Introducing Ask Faraday: Your favorite AI assistant, now with data-backed answers about your customers and market

Today we're launching Ask Faraday. It lets the AI assistant your team already uses answer questions about American consumers and about your own customers, using data from the Faraday Identity Graph, which covers more than 240 million US adults and their households. See Ask Faraday in action and request access.

This post covers the story behind the launch: why we built it, how it keeps personal information out of reach, and a set of prompts you can try once you're connected.

Why we built it

The questions marketers have about their customers are often simple ones. How do my customers compare to the people who live near them? Do the neighborhoods where I sell well have something in common? And they usually come up at decision time, while a campaign is being planned, a budget is being set, or a new market is on the table.

Getting a solid answer has usually meant going the long way around: commissioning a research study, waiting for a data team to pull a custom report, or piecing together public statistics that don't quite match your own customers. An AI assistant on its own will reply right away, but without data about real people behind it, it's working from general knowledge. Either way, the answer tends to arrive too late to help, or isn't solid enough to act on.

We built Ask Faraday so that the people who need an answer can get it when they need it. Your AI assistant asks the question directly and brings back an answer drawn from real data on US households, in the same conversation.

How privacy works

Letting an AI assistant ask questions about people directly raises a fair question of its own: what stops it from asking about one person? We designed Ask Faraday so that the answer doesn't depend on the assistant behaving well. The protections sit in the data and the database, where no prompt can reach them.

• Small groups are refused. Any result that describes fewer than 50 people is refused. If you narrow a question until it describes only a handful of households, you get no answer rather than one that could point to someone.
• The minimum is set on the data, not in the conversation. The 50-person minimum is set on the data itself in Google BigQuery, the database where it's stored. Asking the question a different way can't lower it. It holds even if someone tricks the assistant into ignoring its instructions, and even if our own code has a bug.
• Personal details were never there to begin with. Names, emails, and addresses aren't in the data the assistant can query at all. They aren't filtered out of answers afterward. They were never there.
• It's not a new idea. BigQuery's data clean rooms, which companies use to work with shared data without exposing individual records, rely on the same mechanism.

One more design choice is worth explaining. Behind the scenes, the assistant asks its questions in SQL, the standard language for querying databases, which AI assistants already know well. We could have invented a Faraday-specific query language instead, but that would only have given the assistant something new to learn. Safety comes from what the data contains and what the database will return, not from how the question is written.

"Every SaaS company is bolting a chat box onto a pile of data and getting back confident answers nobody can check," said Andy Rossmeissl, CEO and co-founder of Faraday. "We let the assistant write real queries against real data, and we let the database refuse the ones it shouldn't answer."

Use it alongside your other tools

Because Ask Faraday runs inside your own AI assistant, it isn't a separate app you have to switch to. It works alongside the other tools you've connected to that assistant, such as your CRM, so a single request can draw on more than one of them.

If you've connected HubSpot, for example, try a prompt like this:

"Pull my closed-won and closed-lost deals from HubSpot, then use Faraday to compare who those people are against the US market."

The assistant passes the work between the two tools, so you don't have to move a list from one to the other yourself.

Getting started

If you'd like to try it yourself, here's what a first session looks like, from connecting your assistant to turning answers into next steps.

Connect your assistant

Setup takes about five minutes. In Claude or ChatGPT, add Faraday as a custom connector, then sign in with your Faraday account. To check that everything is working, ask:

• "Give me an overview of my Faraday account."

Ask about the market

Questions about the US market don't need any data from you. You can start asking right away:

• "How many homeowners aged 35–54 live in Colorado, and what's their household income range?"
• "Which counties in the Southeast have the most households with young children?"
• "What share of households in Phoenix are renters vs. owners?"

Compare against your own customers

The most useful answers often come from comparing the market with the people who already buy from you. To do that, add a customer or lead file to Faraday and ask the assistant to create a group from it. Then you can ask questions like:

• "How are my customers different from the general US population?"
• "How do repeat buyers differ from one-time buyers?"
• "Which of my markets have fewer customers than you'd expect, given how many similar households live there?"

Turn answers into next steps

Once you know more about who your customers are, the same conversation can help you decide what to do about it:

• "Based on my best customers, which channels or creators would reach more people like them?"
• "Write a one-page summary of who to target and why."

Tips for better answers

A few habits make a noticeable difference:

• Be specific. Name the place, the age range, or the kind of household you care about.
• Ask for comparisons. "How do my customers compare to…" usually tells you more than a single number on its own.
• Ask follow-ups. Questions like "break that down by region" build on the previous answer, so you don't have to start over.
• Widen the question if it's turned down. If a question is rejected for being too narrow, that's the 50-person minimum at work. Widen the area or loosen a filter and try again.

Availability and pricing

Ask Faraday is available as a $500/month add-on for Faraday Enterprise customers. To add it to your account, contact your account team. If you're not on Enterprise yet, you can talk to our sales team.

To see what a conversation with Ask Faraday looks like, or to request access, visit the Ask Faraday page.