How to Use AI in Your Field Sales Process

The short version: AI helps a field team most at the two ends of a customer visit, the preparation before and the capture after. It helps least in the middle, during the visit itself, which is the opposite of what the sales AI market is built around.

That inversion is why so much of this technology has not reached route teams. Most sales AI assumes a call it can listen to. A plant visit, a counter conversation, a walk around a job site: none of them produce a recording. So the tools that transformed inside sales arrive in field sales with nothing to work from.

This is a map of where AI genuinely fits an industrial field sales process, stage by stage, and what to adopt first if you only do one thing.

An industrial field sales rep sitting in a parked pickup truck outside a plant, phone resting in a dashboard cradle, listening to a briefing before walking in, vivid cobalt sky above

Field sales got skipped, and the numbers show it

AI is close to universal in sales overall and nearly absent in the field.

SPOTIO's State of Field Sales 2026, a survey of 452 sales professionals of whom 388 work with field sales teams, found that 33% of field sales teams are not using AI in any capacity. The same report notes that Salesforce's 2026 State of Sales put adoption across sales organisations generally at 87%.

A third of field teams at zero, against something close to universal adoption elsewhere. That gap is the whole story.

It gets more specific. Where field teams do adopt AI, they adopt the light end of it. Automated email and message personalisation is the most common use at around 30% of respondents, with conversation intelligence close behind, and automated CRM data entry at roughly 24%. The higher-value uses, lead scoring, predictive analytics and customer behaviour analysis, are each used by fewer than 20% of teams.

So the pattern is not that field sales rejected AI. It is that field sales got the version of AI that was easiest to bolt on, and not the version that addresses what actually costs a route team time.

Where AI fits, stage by stage

Walk a normal day and the picture gets clear quickly.

Before the visit: preparation

This is the strongest and least controversial fit. A rep with eight stops cannot research eight accounts properly, so most preparation happens in the car park in the ninety seconds before walking in.

AI closes that gap well because everything it needs already exists in your systems: order history, open quotes, recent service tickets, who was spoken to last and about what. Pulling that into a two-minute briefing is a solved problem, and it works whether the briefing arrives as a phone call, a message or a screen.

During the visit: live answers

Narrower than vendors suggest. In practice this means quick factual lookups: is this in stock, what is the lead time, what did we charge this customer last time.

Genuinely useful, occasionally decisive in front of a customer, but it is a convenience rather than a transformation. Be careful of anything that asks a rep to interact with a device during a conversation. Reps will not do it, and customers notice.

After the visit: capture

This is the one that matters, and it is where the market has the least to offer.

Everything a company knows about what happened at that plant is in the rep's head between leaving the site and sitting down that evening. Inside sales solved this with recording and transcription. Field sales cannot, because there is no recording, which is why conversation intelligence tools break here. We went through that specific mismatch in whether AI note-takers can work for in-person field sales.

The approach that does work is a spoken debrief. The rep talks for two minutes from the truck, and the system turns that into the CRM update, the follow-up draft and the quote action. No recording of the customer, no typing, no app to open.

CRM and pipeline: accuracy as a by-product

Nobody buys AI to improve CRM hygiene, and yet this is where the return usually shows up.

A pipeline is only as good as what got entered, and in field sales what gets entered is whatever a tired rep remembers at nine in the evening. Fix capture and forecast accuracy improves without anyone running a data project, which is the mechanism behind why field sales pipelines are so often wrong.

Follow-ups and coaching

Follow-up drafting is a good fit: the AI knows what was discussed, so a first draft costs the rep nothing and gets the email out the same day rather than Thursday.

Coaching is real but slower. A manager who can see what actually happened across a territory, rather than what got reported, can coach on patterns instead of anecdotes. That only works once capture is reliable, so treat it as a second-year benefit rather than a reason to buy.

A three-part sequence of the same field rep: arriving at a plant gate, shaking hands with a maintenance manager on the shop floor, then speaking into his phone back in the truck

What to adopt first

If you do exactly one thing, do capture.

The argument is simple arithmetic. Preparation saves a rep a few minutes per visit and makes those visits better. Capture removes an entire second shift: the evening hour of typing that happens after the working day ends, every day, for every rep. Reps already lose a large share of the week to non-selling work before that evening even starts, which we broke down in how much time sales reps spend on admin.

There is a second reason to start there, less obvious and more important. Capture is the input that every other use depends on. Briefings are only as good as what was captured last visit. Forecasting is only as good as what got logged. Coaching is only as good as the record of what happened. Adopt forecasting first and you get a confident model reading thin data. Adopt capture first and everything downstream gets better on its own.

