landlord disclosure statements, monthly asset reports, distribution statements or investment committee papers.

REPEAT DOCUMENTS, BUILT ONCE
PRODUCED OVER AND OVER
Landlord disclosure statement
Monthly asset report
Distribution statement
Investment committee paper
BUILT ONCE AS A SKILL
Disclosure statement skill
Asset report skill
Distribution statement skill
IC paper skill
Taught once: your process, your template, your standards. Then it runs the same way every time.
5.3

Teach people a simple way to prompt

Two people can use the same AI tool and get completely different results. A clear prompting method turns generic answers into useful work with fewer rewrites. Teach people to provide context, define the task, show what good looks like and refine the first response.

ONE TASK, TWO WAYS
THE LAZY VERSION

Dump every file and the whole task into one prompt, ask for the finished report, get something plausible back, spend an hour checking every line, lose trust in the whole exercise.

THE BEST VERSION
  1. Break the job into small individual tasks
  2. Validate the output at each step before moving on
  3. Tell the AI what your takeaway is. It cannot know what you think matters
  4. Have it build the final piece from the validated work plus your judgement
Same task, same model. The prompting is the difference, and technique can be taught.
06TRACK TWO

Track two: redesign your costliest workflows

Start by finding where the company loses the most time and money. Look for manual handoffs, repeated checks, rework, delays and decisions slowed by scattered information.

Then redesign the workflow. AI may be part of the solution, often working quietly in the background, but it is not the starting point.

Start with the problem. The most expensive workflows in your company are the same ones they were five years ago. Those do not go out of date. What can be done about them changes every few months, and that is our job to keep up with, not yours.

The Impact First Method

Our method quickly identifies the opportunities worth solving, validates the ROI before anything is built, and turns the strongest cases into workflows that run differently.

  1. 01

    Target

    Find the workflows that cost the most. Not the ones that are most annoying, the ones that consume the most time, money and senior attention. At property funds it’s usually month-end, investor reporting or capex approvals. Your numbers confirm which.

    Output: a longlist of candidate workflows

  2. 02

    Prioritise

    Score each workflow on impact and on how fixable it looks. Impact is what it costs today in time, money and senior attention. Fixable means visible waste, a fix proven elsewhere, a clear right answer, what it needs within reach, and an owner who wants it solved. Pick the two that score highest on both and refuse to start on the rest.

    Output: a ranked shortlist, and an agreed first two

  3. 03

    Map

    Document how the work actually runs today, including the manual steps nobody has written down and everybody performs. This is where the real cost becomes visible.

    Output: the workflow as it truly is, not as the process document claims

  4. 04

    Ideate

    Design the options. Sometimes the answer is AI. Sometimes it is a data connection, a form, or deleting a step nobody needed. If another fund has already solved this workflow, we start from that fix. This is the stage where knowing what is currently possible matters most, and it is the stage companies cannot easily do alone.

    Output: two or three options with a recommended one

  5. 05

    Validate

    Test the approach on real work before committing to it. Does it hold up on your documents, your data, your edge cases?

    Output: evidence it works, or evidence it does not

  6. 06
    THE GATE

    The business case

    Validation tells you it can be built. The business case tells you whether it should be. What does it cost, what does it return, who owns it, and what happens if you do nothing. If the numbers do not work, the right answer is to stop here, and stopping here has cost you a fraction of what a failed build costs.

  7. 07

    Build and deploy

    Build it, put it in front of the people who will use it, and stay with it until it is genuinely part of how the work runs. A deployed tool nobody adopts is a failed project with better paperwork.

    Output: a workflow that runs differently from before

A GOOD FIRST TARGET

Why finance is often where to start

Finance work repeats every month, follows clear rules, and still runs on a lot of manual effort: coding invoices, matching bank lines, chasing accruals, rewriting the same commentary. Work like that is easy to check, so a fix proves itself quickly. That’s why the first result often shows up in finance.

