Ideas · October 19, 2026
Your Best AI Product Idea Is Sitting With Your Frontline Team
The best AI product ideas in a service business come from the people who talk to clients every day. The exact questions we ask them, and how to pick one idea.
When owners ask us where to start with AI, they usually expect the answer to come from a strategy session. A whiteboard, the leadership team, a list of trends. That session rarely produces the idea worth building.
The best ideas we find almost never come from the top of a company. They come from the people who deal with clients every day: the account managers, the coordinators, the technicians, the people answering the phone and the inbox. They know exactly where clients get stuck, what they keep asking, and what they would happily pay to have solved. Most of them have already thought about it.
This post walks through how we find those ideas: the questions we ask, how we run the conversations, and how we decide which idea is worth turning into a product.
Why the frontline knows more than the boardroom
Leadership sees the business through reports. Revenue, margins, utilization, churn. Those numbers tell you what happened. They do not tell you what a client said on a Tuesday afternoon when they were confused, frustrated or delighted.
Frontline employees hear that every day. They answer the same question for the fortieth time this month. They see the moment a client hesitates before signing. They know which part of the service clients value most and which part they barely notice.
That is exactly the information you need to invent a product. A good customer-facing AI tool removes a friction the client feels, or gives them something they wanted and could not get before. The people closest to the client already know where those are.
There is a second reason. Since AI tools became part of everyday work, a lot of employees have quietly started imagining what they could build. They have an idea. They just never had the time, the skills or the permission to pursue it.
The four questions we ask
We keep the conversation simple. Four questions, asked in this order, each building on the last.
1. What is your experience working with the client?
Start broad. Let them describe the relationship in their own words: who the clients are, how they communicate, what a good week and a bad week look like. This warms up the conversation and often surfaces things you would not think to ask about directly.
Listen for emotion. Where they sound tired, there is usually friction. Where they sound proud, there is usually value the business is not charging enough for.
2. What are you asked most often?
This is the most important question. Repetition is a signal. If clients keep asking the same thing, there is a gap between what they need and what the service currently gives them on their own terms.
Push for specifics. Not "status updates," but "clients email every few days asking where their application is, and I have to check three systems to answer." The specific version is where the product idea lives.
3. You have probably already had an idea for something you could build with AI. What was it?
Ask it exactly like that. The assumption matters. Most people have had the idea, and framing the question this way gives them permission to share it instead of feeling like they are pitching.
You will hear a wide range. Some ideas are small internal shortcuts. Some are surprisingly ambitious. A few will make you sit up, because they describe something your clients would use every day and your competitors do not offer.
4. What stopped you from building it?
This tells you what kind of help the idea needs. The answers usually fall into a few groups:
- It was too hard. They did not have the technical skills, or they tried and got stuck.
- It would take too much time. They have a full-time job serving clients.
- They were unsure what it should look like. They could describe the problem but not the design.
- They did not think they were allowed. Nobody asked, so they assumed it was not their place.
Every one of those blockers is solvable. The last one is solved the moment you ask the question.
How to run the sessions
A few things make these conversations work.
Talk to people one at a time. In a group, people defer to whoever is most senior or most confident. One-on-one, the quiet coordinator who has been thinking about something for months will tell you about it.
Keep it short. Thirty minutes is enough. Longer, and it starts to feel like a performance review.
Go wide across roles. Talk to the person who onboards new clients, the person who handles complaints, the person who delivers the work and the person who sends invoices. Each sees a different part of the client relationship.
Take their words down exactly. The way they describe a client's frustration is often the way you will later describe the product to clients. Do not paraphrase it into management language.
Do not judge ideas in the room. If an employee sees you dismiss their idea, the next person will hear about it and hold back. Collect everything. Sort later.
Sorting the answers: back office or customer-facing
After the conversations you will have a long list. The first sort is the most important one: does this idea make internal work faster, or does it give clients something new?
Take an example. A firm that handles visa applications runs these sessions. One coordinator says she spends hours each week re-formatting documents clients send in the wrong layout. Another says clients constantly ask whether their case is on track, and nobody can answer quickly without digging through files.
The first idea is back office. Automating document formatting is worth doing. It is cheap, it works, and it saves real time. It also does nothing for the firm's position in the market, because every competitor can do the same with the same tools.
The second idea is customer-facing. A tool that lets clients see exactly where their case stands, what is missing and what happens next, at any hour, is something clients feel directly. It changes what it is like to be this firm's client. That is the kind of idea that can win share.
We recommend doing both. But only the customer-facing list contains the product. We explain the reasoning behind that split in everyone uses AI, almost nobody profits from it.
Picking the one idea
From the customer-facing list, pick one. Not three. These questions help narrow it down:
- How often does the problem come up? An issue clients hit every week beats one they hit once a year.
- Would clients notice if it disappeared? If they would, it has real value.
- Does it use what makes you different? Your expertise, your data, your process. If any firm in your industry could build the identical tool, it will not set you apart for long. See the commoditization test.
- Could it be sold, or does it help you sell? The strongest ideas either become a product clients pay for or become the reason they choose you.
- Is there a person who cares about it? Often the employee who raised the idea becomes its best champion. Involve them.
If two ideas tie, pick the one the frontline team is most excited about. They will be the ones introducing it to clients.
What to do next
Once you have the idea, resist the urge to write a long specification. Get to something clients can see and click as fast as possible. A working prototype tells you more in a week than a scope document does in a month. We explain why in why we build a prototype instead of writing a scope.
Bring the employee who had the idea into the prototype review. They know what clients will try first, and they will catch problems you will not.
How we help
This is how every engagement starts for us. We talk to the people closest to your clients, find the idea that uses what makes your business different, and build a working prototype you can click through before any money changes hands. If it lands, there is one fixed build fee, and after launch we stay on as product lead, paid as a share of what the product earns.
If you already suspect the idea is somewhere in your team, book a call. We will help you find it.