A practical stack for agents and brokerages: writing tools, design tools, CRM, and automation, plus the mistakes that quietly tank AI-driven marketing.
Last revised May 11, 2026
I have spent the last 18 months helping small real estate teams (one to three agents, plus a virtual assistant) build AI workflows that actually move the needle on listings, lead nurture, and farming. Most "top tools for real estate" lists are written by content teams that have never run a comp or written a follow-up to a lukewarm seller. So here is what I actually use, and what I would tell an agent or wholesaler asking where to start.
The honest hierarchy
Three categories matter, in this order:
- Listing and content tools. Anything that compresses the time between getting a listing and putting strong copy and visuals in front of buyers.
- Lead response tools. Speed of response is the single biggest predictor of whether a lead converts, and the Lead Connect study from the National Association of Realtors has shown this for years.
- Farming and nurture tools. The slow burn that pays off in months six through twelve.
Tools below are organized into these buckets. I have skipped the major CRMs (Follow Up Boss, kvCORE, etc.). Those are infrastructure, not AI tooling, and the choice depends on team size and budget more than features.
Listing and content
ChatGPT, Claude, or Gemini for listing descriptions. All three are competitive in 2026. Claude tends to produce slightly less generic prose; ChatGPT is faster at iterating variants. The difference is small enough that you should use whatever you already pay for.
The prompt structure that consistently beats the default: paste the property details, and paste a recent listing description you wrote that you were proud of. The model uses your description as a voice sample. Generic prompts produce generic copy; voice-anchored prompts produce yours.
Canva's AI image tools (documented here) for thumbnail and flyer work. The AI fill and background removal are good enough for MLS-quality output in a few minutes.
Lead response
Anything that gets you to the lead in under 5 minutes. The 2024 NAR study and Velocify's earlier benchmarks both show the conversion curve drops off a cliff after 5 minutes. The tool matters less than the SLA.
For automated first-touch replies (not follow-ups, just acknowledgment), a simple AI-written reply against a high-quality lead form beats the average human response. The phrasing that works:
Hi [name], saw your inquiry on [property]. Got two quick questions for you: (1) [specific question about their situation], (2) [easy yes/no question]. If you're easier on text, here's my cell: [number].
That structure is short, asks one thing, and gives an out. Most AI-written first replies are 200 words of throat-clearing.
Farming and nurture
This is where AI quietly earns its keep. The math: a sphere of 200 contacts requires roughly 5-7 meaningful touches per year to stay top-of-mind (Tom Ferry's data on agent referral business is consistent on this number). That's 1,000-1,400 individual touches across the year. AI compresses the drafting time on each touch from 5-10 minutes to under 2.
The prompt structure I use:
Write a 4-sentence check-in message to [contact name]. We last spoke about [topic, date]. Their current situation is [what you know]. The market update from this week relevant to them is: [paste a single specific data point]. Voice: [paste 2 sentences of yours]. Banned phrases: "Just touching base," "I hope this finds you well."
The banned phrases are critical. They appear in 80% of every other agent's nurture messages, which is why nobody reads them.
For market-update data points, Redfin Data Center and the Realtor.com Economic Research Hub both publish weekly with citable numbers.
What I would not pay for
The AI features bolted onto most general-purpose CRMs in 2024-2025 are mostly worth what you'd pay if they were free, which they aren't. The "AI assistant" inside [popular real estate CRM brand] is, in my experience, slower and lower-quality than just using Claude or ChatGPT in a separate tab. The integration is the only real value-add, and for solo agents the friction of copy-paste is not worth the premium subscription tier.
What changed in the last 12 months
Two things, both quiet:
- Long-context models can now read a full comp report. Pasting in 20 pages of MLS data and asking for a one-page narrative summary works now in a way it did not work in 2024. Useful for buyer pre-qualification and seller pre-listing.
- Voice-to-text in Apple Intelligence and Google Recorder is finally good. Walking around a property dictating notes, having them transcribed, and asking the model to draft a one-page listing brief is now a 15-minute workflow.
The boring through-line
Tools matter less than the habits you build around them. The agents I see winning with AI right now have built personal prompt templates, a voice sample they paste into everything, and a discipline of iterating ("tighten this," "cut by 30%," "what's missing?") before sending anything.
If you want a starting point, our Real Estate AI Prompt Pack collects the prompt scaffolds I have iterated on most. But the muscle that matters, knowing when a draft is yours, is something you build slowly by editing the model's first attempt every single day.