MONTH-END CLOSE · BANK RECONCILIATION · INVOICE CODING · VARIANCE COMMENTARY · INVESTOR REPORTING

WHAT WE TYPICALLY SEE AT PROPERTY FUNDS

People are acting like the glue between systems

Tap the one that eats most of your team’s week. The calculator below turns it into a number, with your inputs.

What is the glue work costing you?

6
8
$120
ANNUAL COST OF THE GLUE WORK
$264,960

That is roughly 1.3 full-time people spent connecting systems by hand.

Your inputs, your number. We assert nothing here.

Today’s AI is powerful, but it can’t fix bad data. It inherits it.

Your people are the layer between systems that do not talk to each other. Every answer gets rebuilt by hand because there is no common layer underneath. Point AI at that and you get fast, confident, wrong answers. There is a better way: connect the systems once, then reuse the foundation. Every dashboard, report and AI tool after it starts from numbers you trust.

Multiple versions of the truth, high manual effort, reconciliation headaches, re-coding and recalculating data.

One connected layer underneath, so every answer starts from numbers you trust.

What connected systems make possible

  • Live, trusted data dashboards
  • Consolidated financials
  • Automated monthly and investor packs
  • Ask questions of your data in plain language
  • Custom applications with AI built in
  • End-to-end workflows with AI running in the background
HOW WE HELP

We find the workflows costing you the most, redesign them, and build the systems needed to make the new process work.
07THE OFFER

How we can help

Three ways to start. Pick the one that matches your biggest problem.

Choose an option above.

See what it includes and how we deliver it.

QUESTIONS PROPERTY FUND LEADERS ASK

Straight answers

Why is our Copilot or ChatGPT rollout not delivering anything?

A licence gives people access, not capability. Most property funds skip two things: teaching people to use the tool on their own documents, and connecting the data underneath so the tool has reliable numbers to work with. Without both, usage stays at tidying up emails and nobody can point to an hour it has given back.

What is AI fluency?

AI fluency is knowing when to use AI, how to get a strong result, and how to check the output. It is built through practice on real work, not a one-off seminar. Three practices build it fastest: a knowledge base for each project or role, repeated documents turned into reusable AI skills, and a simple way to prompt that people remember.

What is glue work in a property fund?

Glue work is the time your highest-paid people spend carrying numbers by hand between Excel, MRI, SharePoint, Forbury and NetSuite for reports, reconciliations and investor packs, then checking and re-checking them. Estimate it with people doing it, hours each per week, 46 working weeks and a loaded hourly rate. Ten people at five hours a week is roughly 1.4 full-time staff.

Should we fix our data before we use AI?

Do both at once. AI fluency can start on Monday with the licences you already own. But AI cannot fix bad data, it inherits it, so any workflow build starts by connecting the data underneath. Winners run the two tracks in parallel: fluency across the team, and workflow redesign starting with the costliest workflow.

What can property funds actually use AI for today?

Five kinds of work: answering questions, research and analysis, creating documents, running processes, and completing whole pieces of work. Real builds include an asset dashboard on connected data, a landlord disclosure statement that went from 60 minutes to 4, a capex request that clears in days not weeks, a month-end close where only the exceptions reach a person, a deal-screening analyst, a research library you can question in plain language, and a DA watcher.

Is it safe to use AI on our financial data?

It can be, when it runs on tools your firm has approved. The business versions of Copilot, ChatGPT and Claude don’t train on your data by default. Anything we build lives in your own Microsoft environment, on your accounts, and we explain where the data goes before we build anything.

What does it cost to start?

Team training on Copilot, ChatGPT or Claude is from $5,995 for up to 12 people, with 30 days of support and a guarantee. Connecting your data, and running it for you if you want, is scoped on a short call. Building one application in a fixed-price Sprint is scoped on a short call. All prices exclude GST.

The next step is a conversation

If any of this feels familiar, let’s spend 30 minutes identifying where your company is losing the most time and whether fixing it would deliver a worthwhile return.

No pitch. Just an honest assessment of the opportunity and whether we can help.

Email Regan