I spent about ten years in revenue operations, and the complaint I heard most from RevOps leaders barely changed in that time: reps aren’t filling in the CRM the way the process says they should. Fields sit empty, stages are out of date, and the handoff from BDR to AE, or from sale to implementation, arrives with half the story missing.
AI was supposed to fix this. In a lot of the teams I work with now, it hasn’t, and the reason is surprisingly simple: we bolted AI onto the old process instead of replacing it.
Why CRM data matters more than it gets credit for
Patchy CRM data isn’t a tidiness problem. It’s the raw material for nearly everything RevOps is asked to run. You need it to recognize revenue and to quote properly. You need it so an account moves cleanly from your BDR team to your AEs, and so your implementation team knows enough to get a new customer to value quickly. And you need it to answer the question every leadership team keeps asking: why are we winning, and why are we losing?
When reps skip fields, every one of those things breaks a little, usually quietly, and usually right before a board meeting.
The two ways teams use AI on CRM data, and why both stall
1. Using AI to chase reps. Some teams point AI at the problem by getting it to remind or pressure reps into completing their fields. I think that’s a misaligned use case. If the AI already knows what the field should say, why is a person still typing it in?
2. AI writes, a human checks every value. Other teams let AI draft the update and then require someone to review each output before it lands in the CRM. It feels responsible, and I understand the instinct. But if a human has to read the output, approve it, and then put it in the CRM, you haven’t removed a single step from the process. You’ve just given the step a new name.
In both cases the rep is still doing admin, the RevOps team is still chasing, and the business still doesn’t trust the data.
What to do instead
The fix isn’t to remove humans from the process for the sake of it. It’s to put the effort in the right place.
- Invest in the prompt, not the review. Most of the work belongs upfront: what each field means, where the answer should come from (which call, which stage of the deal), and what a good output actually looks like. When the instructions are that clear, the outputs become consistent enough that nobody needs to sign off every single one.
- Keep reps out of routine review. The goal is that reps aren’t manually checking everything that goes into the CRM. Save human attention for the decisions that genuinely need judgment.
- Automate the handoff, not just the field. Populating fields is half the job. The other half is getting that information to the next person: an agent that sends the AE or AM everything they need once an account is qualified, and one that gives implementation the details they need post sale so the customer reaches value sooner. We cover that second part in more depth in our guide to the sales to post-sales handoff.
The bigger question underneath
Once AI can fill in every field, it’s worth stepping back and asking whether every field needs to exist. A lot of RevOps teams are working hard to get every piece of information into the CRM, by hand or by AI, without asking whether the CRM is the right home for it. That question deserves its own piece, and it starts with deciding who owns CRM data quality in the first place.
See it in action
Airspeed captures the details your handoffs and analysis depend on straight from customer conversations and puts them where your team needs them, so reps can get back to selling. See how it writes those fields on our CRM automation page, or book a demo.
Frequently asked questions
What does human in the loop mean for CRM automation?
It means a person reviews or approves what AI writes before it is saved to the CRM. It adds a safety check, but if it applies to every field, it keeps the same amount of manual work the automation was meant to remove.
Should AI update the CRM without human review?
For routine fields, it can, as long as the prompts are specific about what each field means and where the answer comes from. Human review is better spent on high judgment decisions than on every value.
Why don't sales reps update the CRM?
Because it takes time away from selling and rarely helps them directly. That's why asking AI to nag reps rarely works, while having AI capture the information from their conversations does.
What CRM data matters most for handoffs?
The qualification details an AE or AM needs to pick up an account, and the implementation details the post sale team needs to get a customer to value quickly.
