# Why Sellers Still Copy AI Notes Into the CRM

**Analysts:** Ines Bahr and Aurora Martina Granata

*G2 AI Custom Research, commissioned by Pipedrive. Published November 6, 2026. Based on in-depth interviews with 113 active sellers.*

Sellers already use AI to draft emails and capture meetings. Getting that work into the customer record is still their job. Across interviews with 113 active sellers, most of them solopreneurs or on small teams, only 10% have AI updating the CRM for them, and most want to keep the final say on anything AI writes.

The last mile is the gap between an AI-generated summary and a customer record a seller trusts: the copying and re-keying that today’s AI note-takers leave behind.

## Key findings

- **67%** of sellers who use AI meeting capture still copy its output into the CRM by hand.
- **57%** of sellers have no dedicated CRM or keep their work spread across other tools.
- **76%** of sellers say admin takes time from selling or spills into their personal time.
- **72%** want high human review and final sign-off on AI-generated notes, summaries, emails, and research.

## Executive summary

Among sellers who already use AI to capture meetings, 67% still handle customer relationship management (CRM) data entry by hand, copying AI output into the record themselves. That is the clearest finding from in-depth interviews with 113 active sellers, most of them solopreneurs or members of sales teams of up to 10 people. AI has a place in the selling day, and 45% of sellers mainly use it for drafting, research, and content support. The work stops at the customer record, where only 10% of all sellers have automated that step. Sellers also want to keep control, and 72% ask for high human review and final sign-off on AI-generated work.

## Chapter 1: The admin tax — How much time do salespeople actually spend on CRM admin instead of selling?

**Short answer:** More than most would like. About half of sellers say selling leads their day, but admin rarely stays small: 57% say it takes time away from selling or prospecting, and another 19% say it spills into their personal time.

- **Even a selling-led day carries admin.** Among sellers with selling-led days, 65% still say admin takes time from selling or spills into their personal time. Only 24% of all sellers call their weekly admin low or contained.
- **The cost lands harder on small teams.** On teams of 2 to 10, 62% say admin displaces selling, compared with 49% of solopreneurs. Solo sellers are also more than twice as likely to call their admin light.

### About half have selling-led days, while 48% juggle mixed or admin-heavy workloads

| Response | Share |
|---|---|
| Selling-led, limited admin | 52% |
| Mixed selling and admin | 29% |
| Admin- or operations-heavy | 19% |

*Source: G2 AI Custom Research interviews for Pipedrive, 2026. Based on 113 sellers who answered this question.*

> “It takes about seven hours per week, and that takes away from our personal time and time that we get to spend with our son.”
> — Owner, small sales team

**What this means:** Remove the recurring admin that crowds out prospecting. The clearest benefit of automating data entry in the CRM is time given back to selling, with a quick review step: 76% of sellers say admin takes time from selling or their personal time. Small teams feel the cost most, so start there.

## Chapter 2: Post-call capture — Why do CRM records end up incomplete or out of date at small businesses?

**Short answer:** Capture is still mostly manual. Capture after the call is still mostly manual, and gaps are common. After a call, 41% of sellers rely on manual or fragmented note-taking. One in three log notes later in the day, in batches, or inconsistently. And about half work from notes with omissions, errors, or gaps, so the record often trails the conversation.

- **Most sellers log notes promptly.** Two-thirds do it during or right after the call. The other third log notes later, in batches, or inconsistently, and that is where records fall behind.
- **Half of sellers trust their notes, and half work around gaps.** Just over half call their notes complete and accurate, and most of the rest call them usable but with occasional omissions or errors.

### Manual capture edges out AI-assisted capture and CRM logging

| Response | Share |
|---|---|
| Manual or fragmented | 41% |
| AI-assisted or automated | 33% |
| CRM or sales-system logging | 26% |

*Source: G2 AI Custom Research interviews for Pipedrive, 2026. Based on 112 sellers who answered this question.*

> “I threw my notes together sometimes on post its, sometimes in Outlook. And then I schedule a follow-up email or call. But I do it all manually.”
> — Owner, solo seller

**What this means:** Design post-call workflows for fast capture and quick review. Waiting for sellers to finish records later is risky when a third already fall behind. The priority is a structured record the seller checks right after the conversation, with their approval as the check on accuracy.

## Chapter 3: The shadow CRM — Where does sales work happen, if not in the CRM?

**Short answer:** Outside a single CRM. For more than half of sellers, outside a single CRM. Overall, 31% have no dedicated CRM. Another 26% have one but keep their work spread across other tools.

