Stupidly simple: designing AI tools that field teams actually use
By Vishal Kumar, Founder and CEO · Last updated · 7 min read

In short: Field teams don't avoid software because they're lazy. They avoid it because it asks them to type, scroll and pick from drop-downs while they're doing a physical job. The fix is to design every screen around one task and let AI do the paperwork: a selfie to start the day, one big red button for problems, a photo instead of a form, and a voice note instead of an SAP (enterprise software) screen. Design thinking, starting from the person and their few critical tasks, is how you find that version.
Take a selfie. Press start.
That's how a tractor driver begins his day at a sugar cane producer in West Africa we built a platform for.
No login screen. No menu. No training manual.
The phone is a work phone with one app locked on it. The app knows who he is, where he needs to go and what the next three months of farming look like. It just shows him the next step.
Here's how I described it to a partner last week:
"We started simple. Each persona gets their own login. We actually gave them phones where the app is locked in, so that's the only thing that works. And then just make it very stupidly simple to work. So if your job is just driving the tractor, that's all you see."
"Stupidly simple" isn't a slogan. It's a design brief.
Why do field teams ignore new software?
Because the software was designed for the office.
Most business systems assume a person at a desk, with time, a keyboard and patience. Field staff have none of those. They're on a farm, a factory floor, a customer's shop or the side of a road.
So they do the job, skip the system, and fill in the gaps later. Or never.
One operations manager at a manufacturer we work with summed up the result. His field team wouldn't type updates into SAP. But if he called them, they'd tell him everything in thirty seconds. So he ended up doing a daily round of phone calls just to collect what the system should already know.
As I put it in a session this week, data entry into SAP, Salesforce or your own system is complicated, and nobody likes it.
The real question isn't "how do we make them use the system?"
It's "why does using the system feel harder than not using it?"
- 30 seconds
- for a field worker to give a full update on a phone call
- One operations manager’s experience
What does "stupidly simple" look like?
On that West African farm platform, the field app comes down to four moments:
- Start of day: take a selfie, press start. The app shows where to go, with turn-by-turn directions.
- During the day: if anything goes wrong, press the big red button. That's the alert.
- End of day: photograph the odometer and the fuel gauge. Done.
- Everything else: handled by supervisors and managers on their own dashboards.
The photos aren't decoration. Fuel and equipment misuse was a real cost for the business. A picture of the gauge at the end of each day answers the question without anyone filling in a log.
The same platform also manages crop cycles, expected yields, tractors and outgrowers (independent farmers who supply the mill), with a GIS (geographic information system) map behind it. The complexity is still there. It's just not on the driver's screen.
When we trained the teams, we walked through every screen, day one, day two, day three, and removed anything that was noise:
"It just removed everything that's noise. So the UI and UX is just extraordinarily simple."
| Moment | What the field worker does |
|---|---|
| Start of day | Take a selfie, press start |
| During the day | Press the big red button if anything goes wrong |
| End of day | Photograph the odometer and the fuel gauge |
| Everything else | Handled by supervisors and managers on dashboards |
Why are voice dumps better than SAP forms?
Because talking is the fastest way most people can share what they know.
At that manufacturer with the daily phone calls, we built a simple alternative:
"All you need to do is a voice dump of what you wanted to update. The AI will just transcribe that, structure it into a nice table. You click submit, you are done."
A real update sounds like this: "Order 502 is running a little late. This supplier hasn't come in, and I think that city is shut down, so we won't get anything for the next two weeks."
The AI works out which order, which supplier and which person needs to know, and files it in the right place.
We did the same thing inside Quantana for timesheets. Our team has to log hours for client work, and everyone hated it. Now you press a key and say what you did, from nine to ten, ten to twelve, and so on. The system builds the timesheet.
"Compliance has increased significantly. What used to be very tedious, 'I hate my life, why do I have to do these updates', is now press, speak, done."
I'll admit a bias here. Voice to text is one of those features that even a lot of executives don't use much, and I aggressively prefer it.
| Field | Value |
|---|---|
| Spoken | “Order 502 is running a little late. This supplier hasn’t come in, and I think that city is shut down, so we won’t get anything for the next two weeks.” |
| Order | 502 |
| Status | Running a little late |
| Cause | Supplier hasn’t come in; city may be shut down |
| Outlook | Nothing for the next two weeks |
| Filed to | The order, the supplier and the person who needs to know |
Can a photo replace a form?
Often, yes.
A utilities company in Australia we consult for has crews who fix pumps and pipes. Their app has a comment field and an upload field. The crews tap "done" and leave both empty.
So the data gets worse every year, and nobody can plan maintenance properly.
Our approach: ask for one photo of the equipment's outside label. From that, AI can prefill the make, model and details. It can also note the condition, such as new, old, cracked or rusted, which the crew was supposed to type and never did.
Easier entry means better data. Better data means they can plan next season's budget with real numbers.
The same thinking works with QR codes. On a factory floor, scanning a code on a steel coil and a code on the shelf beats writing the location on paper and typing it into Excel later.
How does design thinking get you there?
You start with people, not technology.
Here's how I brief my own designers:
"Our approach is always: forget the tech. Focus on the design. Focus on the user. So we have a design thinking practice where first we identify the people, the personas. We identify the critical paths: what is your job, what are the five or six steps you need to do it. Then we map out one clear screen or a step that just achieves that outcome."
Most applications come down to five or six critical paths. Get those right and the rest is detail.
There's a second rule that engineers don't always love. What's easy for the user is usually hard to build. If the design doesn't make the engineers sweat, the designer hasn't done their job.
We also design with clients, live. In workshops, people describe what they need out loud and our designers build it on screen as they talk. By the end of an hour or two, the client has watched their own tool take shape. That builds a kind of ownership that no requirements document can. And the first thing we show is a clickable prototype, not a document. If the client isn't impressed by the prototype, we haven't done enough.
What about people who still won't use it?
Then look at incentives, not features.
"Ultimately, every business is people and their incentives. Tech is secondary."
So I ask two questions. First, why is this hard to do? Fix that. Second, even if it's easy, why would this person not do it? Sometimes the answer is that the update makes someone's work visible in a way they'd rather avoid.
One approach I've suggested, and I'll say upfront it's controversial: make compliance visible. A simple daily view, for leaders and the team, of who updated on time. Positive recognition for the people who do it.
Simple design removes the excuse. Visibility removes the hiding place.
The tempting choice and the better choice
run more training. Workshops, manuals, videos, refresher sessions. When adoption is still low, run more.
remove the barrier to entry. If someone needs training to report a problem, the screen is too complicated. A big red button doesn't need a workshop.
For more on how we approach AI adoption across a business, see how we work, our piece on voice ordering and speech AI for field sales, and the manufacturing page.
Your next step
Ask one of your field staff to show you how they report a problem today. Count the taps.
If it's more than three, what would a big red button replace? Get your AI roadmap.
Frequently asked questions
Why do field workers resist new apps?
Usually because the app was designed for a desk. Typing, drop-downs and long forms don't fit a physical job. When the app takes one tap, a photo or a voice note, resistance drops.
What is a voice dump?
A short spoken update that AI transcribes and turns into structured data, such as an order status or a timesheet entry. The person reviews the result and submits it.
What is design thinking?
A way of designing that starts with the people who will use the product. You identify personas, map their few critical tasks, design one clear screen per step, and test with real users before building the rest.
Do field teams need smartphones for this?
They need a device with a camera and a microphone. In some projects we supply locked work phones so the app is the only thing on them.
How do you measure adoption?
Track daily active use, on-time updates and how much data arrives complete. If people are still calling each other for updates, the tool isn't working yet.





