Your CRM's AI can't read visits nobody logged.

Salespeople in the field don't update the CRM because logging a visit costs them more, at the worst possible moment, than it gives back. That matters more now than it used to. The big CRMs hold on to customers through the data stored in them, and they are adding AI that runs on that data. Neither works if visits never get written down. Buying more AI won't fix that. Making capture cheap enough to happen at the customer's gate will.

This article covers what really keeps a large CRM in place, why the field is where its data goes thin, and a one-week check you can run on your own CRM before you spend money on anything.

What keeps a big CRM in place

People rarely leave a large CRM because they love it. They stay because leaving is hard. Three things do most of the holding:

• Data gravity. Years of records, custom objects and workflows built on top of them. Custom code doesn't move to a new system, and some contracts give you as little as 30 days to get your data out after you cancel (more on lock-in).
• The ecosystem. Thousands of add-on apps and a large network of certified partners who build and change the system for you.
• AI on top of the data. The newest layer: assistants that summarise accounts, draft follow-ups and answer pipeline questions, often billed per action on top of the seat price. One large vendor publishes a rate of about $0.10 per standard AI action, roughly ₹10.

All three are real advantages for a head office. Notice what they have in common: each one gets more valuable as more data goes in. None of them does anything to get the data in.

The step every moat depends on

Somebody has to type it.

At a desk, that happens during the call, with a keyboard, a second screen and the CRM already open. In the field it happens between two visits, on a phone, standing next to a vehicle, or it doesn't happen at all.

The best numbers we found on this come from a 2026 field-sales survey (vendor survey, mostly US respondents). It is not independent research, and not data from India. In it, B2B field salespeople said they spend about a third of their time actually selling, and 65% of respondents said they spend five or more hours a week on manual CRM entry. Even read cautiously, the direction is clear: data entry is a large share of the job, and it's the share salespeople resent most.

Why field salespeople skip it

Field salespeople are not lazy. Logging fails in the field for structural reasons.

The form

A typical visit log asks the salesperson to open the app, pick the activity type, choose an outcome, write notes, set a follow-up date and save. For one visit, that's fine. After the tenth visit of the day, it gets pushed to “tonight”, and tonight it loses to dinner. Logs written from memory at the end of the week describe a much quieter week than the one that happened.

The phone

Many CRMs were designed for the desktop and squeezed onto a phone. Salespeople judge a CRM by its mobile app, and when that app is a cut-down version of the desktop, they log less. What salespeople want is to say what happened in their own words, often in their own language, not squeeze it into dropdowns.

The signal

In the large CRMs, working offline isn't something the app just does. An administrator has to set it up, choosing which records salespeople can see without a connection, within limits such as 50,000 records per object (another large CRM works the same way). A salesperson in a basement warehouse or a rural industrial estate only has what someone loaded for them in advance.

The setup

Small teams rarely have an admin at all. When a CRM is built for companies with a dedicated administrator, and nobody owns the configuration, the field version never gets set up properly.

What AI does with an empty CRM

AI in a CRM is a reader. It summarises, ranks and answers from what's stored. Ask it “which deals are going cold?” and it looks for the last logged activity. If the last real visit was never logged, the deal looks cold when it's warm, or warm when the customer stopped answering three weeks ago.

So a team with patchy visit logs pays twice: once for the AI, often metered per action, and again in decisions made on a picture of the week that isn't true. The AI isn't wrong. It's faithfully reporting the gaps.

That's why “add AI” is the wrong first move for a field team. The first move is to make sure the visits exist in the system at all.

Make capture cheaper, not smarter

If logging fails because it costs too much at the wrong moment, the fix is to make it cost almost nothing at that moment. Three things matter.

1. Say it instead of filling it in. The salesperson describes the visit the way they'd tell a colleague, for example “met the purchase head, discussed pricing, demo next week”, and the system turns that into the activity type, outcome, notes and a follow-up task. One sentence becomes one logged visit. This is how visit logging in Field CRM works, by voice in English, Hindi or Hinglish, or by typing the same sentence.
2. Photograph instead of typing. Field leads arrive as visiting cards, not web forms. Pointing the phone at a card should create the company record, read details like the GSTIN, and check for duplicates. See the visiting card scanner.
3. Verify at the moment of logging. A manager's quiet question about field data is “did this visit actually happen?” If every log is tagged on-site or off-site when it's saved, the manager has a reason to trust the data without a Monday phone call, and data that's trusted gets used. Field CRM doesn't track location in the background. It captures location only at the moment a visit is logged.

Once capture is cheap, the AI has something to read. In Field CRM, the AI reporting assistant answers plain-language questions from live CRM data, and the MCP connector brings the same reports into Claude or ChatGPT; both are on the Advance plan (what MCP is, explained). That only helps because the visits are there to report on.

A one-week check you can run on your own CRM

Before you buy anything, measure the gap you actually have.

1. Pick one salesperson and one week. Ask them to keep a plain paper or WhatsApp list of every customer conversation: visits, calls, chats.
2. Compare it to the CRM at the end of the week. Count how many made it in, and how many were logged the same day.
3. Time one log. Watch the salesperson log a single visit on their phone, from pocket to saved. If it takes longer than the walk back to the bike or car, you've found the problem.
4. Ask the pipeline a question. “Which deals haven't been touched in two weeks?” Check the answer against what the salesperson says. The difference is what your AI is working from.
5. Repeat on the oldest phone in the team. That's where slow apps hurt most.

If most conversations made it in on the same day, your CRM is fine and AI will help. If they didn't, fix capture first. For what to look for in a field CRM, the five-minute buyer's test is a good place to start.

In short

Large CRMs keep customers through stored data, an ecosystem and, increasingly, AI. All three depend on salespeople entering what happened, and the field is exactly where that breaks: long forms, phone apps built for desks, offline that needs an admin, setup that needs an expert. Measure your own gap for a week. If visits are going missing, the answer is cheaper capture, not more AI.

Frequently asked questions