Artificial intelligence was supposed to transform business. According to Torsten Slok, chief economist at Apollo Global Management, it hasn’t, at least not on the balance sheet. Outside the big tech companies, he says, AI has so far delivered “zero financial payoff” (TechCrunch).
That’s a striking claim, and for many large firms a fair one. But small businesses tell a different story, especially those whose days are filled with the same customer questions. Estate agents, letting agents and property managers fall squarely in that group.
Below we look at what Apollo found, a real small-business case study with full numbers, and how the same thinking applies to the property sector.
What Apollo found
Slok studied the “S&P 493”: America’s 500 largest listed companies minus the seven tech giants. For these firms, profit margins and revenue per employee have stayed flat, with no clear sign that AI is lifting productivity.
His explanation is that business runs on judgment, trust and accountability, which AI doesn’t understand. There’s also a practical reason: most large companies adopted AI broadly, spreading it across thousands of staff without pointing it at a specific job. The gains are real but too thin to show up in the accounts.
Why small businesses are different
In a company of 50,000 people, a productivity gain disappears into an average. In a business of five, it’s one named person getting hours of their day back.
That makes AI both more rewarding and more risky for small firms. A large company can absorb a failed experiment. A small business that spends £500 on the wrong tool has lost real money. So the right question isn’t “should we use AI?” but “will automating this one job save more than it costs, and how fast?”
The case study: one WhatsApp AI agent, full numbers
Qaushik Labs, a small AI agency, recently built a WhatsApp AI agent for a small UK business. Customers contacted the business mainly by phone and WhatsApp, asking the same questions every day: prices, availability, opening times, order status. A receptionist answered each one by hand, which took 2 to 3 hours a day.
The agent was trained on the business’s own FAQs and policies. It answers routine questions instantly and passes anything complex, sensitive or money-related to a person, with the chat history attached. It now handles about 80% of queries, and routine reply time fell to 15–20 minutes a day.
| Monthly figure | Before | After |
|---|---|---|
| Staff time on routine replies | 55 hours | 6.4 hours |
| Staff cost (£15.35/hr fully loaded) | £844 | £98 |
| AI running cost | £0 | ~£30 ($40) |
| Total | £844 | £128 |
| Net saving | £716 a month |
With a one-off build cost of £500, the project paid for itself in about 15 working days. First-year net benefit is roughly £8,090, a return of about 940% on total first-year cost. Even in the most cautious case (2 hours down to 20 minutes a day) the saving is about £533 a month.
The saving is staff time freed, not wages cut. The team spent those hours selling and serving customers, which is where the profit gain came from.
Why this applies to property businesses
The business in the case study wasn’t an estate agency, but its problem will sound very familiar to anyone in property. Agents and property managers field a constant stream of near-identical questions:
- Is this property still available?
- What’s the rent or asking price, and what’s included?
- Can I book a viewing on Saturday?
- Are pets allowed? Is there parking?
- What documents do I need to apply?
- My boiler isn’t working. Who do I call?
Most of these can be answered from information the agency already has: listings, viewing slots, tenancy policies and maintenance contacts. That’s exactly the kind of work an AI agent handles well. Negotiations, offers, disputes and anything involving money or legal advice should stay with a person.
A quick test for your agency
Before spending anything, run the numbers the way the case study did:
- Pick one task, such as answering viewing and availability enquiries.
- Time it for a week. How many staff hours go on it each day?
- Cost it: hours × working days × fully loaded hourly cost (wage, employer NI and pension).
- Price the AI: one-off build plus monthly running cost.
- Check payback: build cost ÷ monthly saving. Under three months is a strong yes. Over six months, start with something simpler, like saved WhatsApp quick replies or a clear FAQ page.
As a rough illustration, not a real client: a letting agency spending 2 hours a day on routine enquiries at the same £15.35 hourly cost is spending about £675 a month on them. Cutting that to 20 minutes a day would save around £530 a month, a similar result to the conservative case above. Your own numbers will differ, which is why measuring first matters.
Why it worked when so many AI projects don’t
- One task, not “AI transformation”. A single, measurable bottleneck.
- Numbers before code. The baseline justified the build in advance.
- The channel customers already use. WhatsApp meant nothing new to learn.
- The business’s own knowledge. Answers come from real policies, not generic internet text.
- People keep the judgment calls. Complaints, payments and exceptions go to a human, covering the trust and accountability Slok says AI lacks.
The bottom line
Apollo’s warning is fair: AI spread thinly with no clear target rarely pays. But a small business, or an estate agency, that picks one repetitive job and runs the numbers first can see a return in weeks.
Want to know if AI would pay off for your business? Qaushik Labs runs the same cost-benefit check and will tell you honestly if it isn’t worth building yet. Visit qaushiklabs.com.
Assumptions: £13/hour receptionist pay based on 2026 UK job ads; 15% employer NI and 3% pension (≈ £15.35/hour fully loaded); 22 working days a month; $40 converted at about £0.75 per $1; WhatsApp Business messaging fees not included. The letting-agency example is illustrative only. News source: TechCrunch, 25 October 2026.