- **Solo sellers carry most of the gap.** Without a CRM, their customer record lives in spreadsheets, inboxes, and informal notes. That population is large: the U.S. Census Bureau counts more than 28 million nonemployer businesses, and the SBA’s Office of Advocacy reports that roughly four in five U.S. small businesses have no employees.
- **Owning a CRM still leaves gaps.** Only half of sellers with a CRM keep most of their selling work in it. The rest split their work with other tools or do most of it elsewhere.
- **When work stays in the CRM, sellers credit an integrated, single-system workflow.** When it leaves, they point to specialized channels and other tools, or to a CRM that takes too much time or is hard to use.
- **Consumer retail and local-service sellers are far more likely than B2B sellers to run without a CRM.** In this group, 41% have no dedicated CRM, compared with 24% of B2B professional, technology, and industrial sellers.
- **Sellers who use AI every day are far more likely to work from a CRM.** Among sellers we matched to screener data, 61% of those who use AI less than daily have no dedicated CRM, compared with 30% of daily AI users. AI adoption and CRM adoption move together, so the sellers who use AI least are also the furthest from a structured customer record.

### Fewer than half of sellers run on a dedicated CRM

| Response | Share |
|---|---|
| Dedicated CRM in use | 43% |
| No dedicated CRM | 31% |
| CRM exists, work fragmented | 26% |

*Source: G2 AI Custom Research interviews for Pipedrive, 2026. Based on 112 sellers who answered this question.*

### 61% of less-than-daily AI users have no dedicated CRM, against 30% of daily users

| Group | Have a CRM | No dedicated CRM |
|---|---|---|
| Use AI daily (67) | 70% | 30% |
| Use AI less than daily (23) | 39% | 61% |

*Source: G2 AI Custom Research interviews for Pipedrive, 2026, matched to screener answers. Based on 90 sellers with screener data who answered this question. “Have a CRM” includes sellers whose work stays fragmented across tools. Most of the less-than-daily AI users came from one recruitment panel, so read this comparison as directional.*

> “I don’t know that a lot of the real selling happens there. Again, like, selling is the conversation. It’s the pitching. It’s the discussing and understanding their needs.”
> — Seller, small sales team

**What this means:** Meet sellers where their records already live. Give the 31% without a dedicated CRM, and especially solopreneurs, a simple way in from spreadsheets and email. For the 26% whose work stays fragmented, connect the tools they already use before asking them to consolidate.

## Chapter 4: AI at work today — How are small sales teams using AI in their day-to-day selling?

**Short answer:** Mostly for writing and research. Mostly for writing and research, not yet as an end-to-end AI sales assistant. Overall, 45% of sellers mainly use AI for drafting, research, and content support, and 42% mainly for meeting capture or broader workflow support. Few use it all the way through to the customer record.

- **Sellers split almost evenly on what holds them back.** Trust, accuracy, and privacy concerns lead at 36%, just ahead of practical hurdles such as skills, cost, and setup. A third say they already use AI as much as they need, so no single fix reaches most sellers.
- **The barrier shifts with team size.** Trust is the main concern on small teams (41%), while solopreneurs more often point to practical hurdles (38%).
- **AI tool choice is often informal.** About a third of sellers pick their own AI tools with no approval process, and some use tools their company has not approved. Any AI built into the CRM starts alongside tools sellers already chose.
- **Unapproved AI shows up on teams far more than with owners.** Based on screener answers, about a quarter of team sellers use AI tools their company has not approved, compared with fewer than one in ten founders and owners, who mostly choose their own tools.

### Drafting and research edge out meeting capture as sellers’ main use of AI

| Response | Share |
|---|---|
| Drafting, research, content support | 45% |
| Meeting capture or broader support | 42% |
| No or limited AI use | 13% |

*Source: G2 AI Custom Research interviews for Pipedrive, 2026. Based on 113 sellers who answered this question.*

> “I use AI for prospecting, a lot, and then I do use AI to help me, like, write the follow-up email. You know, I will do a quick draft with the main points but AI helps me polish it and make it look a lot better.”
> — Solo seller

**What this means:** Start with the tasks sellers already trust AI with, such as drafting, research, and content support. For teams, where trust concerns lead, put accuracy, privacy, and review-before-save controls at the center of the product and the message. For solopreneurs, lower the setup and cost hurdles.

## Chapter 5: The last mile — Why do AI meeting notes still leave the CRM out of date?

**Short answer:** The CRM step is still manual. Because getting AI meeting notes into the CRM is still mostly a manual step. Among sellers who use AI meeting capture, 67% review, copy, and re-enter its output into the CRM themselves, and only 18% have automated that step.

- **The last mile is rarely automated.** Across all sellers, only 10% have automated CRM data entry with AI, so for nine in ten sellers, the record still depends on them.
- **Solopreneurs are far less likely to use an AI meeting note taker.** Among solopreneurs, 61% do not use AI meeting capture, compared with 36% of sellers on teams of 2 to 10. For solo sellers, the first step is capture itself.

