September 25, 2026 · 8 min read
Your Clients Want AI Help and a Human Agent. New American Express Research Shows How to Give Them Both
American Express asked 528 Card Members how they feel about AI in customer service. They want it for the routine, they want a person for the moments that matter, and they want accuracy over speed. Here is what that means for a brokerage's marketing.
On September 15, 2026, American Express published research on how its Card Members feel about AI in customer service. The short version: they want AI to make human support better, not to take its place.
Those Card Members are also your buyers, sellers, past clients and recruits. They let an app pick the restaurant, book appointments without a phone call, and fire quick questions at a chatbot. When something important is on the line, they still want to reach a person who knows what they are doing.
Buying or selling a home is about as important as it gets. That is why this research deserves a closer look from anyone who runs marketing for a brokerage or a team. It shows, in numbers, where people welcome automation and where they expect a human. Here is what American Express found, what it means for real estate marketing, and how to use AI without losing the relationships the business runs on.
What the American Express research found
In February 2026, American Express surveyed 528 U.S. Consumer Card Members from its Customer Research Community about the role of AI in customer service. Four findings stand out.
People want a choice. 95% said it is important to be able to choose between an AI-powered voicebot and a human representative.
Accuracy matters more than speed. Asked what makes an AI voicebot experience positive and memorable, 67% named issues resolved accurately. Only 35% named issues resolved quickly. That is nearly two to one in favor of getting it right.
People accept AI acting on their behalf when they confirm the action. 71% were comfortable with an AI assistant scheduling meetings and appointments for them with their confirmation, and 65% said the same for booking a hotel.
People welcome useful personalization. 79% were somewhat or very comfortable with generative AI analyzing their data to recommend a restaurant, and 73% said the same for hotel recommendations.
There is a fifth finding that matters most for our industry. Comfort dropped in more sensitive situations, such as managing personal finances or getting budgeting recommendations. Mohammed Badi, President of Global Servicing at American Express, put the company's position in five words: "accuracy is what earns trust."
Comfort drops as the stakes rise
Read the numbers as a scale and a pattern appears. People are more comfortable with AI recommending a restaurant than a hotel. They are more comfortable with AI scheduling an appointment than booking a room. They are more comfortable with AI suggesting something than doing it for them. And when the subject turns to their money, comfort falls again.
The more money and consequence involved, and the more the technology acts on its own, the more people want control and a person within reach.
Now put a home on that scale. For most households it is the largest purchase they will ever make, wrapped in contracts, deadlines, inspections, financing and a lot of emotion. If a hotel booking makes people a little cautious about automation, a purchase agreement sits in a different category entirely.
Your clients will gladly let technology handle the small stuff. For the big decisions, they are hiring your agents. Marketing that understands that split wins on both ends.
Let AI take the routine and keep the path to a person obvious
A lot of real estate marketing is repetitive: confirming showing times, sending open house follow-ups, delivering new listing alerts, answering basic questions about a property, reminding clients about upcoming dates. AI and automation are well suited to all of it.
The 95% finding is the guardrail. Every automated touchpoint should make it easy to reach a human. Put the agent's name and direct number in automated emails. If the brokerage website uses a chat assistant, label it as an assistant and give visitors a one-tap way to call, text or book time with an agent. Automation that gets people to a person faster builds trust. Automation that traps them in a loop spends it.
Build in a confirmation step, too. Comfort with AI scheduling in the survey depended on one condition: the customer confirms. Let an assistant propose three showing times and have the buyer pick one. Let an email suggest consultation slots and have the seller choose. People accept automation when they keep the final say.
Here is what that looks like in practice. A buyer browses a listing at 10 PM. The site assistant answers a question about HOA dues from the listing data and offers three showing times. The buyer picks Saturday at 11:00 AM. The next morning the agent calls to confirm and asks the questions a chatbot cannot: what is drawing them to this home, and what would make them walk away?
Put accuracy ahead of speed in everything you publish
AI tools write fast, and they can state things with total confidence that are not true. In real estate the details carry the weight: square footage, lot size, HOA dues, tax figures, days on market, the open house date. A listing description with the wrong fee, or a market update built on an invented statistic, costs credibility with the exact people deciding to trust you with a sale.
Use AI for the first draft. Then check every number against the MLS, public records or another reliable source, and read the whole piece to make sure it sounds like the agent. The Card Members in this survey named accuracy nearly twice as often as speed, and nobody feels differently about the biggest transaction of their life.
This is the standard we hold our own team to. Every fact in a post, a caption or a monthly newsletter is copied from the source and checked before it goes to the client for approval. Fast is nice. Right is required.
Personalize with what clients tell you
Most respondents were comfortable with AI tailoring recommendations to them, which tells you personalization is welcome when it is useful. In real estate, useful personalization looks like:
Listing alerts matched to a buyer's stated price range and neighborhoods.
Market updates for the specific area a homeowner lives in.
Separate email tracks for active buyers, past clients and homeowners thinking about selling.
Recruiting messages that speak to where an agent is in their career, based on what they told you.
