Sales Force Automation is gone as an analyst category. What replaced it is scored on something most RFPs have never asked a vendor to demonstrate: whether you can see what the agent inside the CRM actually did.
On 12 August, Gartner published the first Magic Quadrant for CRM Sales Platforms, retiring the Sales Force Automation Platforms name it had used for years. Vendors put out the usual placement announcements within a day. Microsoft noted a sixteenth consecutive year as a Leader; Salesforce joined it in the Leaders quadrant. Oracle, Zoho, HubSpot and Pega were placed as Challengers, Creatio as a Visionary, and a longer tail — BUSINESSNEXT, SAP, Neocrm, monday.com, Vtiger, SugarAI — as Niche Players.
A category rename reads as housekeeping. It is not. The name of an analyst category is a summary of its evaluation criteria, and when the name changes, the criteria changed first.
What "automation" meant, and why it stopped working
Sales Force Automation described a coherent idea: the software takes deterministic work off a rep. Route the lead, log the activity, advance the stage, generate the quote. Every one of those is a rule someone configured, and evaluating a vendor on them is straightforward. Does it do the thing, how much configuration does the thing require, and what does it cost per seat to do it.
That framework assumes the system only ever does what it was told. Agentic features break the assumption. When a CRM contains something that decides which accounts to work, drafts the outreach, updates a record on its own reading of a call, or advances a stage because it inferred that a stage-gate was met, "does it automate this task" stops being the question. The question becomes what it did, on what basis, and whether anyone can reconstruct that afterwards.
The renamed report reflects that. Its stated weighting covers how vendors embed predictive, generative and agentic AI into sales workflows with governance oversight, and it assesses visibility into agent actions rather than basic automation coverage. That second clause is the one that should change your scorecard.
The old category asked whether the system could do the work. The new one asks whether you can audit it having done so.
Why this is a systems decision, not a feature comparison
Agent visibility is not a feature that appears in a demo. It is a property of the platform's data model, and it either exists or it was never built.
Consider what has to be true for a RevOps owner to answer a simple question — why did this opportunity move to Commit last Tuesday? If a human moved it, the audit trail names the human. If an agent moved it, the audit trail needs to record that an agent acted, which agent, under whose authority, on what inputs, and with what confidence. Most CRM audit logs were designed to record user actions. An agent operating through an integration user shows up as a user, and the entire distinction collapses into a single row that says the record was updated.
The consequence is not philosophical. It shows up in the forecast, where an inflation caused by agent-advanced stages is indistinguishable from real movement; in compensation disputes, where attribution to a rep or an agent decides a payout; and in territory design, where an agent's account-selection behaviour quietly becomes the coverage model nobody signed off.
Figure 1
Evaluation dimension | The SFA-era question | The question the new category implies | Where the answer lives |
|---|---|---|---|
Record changes | Can the system update the record automatically? | Can you distinguish an agent-originated change from a user-originated one in the audit log? | Platform audit model — not configurable after purchase |
Agent authority | Which automations are switched on? | What is each agent permitted to do without a human confirming, and who set that boundary? | Permissions and governance layer |
Inputs and reasoning | What triggered the workflow? | What data did the agent read, and is that retained long enough to review a disputed action? | Retention configuration and logging depth |
Forecast integrity | How accurate is the predictive score? | Can you exclude agent-advanced stage changes from a forecast roll-up? | Reporting model |
Commercial model | Cost per seat | Cost per agent action, and who controls the consumption ceiling | Contract, not product |
The scorecard shift. Column two is what most existing RFP templates still ask. Column three is what the renamed category is scored on.
What to do with an RFP written for the old category
Most sales-systems RFP templates in circulation were written between 2021 and 2024 and inherit the automation framing. Three additions carry most of the value.
Ask for an audit-log export, not a demo. Request a sample export covering a week of agent activity in a reference environment, and check whether agent and human actions are separable in the raw data. A vendor that can produce it in a day has the model. A vendor that offers to walk you through the interface instead does not.
Put the consumption ceiling in the commercial section. Where agent activity is metered, the operative risk is not the unit price but who can raise usage. If a sales manager can enable an agent across a team without a systems owner approving it, the ceiling is nominal.
Score the ability to switch an agent off cleanly. Reversibility is the cheapest form of governance and the one least often tested before purchase. If disabling an agent leaves the records it created in an indeterminate state, that is a data-model answer, and it is the same answer as the audit question.
The wider signal
Analyst categories are lagging indicators — they get renamed once buying behaviour has already shifted enough that the old name misdescribes what buyers are asking about. The useful information in this rename is not the quadrant placements. It is confirmation that enough enterprise buyers are now asking governance questions during CRM procurement that the benchmark had to be rewritten around them.
If your next renewal or replacement is being evaluated on automation coverage and cost per seat, you are running last cycle's scorecard against this cycle's products.
Sources and notes. Gartner Magic Quadrant for CRM Sales Platforms, published 12 August 2026, replacing the Magic Quadrant for Sales Force Automation Platforms. Placements and criteria described here are drawn from vendor announcements (Microsoft, 11 August 2026) and trade coverage (CX Today); SalesHubMedia has not reviewed the full report, which is licensed, and vendor placement announcements are inherently selective. Gartner does not endorse any vendor and its research should not be construed as a recommendation. Journalism, not procurement advice. Corrections welcome.