### Only 18% of AI-capture users have automated the step into the CRM

| Response | Share |
|---|---|
| Copy into the CRM by hand | 67% |
| Automated AI-to-CRM workflow | 18% |
| No workflow for AI notes | 15% |

*Source: G2 AI Custom Research interviews for Pipedrive, 2026. Based on 61 sellers who answered this question.*

### 55% of sellers use AI meeting capture, and 45% do not

| Response | Share |
|---|---|
| Does not use AI meeting capture | 45% |
| Regular AI meeting capture | 29% |
| Occasional or selective use | 26% |

*Source: G2 AI Custom Research interviews for Pipedrive, 2026. Based on 112 sellers who answered this question.*

> “So nothing happens automatically, which is one of the problems. We effectively have to copy and paste all of those notes onto an individual client record, which obviously takes time rather than it just automatically update.”
> — Seller, larger sales organization

**What this means:** Plan AI around the seller’s path to an updated record. For the 45% who don’t use AI meeting capture yet, including most solopreneurs, start with simple capture. For the two-thirds of users who copy AI output into CRM fields by hand, offer structured updates they check and approve. Full automation comes later, once sellers see value from the first steps.

## Chapter 6: Human in the loop — How much should sellers trust AI to update the CRM on its own?

**Short answer:** With review, not on its own. Sellers want human-in-the-loop AI, and automation with controls is the opening. Asked about automatic CRM updates, 41% of sellers would accept them with review, audit trails, or field-level controls, while 43% reject updates without a review step. Solopreneurs are twice as open as teams to skipping review altogether.

- **Accuracy drives verification more than polish.** Half of sellers (52%) say AI makes factual, numerical, or context errors that need checking, while others mainly fix tone and wording. Fewer than one in five report few or no real errors, so review stays part of the job.
- **Full autonomy remains a minority view.** Only 15% of sellers are comfortable with unreviewed updates, even for low-risk fields, rising to 22% among solopreneurs.

### Only 8% of sellers are open to low-touch review or more AI autonomy

| Response | Share |
|---|---|
| High review, final sign-off | 72% |
| Tiered oversight | 20% |
| Open to more autonomy | 8% |

*Source: G2 AI Custom Research interviews for Pipedrive, 2026. Based on 113 sellers who answered this question.*

> “I think 90% of it would be correct. But just letting it freely do it without any oversight, it would make me nervous.”
> — Seller, small sales team

**What this means:** Build review-and-approve steps into AI sales workflows, in line with the human-oversight guidance in the NIST AI Risk Management Framework. Let sellers check and fix AI-generated notes, numbers, commitments, and CRM updates before anything gets saved or sent, with the source material visible. Treat sign-off as the feature that makes AI usable, the same boundary G2 found among enterprises scaling AI.

## Strategic patterns

### The automation gap sits between useful output and a finished workflow

Sellers already use AI for drafting, research, and content support. Meeting capture is uneven, and manual or fragmented recordkeeping after the call is still the most common approach. A seller gets a useful AI output and still has to update the customer record.

**Implication:** The conversation about AI-native CRMs should center on the work sellers still finish by hand: turning conversations into usable, reviewable records.

### Seller control sets the boundary for AI adoption

Sellers welcome AI support they see and check, and they want to keep the final decision. Everyday use of AI for support tasks goes hand in hand with a strong preference for human review and final sign-off.

**Implication:** Position seller-centric AI as support that gives time back to the seller and preserves their authority over customer information and next actions.

### Solopreneurs and small teams start from different places

Solopreneurs are three times as likely to run without a CRM and less likely to use AI meeting capture, and skills, cost, and setup hold them back most. Sellers on small teams more often have a CRM and capture tools in place, and trust in AI accuracy and privacy is their main concern.

**Implication:** Give solopreneurs a simple path into capture and a CRM. Give teams the accuracy, privacy, and review controls that earn their trust.

## Recommendations

1. **Focus the AI story on the complete recordkeeping workflow.** Frame seller-centric AI around the work that follows a customer conversation: turning useful capture into a usable, reviewable record. Among sellers who use AI meeting capture, 67% still do that step by hand.
2. **Make review and approval a core design principle.** Treat seller review as a condition of trust, especially for customer facts, commitments, and next steps. Offer AI assistance that is easy to inspect and correct, since 72% of sellers want final sign-off.
3. **Meet sellers where AI already works.** Start with drafting, research, and content support, where 45% of sellers already use AI most. Then fix the handoff from meeting notes to an accurate record.
4. **Give solopreneurs their own starting point.** Half of them have no dedicated CRM, and most don’t use AI meeting capture yet, so simple capture and an easy move from spreadsheets come first.