Keep personalization grounded in what people share with you: budget, preferred areas, property type, timeline. Never build it on assumptions about who someone is. That keeps the marketing relevant, respectful and on the right side of fair housing rules.
Keep the agent's voice on the moments that build trust
AI can outline a market update video, draft a caption or suggest subject lines. What it cannot do is be the agent. Clients choose a person, and that choice rests on judgment, local knowledge and how the agent makes them feel when things get stressful.
Spend the time AI saves where a person matters most: the on-camera market update, the story behind a new listing, the personal note after closing, the phone call the day the inspection report arrives. Short videos where an agent explains the market in their own words let people decide to trust them before they ever meet. That is why video editing sits inside every plan we run: the footage is the agent, and the edit just keeps up.
For a team leader, this is also the recruiting story. Candidates can tell the difference between a feed of generated filler and a brand with real people in it. The second one is the one they want to join.
A note on fair housing and AI-written copy
Our founder teaches lunch and learn sessions in brokerage offices on real estate advertising and fair housing, and AI does not change the basics. Fair housing rules apply to everything a brokerage publishes, no matter who or what wrote the first draft. Listing descriptions, social posts, emails, website pages and chat responses all count.
AI tools learn from enormous amounts of existing text, and some of that text describes who a home is perfect for instead of what the home offers. Describe the property, not the people you picture living in it. Rather than calling a home ideal for a certain kind of buyer, say what it has: three bedrooms on one level, a fenced backyard, a short walk to the park.
Keep chat assistants from characterizing neighborhoods or the people who live in them, and send those questions to objective public sources. Review every AI draft before it goes out, the same way you would review work from a new assistant, because the brokerage is responsible for what it publishes.
How Urban Marketing Edge handles this
We use AI the way the Card Members in this survey want it used: for the routine, with a person checking the work and a person on the other end of every conversation that matters.
Drafts start with AI where it saves time, such as first passes on captions, newsletter sections and blog outlines. Every draft then goes through a human edit against the client's voice guide and a fact check against the source, the MLS or the client's own notes. Nothing publishes without the client's approval in their dashboard, and every listing graphic and caption is checked against fair housing language before it is scheduled. Our social media management plans build the month around the agent's own footage and words, so the feed sounds like the people who work there. The automation handles reminders, approvals and reporting. The judgment stays human.
Questions brokers ask
What did the new American Express research find?
American Express surveyed 528 U.S. Consumer Card Members in February 2026 and published the results on September 15, 2026. 95% want the choice between an AI voicebot and a person. 67% said accuracy makes an AI interaction positive, against 35% for speed. 71% would let an AI assistant schedule appointments with their confirmation, and 65% would let it book a hotel. Comfort was lower for personal finance and budgeting.
Will AI replace real estate agents?
Nothing in this research suggests clients want that. As the stakes rise, people grow more cautious about handing decisions to AI, and a home purchase involves major money, legal documents and real emotion. The shift is in how agents spend their time, with more of it going to advice, negotiation and relationships and less to routine follow-up.
Which real estate marketing tasks are best suited to AI?
First drafts of listing descriptions, captions and newsletters. Scheduling with client confirmation. Listing alerts based on a buyer's stated criteria. Turning call notes into CRM entries. Repurposing one long piece, like a market video, into several shorter posts. Every piece still gets a human review before it goes out.
What should always stay human?
Pricing advice, offer strategy, negotiation, contract and inspection questions, and any conversation where a client is anxious or a deal is at risk. Questions about neighborhoods and who lives in them belong with objective public sources, not a chatbot's opinion.
How does an agent start using AI without losing their voice?
Start with one routine task, like open house follow-up emails. Write a short guide to how the agent sounds: the words they use, the words they avoid, how formal they are. Give that guide to the tool every time, then edit each draft until it reads like the agent wrote it. A brokerage can do the same at the brand level so forty agents sound like one company.
Should a brokerage tell clients when they are talking to AI?
Yes. Label chat assistants clearly and make it easy to reach a person. Transparency supports trust, and the survey shows how much people value having a human as an option. Check the brokerage's policies and state rules on AI disclosure as well.
Does speed to lead still matter if accuracy matters more?
Yes. Responding quickly still counts. The lesson is that a fast wrong answer costs more than a slightly slower right one. Use automation to acknowledge a new lead right away and set expectations, then follow up personally with accurate, specific answers.
Is AI-written listing copy a fair housing risk?
It can be, like any copy. The risk comes from language that describes who a home suits instead of what it offers. Review every draft, describe the property itself, and remember that the brokerage is responsible for what it publishes.
Keep the human edge
The American Express research lines up with something the best agents already practice: technology earns its place by making service better, and trust is still built person to person. Use AI for the routine, hold every detail to a high standard of accuracy, and save the agent's time and voice for the moments clients remember.
Urban Marketing Edge builds marketing for solo agents, boutique brokerages and real estate teams, covering content, email, web, video and analytics. If you are working out where AI belongs in your marketing and where your people should stay front and center, let's talk it through.