This is the category now sometimes called an AI sales coordinator: a tool the rep talks to rather than types into, which handles the CRM update, the follow-up and the quote reminder from a short spoken debrief. It sits in front of the CRM rather than replacing it, which is the distinction we set out in what an anti-CRM actually is.

How to judge any AI tool for a route team

Six checks. Any vendor can fail them, and most fail at least one, which is what makes them worth asking.

1. Does it need a recording? If yes, it was built for inside sales. Ask directly what it does with an in-person visit, and listen for whether the answer involves the rep typing.
2. What does the rep have to do that they do not do today? If the honest answer is more than talking for two minutes, you have added a step, and added steps get skipped by week six.
3. Does it write into the fields your managers report on? A transcript or a summary parked somewhere is not a CRM update. Ask to see the actual record it produces.
4. Does it work one-handed, or not at all? A rep in a truck park with a hard hat under one arm has no free hand and nowhere to look. Anything requiring a five-field form is theoretical.
5. What happens when it is unsure? Silently guessing a deal stage is worse than flagging it. Ask to see an uncertain case, not a clean one.
6. Is the integration you need available at the tier you are buying? Native CRM and ERP write-back is frequently enterprise-only in this category, and buyers discover it after the seat maths is settled.

The sixth catches more teams than the other five combined.

A branch sales manager at a desk reading an up-to-date territory summary on a monitor, calm and unhurried, bright window with an industrial yard beyond

What this looks like in practice

Start with one thing, measure one thing.

Pick two reps rather than the whole branch. Run capture for six weeks. Then ask a single question: how many visits did each of them type up last Thursday, before and after. If that number did not move, the tool is decoration regardless of how the software demonstrated.

The teams getting value from AI in field sales are not the ones who bought the most capable model. They are the ones who correctly identified that their bottleneck was the hour after the visit, not the hour before it, and bought for that. Teams cutting field sales admin this way, including with Otto Sales (Otto), the AI sales coordinator for industrial field sales, generally start at capture and let preparation, forecasting and coaching follow from it.

FAQ

How is AI used in field sales? Mostly at the two ends of a customer visit. Before it, AI assembles a briefing from order history, open quotes and service issues. After it, AI turns a short spoken debrief into the CRM update, the follow-up draft and the quote action. During the visit its role is narrow, limited to quick lookups like stock and lead times.

Why have field sales teams adopted AI more slowly than inside sales? Because most sales AI assumes a recordable call and a rep at a desk, and field sales has neither. SPOTIO's State of Field Sales 2026 found 33% of field sales teams use no AI at all, against roughly 87% adoption across sales organisations generally.

What AI should a field sales team adopt first? Capture, meaning the conversion of what happened on a visit into a CRM record without the rep typing. It removes the largest single block of wasted time, the evening admin shift, and it is the input every other AI use depends on. Forecasting AI built on thin data just produces confident guesses.

Can AI update the CRM without a rep entering data? Yes, if it takes a spoken debrief rather than a recording or a form. The rep talks for two minutes from the truck and the system writes the update. Only about 24% of field teams currently use automated CRM data entry, so this remains an advantage rather than table stakes.

What is an AI sales coordinator? An AI sales coordinator is a tool a field rep talks to instead of typing into. It briefs the rep before a customer visit and, after a short spoken debrief, writes the CRM update, drafts the follow-up and flags the quote action. It sits in front of the CRM as a capture layer rather than replacing the system of record.

TL;DR

• AI fits field sales at the two ends of a visit: preparation before, capture after. The middle, during the visit, is the weakest fit and the one most heavily marketed.
• The adoption gap is real and measured. 33% of field sales teams use no AI at all, against roughly 87% adoption across sales organisations generally (SPOTIO, State of Field Sales 2026).
• Field teams that do adopt get the light version: around 30% use automated message personalisation and 24% automated CRM entry, while lead scoring and predictive analytics sit under 20% each.
• Adopt capture first. It removes the evening admin shift and it is the input that briefings, forecasting and coaching all depend on.
• Judge any tool on six checks, starting with whether it needs a recording. A plant visit does not produce one, and everything follows from that.

Work out where your bottleneck actually sits before you look at products. For most route teams it is the hour after the visit, not the hour before, and knowing which one you are solving makes the rest of the decision straightforward.

By Omkar Pandharkame, Co-founder of Otto.