## Frequently asked questions

### How do solopreneurs keep up with CRM admin without a team?

Many work around the CRM. Half of solopreneurs have no dedicated CRM, and most don’t use AI meeting capture. Admin weighs on them a little less than on small teams: 49% say it displaces selling, compared with 62% of sellers on teams of 2 to 10.

### What is CRM data entry, and can AI automate it?

CRM data entry is the work of logging calls, notes, and next steps into the customer record. AI takes on part of it, and few sellers have automated it fully. Only 10% of sellers have AI updating the CRM automatically. Among sellers who use AI meeting capture, 67% still review and copy its output into the CRM by hand.

### What’s the biggest barrier stopping small sales teams from using AI more?

No single barrier dominates. Trust, accuracy, and privacy concerns lead at 36%, just ahead of practical hurdles such as skills, cost, and setup, and a third say they already use AI as much as they need. Trust leads on small teams, and practical hurdles lead among solopreneurs.

### What does AI get wrong in sales work?

Mostly facts. Just over half (52%) of sellers say AI makes factual, numerical, or context errors that need checking. Others mainly fix tone and wording, and fewer than one in five report few or no real errors.

### What is human-in-the-loop AI for sales teams?

Human-in-the-loop AI drafts the work, and a person reviews and approves it before anything gets sent or saved. Most sellers in this study ask for exactly that. Overall, 72% of sellers want high human review and final sign-off on AI-generated notes, summaries, emails, and research. Another 20% want lighter review for internal work and closer review for customer-facing work. The review step is how human-in-the-loop improves AI accuracy: 52% of sellers say AI makes factual, numerical, or context errors that a person needs to catch.

### Do salespeople trust AI to update their CRM without review?

Most don’t. Overall, 43% of sellers reject automatic CRM updates with no review step, and 41% would accept them only with review, audit trails, or field-level controls. Only 15% are comfortable with unreviewed updates, rising to 22% among solopreneurs.

### What is an AI-native CRM?

An AI-native CRM builds AI capture and record updates into the CRM itself, so the seller reviews the update in one place and skips copying output across from a separate tool. It is the difference between an AI CRM designed around the record and a CRM with AI features bolted on, and this study shows why the distinction matters. Among sellers who use AI meeting capture, 67% still re-enter its output into the CRM by hand, and only 18% have an automated AI-to-CRM workflow.

### How does this report define its key terms?

These are the terms as we use them. AI meeting capture means using AI to record, transcribe, or summarize sales calls, regularly or now and then. A dedicated CRM is a CRM or sales system the seller’s company uses as its main record of customers and deals. A fragmented setup has a CRM, but most work happens in other tools such as spreadsheets, email, or notes apps. A selling-led day is one where selling takes most of the time and admin stays limited. Automation with controls means AI updates the CRM only with review steps, audit trails, or limits on which fields it can change. The last mile is the step between an AI-generated summary and a CRM record the seller trusts.

### What are the limits of this data?

The findings describe 113 sellers, most of them solopreneurs or members of teams of up to 10, with 8% in larger organizations. Of the 91 we matched to screener data, 77 are US-based and 14 UK-based, so the sample leans heavily toward the US and we don’t compare markets. We report subgroup figures only for groups of at least 20 respondents.

## Methodology

### How was this research conducted?

This research draws on 116 AI-conducted voice interviews with 113 active sellers recruited in the US and UK, fielded in October 2026, most of them solopreneurs or members of sales teams of up to 10 people. Three sellers took the interview twice, and we kept each person’s first interview. Interviews ran about 10 to 34 minutes, with an average of 18, and let sellers describe their actual practices in their own words.

### Who we interviewed

Of the 91 sellers we matched to screener data, 77 are US-based and 14 UK-based, so the sample leans toward the US and we don’t compare the two markets. Most sell in B2B professional, technology, or industrial categories (60%), and most are solopreneurs (41%) or on teams of 2 to 10 (51%).

### How we analyzed it

Each percentage uses the sellers who answered that question as its base, shown under each chart. We exclude non-answers and responses the interview platform coded as off-topic. We report a group comparison only when both groups have at least 20 answers and differ by at least 10 points. Percentages are rounded to whole numbers, so some totals may not add up to exactly 100. Because this is an in-depth interview study rather than a large-scale survey, we show the count behind each headline figure so readers can see the base for every number. We analyzed all transcripts with AI, and G2’s AI Custom Research team reviewed each one independently.

**Limitations:** Findings describe the 113 sellers interviewed and are not a statistically representative estimate for all sellers. The sample is mostly US-based, so we report no US versus UK comparison. This is a single fielding wave, so the report shows no change over time.

---

*This research was commissioned by Pipedrive. G2 maintains full editorial independence.*

**Topics covered:** AI Adoption, AI Trust & Governance