AI in the CRM: What Agentforce, Breeze and Joule deliver in sales

AI This article was created with the help of AI.

Key takeaways

• AI in the CRM summarizes, makes predictions, and acts as an autonomous agent.
• Agentforce, Breeze and Joule offer specific sales functions (as of 2024/2025).
• 76% of CRM users have patchy data, which blocks AI adoption.
• A capture assistant like Bliro solves the data problem in field sales.

What does AI in the CRM actually mean?

AI in the CRM is not a single feature. Behind the buzzword hide three classes of functionality that intervene in your sales operation to very different depths. If you keep the three apart, you can realistically assess what your CRM platform can do today and where its limits lie.

1. Summarizing: The AI creates meeting notes, account summaries, or email drafts from data that is already in the system.
2. Predicting: Predictive AI scores leads, estimates close probabilities, and shows you which deals are stalling before the quarter-end close reveals it.
3. Acting: Agents carry out tasks autonomously. They qualify leads, update records, or book appointments without a rep clicking a thing.

Agents: from answer to action

The third class is the most demanding, and Salesforce shows what this means: the Einstein SDR Agent, today Agentforce SDR, autonomously reaches out to inbound leads, answers questions in natural-language dialogue, handles objections, and books appointments for your reps. The Sales Coach Agent, meanwhile, simulates a buyer so reps can practice their pitch and objection handling before the real meeting. According to Salesforce, both agents are connected to the company's CRM data so that every answer stays in the context of the deal, the account, and previous correspondence.^[1]

All three classes have one thing in common: they work with what's in the CRM. An agent that acts autonomously needs reliable records. That's exactly where the data question becomes the bottleneck - and faster than many CRM roadmaps suggest.

What do Agentforce, Breeze and Joule offer in sales?

The major CRM vendors have built their AI directly into the platform. The feature sets differ, but the pattern is the same: the AI sits inside the CRM and accesses CRM data. The following overview summarizes the vendors' statements as of the date of each source, without evaluation.

‍

Microsoft Dynamics 365 also builds AI assistance directly into the system along this pattern. Important for context: these are vendor statements as of the time of each source, not a ranking and not a purchase recommendation. What matters is the common thread. Joule updates what the rep tells it to. The Prospecting Agent researches what sits in HubSpot. The AI in the CRM knows only the record - never the conversation that happened at the customer's site yesterday.

Why does CRM AI need complete data?

What that means in practice: a summary based on patchy fields is a summary of the gaps. A close probability based on outdated opportunities is a forecast of a pipeline that may no longer exist. And an agent that acts on its own multiplies the errors baked into the data.

How thin that foundation often is in practice is shown by a Validity study of 602 CRM users and decision-makers: 76% state that less than half of their company's CRM data is correct and complete. 37% report having already lost revenue directly as a result of poor data quality.^[4] For you as a CRM manager, that is not a statistical footnote - it is a description of your daily work: dashboards that reflect your reps' memory instead of reality.

When field sales doesn't enter data, everything is missing

In the office, records get created at the desk. In field sales, they rarely do. The meeting ends at 4 pm, the next one starts at 4:15, and the visit report lands on the 'later' list. Typing it up at the kitchen table in the evening never happens. You know the result from every forecast meeting.

• Forecast rounds are negotiated over recollections instead of facts.
• Agents like the SDR Agent qualify leads based on patchy account data.
• Coaching and pipeline analyses rely on spot checks instead of complete conversation records.

Scale up the missing reports, and your team loses hours every week on admin work in the sales routine that could be converted into selling time. You can work out what that amounts to for your team yourself with the interactive ROI calculator at www.getbliro.io, from capture to reporting.

When is the built-in AI enough?

So the built-in AI is not a bad buy. It goes a long way when three conditions come together. Check them for your team before you think about adding anything.

• Your team works in the office or runs its meetings online via Microsoft Teams, Zoom, or Google Meet.
• The data in your CRM is solid: accounts, contacts, and opportunities are maintained directly at the desk while the conversation is happening.
• The built-in features cover your use cases, such as inbound lead qualification, or notes and field updates on demand like Joule in SAP Sales Cloud.^[5]

In this setup, the AI works exactly where the data is created. The rep sits at the desk, the CRM is open, and updating it costs no extra step. Data quality holds because data entry has no barrier. It comes down to this: built-in AI carries you as far as the barrier to data entry is low.

So the sticking point is not the AI's intelligence, but the rep's situation. As soon as meetings happen outside the office, that chain breaks - and with it the foundation for each of the three feature classes.

When do you need a capture assistant in front of it?

Field sales is different. Your rep is standing in the customer's parking lot at 5 pm, the next meeting is waiting, the laptop stays in the trunk. This is exactly the moment that decides whether your CRM is complete - or whether the 37% revenue loss from the Validity study hits your team too.^[4] Typing on the phone in the car is not an option, and 'later at the office' means, in practice: never.

That is where you need a capture assistant in front of the CRM that takes the data where it is created: by voice. The rep speaks into the hands-free system after the meeting, and the report lands in the system in a structured form. Three ways to get field sales notes into the CRM can be compared cleanly, from the dictation recorder to the Voice-to-CRM assistant that writes directly into the records.

• Visit report by voice after every meeting, with no typing and no delay.
• Create or update an opportunity, directly from the conversation context.
• Fill mandatory fields consistently so records don't slip through the system incomplete.
• Duplicate checks so that every visit doesn't create a new contact and hurt your data quality again.

That's exactly what the Bliro Sales Assistant is built for. It documents conversations onsite and online in real time, GDPR-compliant, without recording and without a bot, proactively fills CRM fields from the transcript, and takes visit reports by voice call while the rep drives away from the appointment. How such automated reports change everyday sales work is covered in a dedicated practical piece on automating visit reports. The built-in AI of Agentforce, Breeze and Joule then works with a data foundation that is actually complete. If you want to see what this means for CRM coverage and admin work in your team, find out more at bliro.io.

Sources

1. https://www.salesforce.com/news/stories/einstein-sales-agents-announcement/
2. https://www.hubspot.com/company-news/spring-2025-spotlight-breeze-agents
3. https://learn.microsoft.com/en-us/dynamics365/sales/use-sales-copilot
4. https://validity.com/resource-center/the-state-of-crm-data-management-in-2025
5. https://news.sap.com/2025/02/sap-sales-cloud-joule-intelligent-selling-now-available/