We should ban manual CRM entry completely, not just reduce it. Manual entry means deciding what goes into which field. The alternative is to say what happened and let the system figure out what to write. Most of what the CRM needs can already be captured without anyone typing. The hard part is what happens in person, and that's where voice comes in.

In Validity's survey of CRM users, 76% said less than half of their CRM data is accurate and complete. Every sales team talks about CRM hygiene, and very few have it. So the rule is simple. If a person typed it, don't trust it.

Why is CRM data always wrong?

CRM data is wrong because it's written from memory, and memory fails even the most careful rep. Most updates happen on a Friday afternoon, when a rep tries to remember what everyone said in a week of meetings. Johan has written about why reps don't update the CRM in the first place.

And we don't remember much. In one study, two people had a conversation and were asked about it a week later. They agreed on only 4.7% of the details.

One conversation, one week laterShare of details
Recalled by both people4.7%
Missed by at least one of them95.3%
Recalled by bothMissed by at least one
Brown-Schmidt, Jaeger, Evans and Benjamin, Cognitive Science (2023). Pairs of participants recalled a conversation one week later.

Why ban manual entry instead of reducing it?

If typing is allowed at all, it stays the fallback. Most teams try to reduce it instead, with fewer required fields, some automation and a weekly reminder. But the moment something isn't captured, someone opens the CRM and types it in.

If a person typed it, don't trust it.

The catch is that not every meeting can be recorded. Video calls and emails can go straight into the CRM, but McKinsey finds that B2B buyers still use in-person meetings for roughly a third of their buying, at every stage. Nobody puts a phone on the table at a customer dinner, some customers don't allow recording, and people talk less openly when they know they're being recorded. For those meetings, the fallback has always been to type it in later.

What if the meeting can't be recorded?

You speak right after it instead. Record a voice note on the way out, while everything is still fresh, and the system turns it into CRM updates, so nobody types anything later. That timing matters. In one memory study, people recalled 81% of a text five minutes after reading it, 54% two days later and only 42% after a week. A few minutes after the meeting beats Friday afternoon by a wide margin.

What do you say in the voice note?

Say what happened, not what to update. You don't say "update the close date to March", you say "they want to start after their board meeting in March". The system works out which records and fields that affects: the close date and the next step.

The same goes for anything else that happens outside a recorded meeting.

  • After a site visit, "their warehouse is at full capacity, and they want a proposal for the second site before Christmas" becomes deal notes, a next step and a close date.
  • After a dinner, "Anna is leaving in January, and her replacement from finance will sign" updates the contacts and who the champion is.
  • From the car, "they went with a competitor for Norway, but Sweden is still open" closes one deal with a reason and keeps the other open.
  • At a trade show, "I met their head of procurement, who wants a follow-up after the budget is set in November" becomes a new contact and a November follow-up task.

Voice also gives better data. Speaking a short message is three times faster than typing it, with 20% fewer errors, so people say more than they would ever type. Medicine is already moving this way. AI scribes listen to the appointment and write the note, and in a trial with 238 doctors they reduced burnout compared with doctors who kept writing their own. Gartner expects 60% of seller work to run through conversational AI by 2028, up from less than 5% in 2023, so sales is heading the same way.

What changes when nobody types

The obvious question is whether the system gets it right. Compared with memory, it isn't close. The system works from the whole conversation or a voice note recorded minutes after the meeting, not from what a rep still remembers on Friday. And since every field has a source, you can check where something came from and fix it. Getting there looks like this.

  1. Capture meetings and email. Let the CRM update from what happened, right after it happens.
  2. Add voice for the rest. Anything outside a recorded meeting goes in as a voice note.
  3. Make the fields read-only for reps. This is the actual ban, and the step most teams skip.
  4. Use the data. Check the source when something looks off, and spend the time on customers instead.

Salespeople will love it, since nobody got into sales to update properties at the end of the week. RevOps will probably take longer, but they'll learn to like it too, because they can stop policing the CRM and start working with data they can check. That kind of trust builds in steps, which I wrote more about in my last article. Compliance gets easier as well, since every change has a source.

If a person typed it, don't trust it. If a person said it, let the system write it.

Frequently asked questions

CRM hygiene is how accurate, complete and current the data in a CRM is. Most teams try to improve it with rules, required fields and reminders. The most reliable way is to stop people typing into the CRM and have the system write the data from meetings, emails and voice notes, so every field has a source.
Stop asking them to type it. Mandates, required fields and training fade within months. Capture updates from the meetings and emails that already happen, and let reps describe the rest in a voice note, so the CRM is written from what actually happened.
Spiich lets reps update HubSpot, Attio and Pipedrive by describing what happened in a voice note. The system decides which records and fields to change, so the rep never has to find the right property.
Record a short voice note right after the meeting saying what happened, who was there and what was agreed. A voice to CRM tool such as Spiich turns it into updates on the right deal, contacts and next steps, so nothing depends on remembering it later.
They should be able to correct it, but typing should not be the normal way data gets in. When fields are written from meetings, emails and voice notes, every value has a source. Letting people type freely brings back data written from memory that nobody can check.