Gartner's latest Magic Quadrant for CRM Sales Platforms arrives as AI reshapes what vendors think a CRM should actually do. The category grew 12.1% in 2025 to $16.9 billion, and vendors have increasingly made agentic AI central to their product strategies. Gartner has already warned that AI agents could reshape enterprise software more broadly, and the CRM sales market looks like an early test case for that shift.

That makes the Magic Quadrant (MQ) more interesting than a simple ranking of vendors. The Leaders and Challengers may be separated by Gartner's broader criteria, but the report points to agent autonomy as an increasingly important dividing line between the vendors best positioned for what's coming and those still relying heavily on predefined workflows. Gartner's assessment suggests the market is moving in that direction, but most vendors still have some way to go.

Microsoft and Salesforce are in the much coveted “Leaders” quadrant. HubSpot, Oracle, Pega and Zoho are Challengers. Creatio is the sole Visionary. BUSINESSNEXT, monday.com, Neocrm, SAP, SugarAI and Vtiger are positioned as Niche Players.

The rankings demonstrate Gartner's broader assessment of “Ability to Execute” and “Completeness of Vision”, not autonomy alone. But AI maturity is increasingly central to the evaluation, making the positioning a useful guide to where vendors sit in the shift towards agentic CRM.

Salesforce's position reflect the breadth of Agentforce, including configurable agents and subagents, alongside predictive scoring, conversation intelligence, and governance and observability capabilities. Gartner also warns that its more advanced agentic workflows still demand considerable composition work, and recommends buyers test confidence handling and error recovery directly rather than assume end-to-end autonomy.

Microsoft's placement rests on Dynamics 365, Microsoft 365 Copilot, its sales agent and Copilot Studio working across applications. Gartner sees that cross-application reach as a strength, but also flags fragmented AI experiences, latency and the technical expertise sophisticated custom agents can still require.

Further down the MQ, Gartner's reservations about current agentic capabilities become more apparent. It says HubSpot's Breeze agents offer a strong seller experience but remain constrained in how much autonomous orchestration they can deliver at this stage.

Being named a leader does not mean either Salesforce or Microsoft has solved autonomous AI. It means they currently combine the capabilities, market position and execution Gartner rates most strongly in this category.

The most revealing part of the report sits underneath the quadrant. Vendors are presenting agentic AI as the next stage of CRM, but Gartner finds that much of the underlying technology remains based on LLM-augmented workflow chains, including predefined graphs, deterministic triggers and administrator-authored instructions. Even where platforms use supervisor agents and subagents, the underlying paths and tools can still depend heavily on human configuration.

That is a meaningful difference from a system that is given a goal, works out how to achieve it, chooses its own tools and adapts as circumstances change. Gartner suggests broadly reliable autonomous agentic sales capability is more likely to emerge after 2026. It points buyers past the polished demos towards the foundations that would actually support it, such as planner runtimes capable of working out how to reach a goal rather than following a fixed path, persistent memory that carries context across a relationship rather than resetting each session, and dynamic orchestration that coordinates multiple AI capabilities as context shifts.

This changes the buying question. It is no longer enough to ask which vendors have launched agents. A better question is what those agents can actually do without someone having designed the path for them first. Similarly, for the wider CX community, the promise of autonomous customer experience depends on agents working across context, systems and interactions rather than simply executing isolated automations.

Autonomy is only useful if an agent has enough context to make a good decision, and Gartner identifies context federation as an emerging architectural battleground. Zero-copy access to platforms such as Snowflake and Databricks, allowing AI systems to access data without duplicating it, is emerging as one marker of maturity, although Gartner says only a handful of vendors currently demonstrate production-grade capabilities of this kind.

The problem goes beyond where data is stored. Memory is inconsistent across the market too. Some platforms lean on isolated session histories or basic CRM fields, leaving sellers or administrators to reconstruct relationship context by hand. An agent dealing with a customer should ideally know what has happened before, what has been promised, and what other systems say about the relationship. Scatter that across CRM records, contact-centre systems and data warehouses, and a more capable model alone doesn't fix it.

There's a constraint that's easy to overlook while vendors demonstrate increasingly sophisticated agents: what it actually costs to run them at scale. Gartner says AI value is increasingly bound up in packaging, data readiness and consumption economics, with agent actions, retrieval calls and workflow execution often metered separately across multiple product layers. Salesforce is a useful example here. Gartner notes that buyers need to weigh dependencies across Agentforce 1, Data 360, Tableau and Slack alongside consumption-based flex credits.

An agent can work well in a demo and still be difficult to reproduce across thousands of users without the right data, permissions, integrations and commercial model behind it. Increasingly, the AI question is an operational one as much as a technical one: can the organisation afford to run the agent it has been shown?

Another shift in the report could prove just as significant as the move towards autonomous agents. Gartner identifies three different models emerging in the market: AI embedded across the applications where sellers already work, AI presented through an assistant sidebar inside the CRM, and the traditional CRM with AI features bolted on. Gartner says the third model is losing ground.

That helps explain Microsoft's position in the MQ, since its sales agents can surface within the Microsoft 365 ecosystem rather than pulling sellers back into a CRM screen for every task. The broader implication is that the CRM may increasingly become the underlying source of customer context and intelligence rather than the place where all sales work has to happen. This extends to service and marketing, where users will increasingly expect AI to bring customer context into the tools where they already work.

The CRM market is being pulled in several directions at once. Agents are getting more capable, but much of the market still runs on predefined workflows. Context is becoming more important, but enterprise data remains fragmented. AI is spreading into more applications, but the economics of running it at scale are harder to predict than a demo suggests. And the vendors that look strongest today aren't necessarily the ones with the longest feature lists.

Gartner warns that decisions made now could shape technology architecture for three to five years. For CX leaders, that makes the MQ more than a vendor ranking. It's an early read on how enterprise software is trying to move from AI that assists people towards AI that can act on their behalf.

The practical test isn't whether a platform has an AI agent. It's whether that agent has the context, memory, orchestration and governance needed to become genuinely autonomous. The evidence is increasingly pointing towards production just being the start.