When AI tools first started showing up in real estate two years ago, every conference had a panel about them. The promises were big. The reality, in most agencies I’ve consulted with, has been more modest.

That’s not a bad thing. The agents getting real value out of AI today aren’t the ones chasing every new tool. They’ve picked two or three things that genuinely save time, and they ignore the rest.

Listing copy generation

Most agents I work with now use ChatGPT or Claude for first drafts of listing descriptions. The trick isn’t writing better prompts — it’s training the model on your own voice with a few examples from your back catalogue.

Stick a handful of your best-performing listings into a prompt with a note like “match the tone of these,” and the output is usable in about half the cases. The other half needs editing, but editing is faster than writing from scratch.

The mistake some agents make is publishing the AI draft straight to REA Group or Domain. Buyers can tell. The descriptions all sound the same — vaguely positive, generic phrases, no actual hook. A good listing has a specific detail that draws the eye, and AI doesn’t know which detail matters in your market.

AI image enhancement

Image enhancement is the quiet winner. The AI tools that brighten interiors, sharpen exteriors, and remove minor distractions are now standard at most agencies. Twilight effects on dusk shots, virtual decluttering, virtual furniture placement — all of it has gotten significantly better in the past 12 months.

The line not to cross is fundamentally altering the property. Removing power lines from a view, painting walls a different colour in photos when the actual walls are stained — buyers see the contrast on inspection day and they don’t forgive it.

CoreLogic and PropTrack have both released analysis tools using image AI for valuation context, which is a different use case but worth knowing about.

Buyer matching

CRMs with AI matching are getting better but they’re still not where the marketing claims suggest. The good ones cross-reference buyer enquiries against new listings and surface matches automatically.

The bad ones flood agents with notifications about every loose match, training agents to ignore the alerts. If your CRM’s matching is generating noise instead of signal, turn it off — most agents I know have done this.

What’s overhyped

A few categories are still selling promises that aren’t delivering:

• AI-driven appraisals for the agent (rather than the AVMs that go to consumers). The accuracy isn’t there for premium properties, and applying AI to a comparative market analysis where you already know the comparables doesn’t add much.
• AI scripts for cold calling. Recipients can hear an AI on the phone within three seconds. The conversion rate is dismal.
• Predictive seller models that claim to know who’s about to list. The data signals are real but the conversion rate of acting on them is statistically indistinguishable from random door-knocking, in my experience.

What I tell new agents

Pick the two or three places where AI saves you 30+ minutes a day. Ignore the rest. The agencies that are getting good at AI aren’t the ones with the most tools — they’re the ones who’ve quietly automated the boring parts and kept the relationship work human.

For agencies looking at deeper integrations across CRM, marketing, and back-office workflows, working with Team400 on a proper rollout makes more sense than buying ten point solutions and stitching them together yourself.

The next 12 months

Voice tools are the next category that will quietly become standard. Transcribing buyer feedback at opens, summarising vendor calls, generating follow-up emails from voice notes. The accuracy is now good enough that the friction is gone.

The bigger shift, though, is going to be in process — not features. Agents who systematise their use of AI across the whole listing-to-settlement pipeline outperform agents who use it ad hoc. The tool isn’t the differentiator anymore. The discipline is.