Most CRM complaints are time complaints. AI can give some of that time back, once you know which part of it reads your data and which part changes it.
Your sales team asked for a better CRM. What they actually described was less typing. Those are two different requests, and solving the first one when the problem is the second is how a company ends up with an eighth tool nobody opens.
Start With the Week, Not the Shortlist
Every few years the complaints stack up. The pipeline is behind, the reports do not reconcile, reps are not updating records. Sometimes the software really is the problem. More often it is fine and the work around it was never designed. So before you shortlist anything, get honest about what a rep’s week looks like.
Where the Sales Week Goes
Sellers spend 40% of the week on selling activities, with the rest splitting across prospecting (18%), data entry (17%), planning (13%), and training (11%). Data entry is the line a new CRM rarely moves. Different screens, same typing. The version worth buying is the one that does part of the keying with your rep.
What Connected Data Makes Possible
The same survey found sales teams running an average of eight standalone tools, with only 33% working from a single platform, and 42% of reps saying the number of tools is a lot to manage.
That matters more this year, because 51% of sales leaders in the same study said disconnected systems are slowing their AI plans. An assistant is only as good as the data it can reach. Consolidate the customer record and the same assistant gets materially better answers, which makes consolidation an AI decision rather than an IT tidiness one. That reframe helps when you ask finance for the budget.
Buyers Still Want a Person in the Room
It is tempting to read all of this as a case for taking sellers out of the loop. The buyer data points the other way: 69% of B2B buyers said they prefer to validate AI-generated insights with a sales rep at critical decision points, according to . Buyers use AI to build the shortlist. They still want a person to confirm it before they sign. So the goal is not fewer reps. It is reps with more hours for the conversations that close, which is a far easier business case to make internally.
"AI in the CRM" Describes Three Different Things
This is the distinction that will save you an evaluation cycle. Ask every vendor to separate them, in plain language, on the call.
The assist layer. It works next to a rep on the record in front of them: summarizes the account, drafts the follow-up, preps them for the meeting. It suggests and writes text, but it does not change your data.
The analysis layer. You ask a question in normal language, such as which open deals have gone quiet, and it reads across your records and answers, often handing you a document or a deck. The good versions read beyond the CRM, into email and files. It still does not change your data.
The agent layer. This one acts. It creates and updates records and runs a process end to end without a person driving each step. Microsoft, for one, : its assistant "helps you work faster," while its agents "autonomously perform tasks" and create, update and manage records. Any credible version comes with supervision: a log of what it did and a way to intervene.
All three are useful, and they are licensed, configured and adopted differently. A demo that blends them makes it hard to tell what you are buying, and the gap shows up later as "we thought it would just update the CRM." Ask which layer you are watching, every time.
Why the Now is the Best Timing
If you looked at AI in the CRM eighteen months ago and were underwhelmed, that is a fair memory of where things stood. It is worth a fresh look. Microsoft is a useful marker for how fast this moved: its agentic only reached general availability in June 2026, and Microsoft now documents alone, covering lead qualification, opportunity research, and close. Whichever vendor you land on, that pace has moved the whole category, so the comparisons you ran last year are out of date.
What to Ask Before You Sign
- Which of the three layers am I looking at right now?
- If this one acts on my data, what does the audit trail look like and how do I intervene?
- What has to be true in our environment before a rep can use this: which licenses, which connectors, which settings?
- Does it respect the permissions our people already have, or does it see everything in the database?
- What does this cost when forty reps use it daily, not when three of them pilot it?
- Who is accountable for adoption ninety days after go-live?
The last one decides whether any of this pays for itself. Capability ships on its own. Adoption gets planned.
Seeing It Instead of Reading About It
Microsoft and Western Computer are running a 30-minute session on Tuesday, September 15 at 12 pm ET / 9 am PT: . It is a live look at both layers running inside a working CRM.
Western Computer has spent nearly 40 years implementing Microsoft Dynamics, with 1,750+ implementations behind us. If Microsoft is not on your shortlist yet, the session still gives you a concrete standard to hold every other vendor to.
on LinkedIn.
