From hours of comps to a scored property report
How a boutique buyer's agency swapped hand-built comps, suburb checks and cash-flow sums for a scored research report from a team of AI specialists, with every figure verified by a person.
01 At a glance
The short version
A boutique buyer's agency researches properties for its buyers. We set up a team of five AI specialists that scores each property in one run, so staff check the work instead of building it from scratch.
Boutique buyer's agency
Researches properties for buyers and bills its hours on comps, suburb checks and cash-flow sums.
Hours on every property
About six hours of comps, suburb checks and rental sums per property. Reports read differently depending on who wrote them.
Five AI specialists at once
One command starts a team that checks value, rent, suburb, upside and market, then scores the property out of 100.
Half the research time
Reduced research to 3 hours a property by month 3, down from 6. Staff verify every figure.
The shift
Before: hours of manual research and a report that depends on who wrote it. After: a scored first draft in one run, with staff time spent on checking and advising.
02 The pain
Every property starts
with hours of sums
Buyers pay for judgement. Most of the clock goes on gathering and rebuilding the same kinds of numbers, one property at a time.
Comps built by hand
Staff search recent sales, pick the closest matches and adjust for size and condition. Two people can pick different comps for the same house.
Suburb checks across many sites
Schools, crime, incomes and market trends sit on different sites. Someone opens a dozen tabs and copies the findings across.
Cash-flow sums rebuilt every time
Rent, vacancy, fees, repairs and loan costs go into a fresh spreadsheet for each property. One wrong cell changes the answer.
Reports that depend on who wrote them
One report leads with price and the next leads with rent. Buyers can't compare properties side by side.
Good properties wait in the queue
When research takes hours, the properties at the back of the line get looked at last. The good ones may be gone by then.
03 The solution
An AI research team,
working on every property
The agency types one command, /realestate analyze, with a property address. Five AI specialists research the property at the same time. Their work is then pulled into one scored report.
The tool builds a 14-point property profile and works out the type: a single home, multi-unit, condo, commercial, land or short-term rental.
Five specialists work at once on value, income, neighbourhood, upside and market. Each scores its area out of 100.
The five scores are weighted into one Property Score, graded A+ to F, with a client-ready PDF report.
Human review stayed in
Staff checked each comparable sale against the agency's own data and checked rent figures against local listings. They also reset assumptions such as vacancy and management fees for the local market. Nothing went to a buyer until a person had signed it off.
04 Implementation
What we ran,
and what it gave the team
Nine of the tool's 14 commands cover the agency's daily work. The team starts with a full analysis, then uses the others to dig into one area.
| Command | What it gave the team |
|---|---|
/realestate analyze | A full property analysis from five AI specialists, scored out of 100 |
/realestate quick | A 60-second snapshot to screen a property before spending real time on it |
/realestate comps | A comparable sales analysis and a fair-value estimate |
/realestate rental | Rental income and cash-flow analysis |
/realestate neighborhood | Schools, crime, walkability, demographics and growth |
/realestate invest | An investment view: buy and hold, renovate and refinance, or flip |
/realestate compare | A side-by-side comparison of two properties |
/realestate mortgage | A loan calculator and affordability check from a price |
/realestate report-pdf | A professional PDF property report for the buyer |
Install the tool, point it at the agency's Claude account and run it on one sample address.
Run it on properties the team has already researched. Compare its results with the human work and note the gaps.
Agree a review checklist, show staff the commands and set the local assumptions.
05 Under the hood
The script behind
the grade labels
A short script builds every PDF. This part turns the score into a letter grade and a plain buy signal, with the rule written once for every report.
Score to letter grade. 85 is an A+, 70 an A, 55 a B, 40 a C, 25 a D. Lower is an F.
Score to signal. The same cut-offs give Strong Buy, Buy, Hold / Watch, Caution, Pass or Avoid.
A score of 56 lands on a B and Hold / Watch. Same rule on every report, whoever runs it.
06 The report
A clear score
for every property
Each specialist scores its area out of 100. The five scores are then weighted into one Property Score.
| Specialist | Weight | What it covers |
|---|---|---|
| Value and comps | 25% | Comparable sales, fair value, asking price |
| Income potential | 20% | Likely rent and cash flow after costs |
| Neighbourhood quality | 20% | Schools, safety, walkability, incomes |
| Investment upside | 20% | Growth outlook and the best strategy |
| Market conditions | 15% | Supply, demand and price trends |
What it is worth
Comparable sales, a fair-value estimate and a suggested offer range.
What it earns
Cautious defaults: 8% vacancy, 10% management, 10% maintenance, 5% big repairs, 25% deposit, 30-year fixed loan.
Who lives there
Schools, safety and incomes are scored, so a buyer sees the suburb in numbers.
How to play it
Buy and hold, renovate and refinance, flip or short-term rental.
07 A sample run
A great suburb,
not a great investment
We ran the tool on a sample family home listed at $1.88 million in an upper-end suburb. This is a test run on one property, not a client file.
| Specialist | Score | What it found |
|---|---|---|
| Neighbourhood quality | 85 | Top schools, low crime, high incomes |
| Market conditions | 70 | A solid local market |
| Investment upside | 38 | Only a long-term hold made sense |
| Income potential | 34 | Rent doesn't cover loan and running costs |
Strong on every local measure
Schools score 100 out of 100. Crime is 49% below average. Median income is $191,000.
Asking above fair value
Seven comparable sales put fair value near $1.73 million, about $144,000 under the ask.
Negative every month
About $8,200 a month short, or $98,000 a year.
Renovating loses money
A short-term renovation showed a loss of $197,000. Analysis confirmed only a long-term hold made financial sense.
08 The outcome
Delivered operational results
Measured results across each operational area, with research time falling from 6 hours to 3 hours per property by month 3.
| Area | Before | Delivered at 90 days |
|---|---|---|
| Comparable sales | Picked by hand, and different from person to person | Listed in one run, then checked by staff against local data |
| Suburb checks | A dozen tabs, copied across by hand | One scored neighbourhood section on every report |
| Cash flow | A fresh spreadsheet each time | The same stated assumptions, reset for the local market |
| Report consistency | Depends on who wrote it | Same layout and the same 0 to 100 score for every property |
| Research time | About 6 hours a property | About 3 hours a property by month 3 |
The 90-day rollout
Staff ran the AI workflow alongside manual workflows. Team members compared outputs and calibrated local assumptions.
Specialist reports became the standard first draft. Staff shifted from manual data gathering to review and verification.
Recovered hours were dedicated to buyer consultations and property inspections.
Verified operational return
These numbers reflect verified operational metrics across active buyer engagements. Automating initial comps, suburb profiles and rental cash flows cut report turnaround by half while maintaining strict human verification on every recommendation.
09 The numbers
Delivered results
at 30 properties a month
Based on $80 an hour staff time and $5 AI cost per report. Review efficiency improved each month as staff transitioned from manual gathering to verification.
Staff review replaces manual report drafting.
Half the first-pass hours across 30 properties.
Staff time plus $5 of AI usage per report.
Month by month
| Measure | Before | Month 1 | Month 2 | Month 3 |
|---|---|---|---|---|
| Properties researched | 30 | 30 | 30 | 30 |
| Hours per property | 6.0 | 4.5 | 3.5 | 3.0 |
| Research hours | 180 | 135 | 105 | 90 |
| Staff cost at $80 an hour | $14,400 | $10,800 | $8,400 | $7,200 |
| AI usage at $5 a report | $0 | $150 | $150 | $150 |
| Total cost | $14,400 | $10,950 | $8,550 | $7,350 |
| Cost per property | $480 | $365 | $285 | $245 |
Hours freed over 90 days. That is a saving of about $16,350 after AI running costs, or time for about 35 more properties at the old pace.