Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
In August 2024, Salesforce CEO Marc Benioff said many customers were disappointed with Microsoft Copilot’s answer accuracy and quality. Microsoft executive Jared Spataro countered that customer numbers and daily use were rising. The exchange was a clash of vendor claims—not independent proof that either product was better. It also came as Salesforce prepared to launch Agentforce, its competing pitch for AI agents in business workflows.
What Benioff said about Microsoft Copilot
Benioff made the criticism during Salesforce’s fiscal second-quarter earnings call on August 28, 2024. Salesforce’s quarter had ended July 31. He said “so many customers” were disappointed with what they had bought from Microsoft Copilot, pointing to accuracy and response quality as problems. CRN’s account of the exchange records the complaint; Salesforce’s earnings release establishes the reporting period.
His argument was broader than a critique of one Microsoft product. Benioff objected to the idea that businesses should build, train, or retrain their own AI models, and argued that Microsoft had emphasized models rather than a complete enterprise platform. He contrasted conversational copilots—which assist when prompted—with agents he said could plan, reason, and take action with less human input. He presented Salesforce’s platform, customer data, and workflow tools as a simpler and less expensive route.
Free tools Windows power users keep installed
One-click scans. No signup required.
Those were Salesforce’s competitive claims, not neutral findings about Microsoft’s technology or the relative cost of the two approaches. Benioff’s remarks also arrived just ahead of Agentforce’s planned launch, giving Salesforce a clear commercial reason to distinguish its product from copilots.
#1 Best Overall
Microsoft’s response: adoption was growing
Microsoft executive Jared Spataro said the company was hearing “something quite different” from Copilot for Microsoft 365 customers. He cited a more-than-60% increase in customers during the prior quarter and said daily users had more than doubled. He also pointed to conversations with CIOs and third-party data as signs that organizations were choosing Microsoft for AI transformation, according to CRN.
Those figures support a claim of commercial momentum, but they do not answer Benioff’s narrower criticism. More customers or more daily users do not, by themselves, show that answers were accurate, that customers were satisfied, or that deployments delivered a satisfactory return on investment. Microsoft did not concede the quality complaint, and the exchange supplied no neutral head-to-head accuracy or ROI study.
Rank #2
These were not simply two versions of the same product
“Microsoft AI” covers several products and services. The product most directly at issue was Microsoft Copilot for Microsoft 365, an assistant integrated with apps such as Word, Excel, PowerPoint, Outlook, and Teams. Copilot Studio provides tools for extending Copilot and building custom agents, while Azure AI is a broader cloud platform for developing AI applications. Microsoft’s later materials describe Copilot expanding beyond assistant-style interactions toward custom agents and workflow automation; see its FY25 second-quarter earnings materials and Copilot Studio licensing guide.
Salesforce Agentforce was positioned as a platform for autonomous agents in CRM-related work, including sales and customer service. It sits within a wider Salesforce environment that includes Einstein AI features and Data Cloud, alongside Salesforce’s customer and business applications. In practical terms, Microsoft’s pitch centered on assistance across a broad productivity and collaboration suite; Salesforce’s centered on agents grounded in CRM data and customer processes. The products can overlap, but they are not interchangeable in every job.
Rank #3
Agentforce was central to Salesforce’s argument
Salesforce expected Agentforce to become generally available in October 2024; later coverage reported a launch on October 24. Benioff said Salesforce expected thousands of customers to be using it by the beginning of the next fiscal year and floated an ambition of one billion agents by fiscal 2026. These were forecasts and ambitions, not verified outcomes. Salesforce was framing enterprise AI as “digital labor” capable of completing tasks, rather than only generating answers.
Salesforce also pointed to early customer examples including ADP, OpenTable, and Wyndham. In a Fortune report, Benioff claimed Agentforce was resolving more than 90% of ADP customer-service inquiries. That should be read as an executive’s account of a particular customer deployment—not an independently audited benchmark or a rate established across Agentforce customers.
Rank #4
What the evidence does—and does not—show
The available reporting establishes that Benioff made a forceful criticism and Microsoft responded with adoption figures. It also shows two companies competing to define enterprise AI: Microsoft could lean on its broad Microsoft 365 distribution, while Salesforce argued that CRM context and autonomous agents offered a different kind of value. It does not establish that Copilot was categorically inaccurate, that Agentforce was more accurate, or that either product delivered better ROI across customers.
Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →Later disclosures show Microsoft continued to gain traction. In its FY26 first-quarter materials, Microsoft said tens of millions of users were using Copilot Chat and more than 90% of Fortune 500 companies used Microsoft 365 Copilot. In its FY26 second-quarter materials, it reported 15 million paid Microsoft 365 Copilot seats and daily active users up tenfold year over year. These are Microsoft-reported adoption measures, not satisfaction scores; they cannot retrospectively settle whether some customers were disappointed in 2024.
Best Value
How a business should compare them
The useful question is not which CEO won the exchange. It is which product fits a specific workflow, data environment, and budget—and whether a pilot proves it works. A sensible comparison starts with:
- Existing systems: Microsoft Copilot may be a more natural fit where employees already work in Microsoft 365, Teams, SharePoint, Outlook, or Azure. Agentforce may fit better when customer records and service or sales processes already live in Salesforce.
- Job to be done: Test document, meeting, email, and collaboration assistance separately from CRM tasks such as handling service inquiries or updating sales workflows. Do not treat a general assistant and a task-specific agent as equivalent.
- Data and permissions: Neither tool becomes reliable simply by connecting it to company data. Check source quality, access controls, retrieval and configuration, and whether users can see only information they are authorized to access.
- Autonomy and safeguards: An agent that takes action needs tighter testing, approvals, audit trails, permissions, and a way to reverse mistakes than a tool that only drafts text. Start with human review for consequential actions.
- Outcomes and cost: Measure answer accuracy against a reviewed set of real tasks, completion and escalation rates, time saved, cost per completed task, and security incidents. Seat counts, prompts, and conversations measure activity—not business value.
- Total cost: Microsoft 365 Copilot has been sold primarily on a per-seat basis, while Salesforce described Agentforce with consumption-oriented pricing. A 2024 report cited approximately $30 per Microsoft 365 Copilot user per month and about $2 per Agentforce conversation, but those are historical price signals, not verified August 2026 terms. Packaging and metering can change, and custom-agent tools may involve separate capacity or usage costs.
That last distinction matters for budgeting: per-seat pricing can be easier to forecast but may leave a business paying for lightly used seats. Usage pricing can tie spend to automated work, yet costs may be harder to predict if conversations or agent actions grow unexpectedly. The right comparison is a like-for-like workload using current quotes, not a historical headline price.
Common deployment failures also cut across vendors: inaccurate or unsupported answers from weak source data; low adoption because staff do not know when to use the tool; pilots that fail on unusual cases at production scale; automating a flawed process; and measuring activity rather than completed work. A disappointing result may reflect the model, but it may also reveal a data-governance, workflow, or change-management problem.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.

