McKinsey's September 2024 analysis of AI in B2B sales found that companies using generative AI to assist sales teams reported efficiency gains of 10 to 15 percent. That number is real and comes from a credible source, but it requires context to be useful: the gains were concentrated in specific tasks, not distributed evenly across the selling process, and they came from AI assisting human judgment rather than replacing it. The broader story about generative AI in B2B sales is more complicated, and the research from the last 18 months tells a clearer story than most vendor materials do.
What McKinsey's research actually says
In "An Unconstrained Future: How Generative AI Could Reshape B2B Sales" (September 2024), McKinsey identified three areas where generative AI in sales produces the clearest documented gains: automating sales research and call preparation, generating first drafts of outreach messages that reps then personalize, and producing follow-up summaries and CRM updates after customer interactions. The 10 to 15 percent efficiency figure applies to these use cases, where the AI handles time-consuming tasks that do not require deep human judgment, freeing reps to spend more time on conversations.
The same report projected that generative AI could eventually automate activities that currently consume up to 80 percent of a sales rep's time, including prospect research, qualification conversations, and routine follow-up. The important qualifier is "eventually." The current state of deployments in production does not look like the projected potential. Most sales teams using generative AI today are capturing efficiency in a subset of administrative and content tasks, not transforming the core selling motion.
Where the analyst data gets more cautious
Gartner's findings from 2025 provide important counterweight to the optimistic projections. In a June 2025 press release, Gartner predicted that over 40 percent of agentic AI projects would be canceled by the end of 2027, citing escalating costs, unclear business value, and inadequate risk controls. Gartner's analysis also found that only approximately 130 of the thousands of vendors marketing AI agent capabilities have real agentic AI functionality. The rest are automation tools with an AI label attached.
For sales teams evaluating generative AI tools specifically, this distinction matters. A tool that drafts a follow-up email based on call notes is generative AI applied to a useful task. A tool that claims to autonomously conduct sales conversations, qualify prospects, and book meetings without human involvement is an AI agent claim that requires significant scrutiny before deployment. The failure rate on the latter category is substantially higher, and the governance complexity is greater.
Gartner also predicted in a November 2025 press release that by 2028, AI agents will outnumber human sellers by 10 to 1, yet fewer than 40 percent of sellers will report that AI agents improved their productivity. That combination, ubiquitous AI adoption alongside majority dissatisfaction with outcomes, is a precise description of a market that has moved faster than the underlying technology's ability to deliver on its promises.
The use cases with the clearest evidence base
- Prospect research and contact enrichment. Generative AI applied to public company data, job posting analysis, news events, and social signals can reduce the time a rep spends building an account brief from hours to minutes. Tools like Clay use AI to synthesize data from multiple enrichment sources into structured research outputs. The value is not that the AI is better at research than a human; it is that the AI can do 50 accounts' worth of research in the time a rep would do 3. The quality ceiling is lower than human research, which matters more on enterprise deals than on SMB volume motions.
- First-draft outreach personalization. Generative AI can produce a first draft of a personalized outreach message faster than a rep can write it from scratch. Tools like Outreach and Salesloft have integrated AI-assisted writing into their sequence builders. The documentation from rep usage suggests that AI-drafted messages require significant editing before they read as genuinely personalized, particularly for senior buyers who are familiar with AI-generated copy patterns. But even the draft-and-edit workflow is faster than starting from a blank page, which is where the efficiency gain materializes in practice.
- Post-call summaries and CRM updates. CRM hygiene is a chronic problem in B2B sales, and AI-generated call summaries have shown consistent adoption because they address a task that reps universally dislike and routinely skip. Platforms like Gong automatically generate structured summaries from call recordings, including next steps, objections raised, and action items, which then populate the CRM record. The accuracy is high enough for most sales motions, and the adoption rate among reps is better than any other AI feature category because the workflow benefit is immediately obvious.
- Content personalization for mid-funnel assets. Generating customized case studies, proposal sections, and comparison documents tailored to a specific prospect's industry and pain points was previously limited by the time cost of customization. Generative AI reduces that cost significantly. For inside sales teams working deals at volume, this means a level of personalization that was previously only possible at the enterprise level, where dedicated sales engineers could build custom materials for each account.
The governance risk Forrester identified
Forrester's 2026 B2B predictions report warned that B2B companies will lose more than $10 billion in enterprise value because of ungoverned generative AI use, through regulatory penalties, legal settlements, and brand damage. The specific risk in sales is AI-generated content that contains inaccurate claims about product capabilities, competitors, or customer results. Generative AI producing sales materials at scale can propagate errors at a scale that manual review cannot catch if the governance process is not designed correctly.
This is particularly relevant for teams using AI to generate proposal content, competitive comparisons, or technical documentation. The efficiency gains from generative AI in these contexts are real, but they require review processes that are often not in place when teams first deploy the tools. The Forrester prediction is not a reason to avoid generative AI in sales; it is a reason to plan the governance infrastructure before turning on the content generation.
What to prioritize when evaluating tools
The research supports a clear prioritization framework for generative AI investments in B2B sales. Start with use cases where the AI handles rote tasks that reps currently skip or do poorly: CRM updates, call summaries, contact research. The quality bar for these tasks is achievable with current technology, and the workflow adoption is higher because reps see the benefit immediately rather than needing to trust a new process.
The more ambitious applications, autonomous prospecting, AI-driven qualification conversations, and fully automated follow-up sequences, have longer time horizons before the success rates are reliable enough to deploy at scale without significant oversight. The Gartner finding on agentic AI cancellation rates applies most directly to these deployments. Teams that jump to full automation before establishing strong performance on assisted tasks are building on an unstable foundation.
For teams currently evaluating data enrichment as the foundation for any generative AI-assisted outreach, the comparison between Clay and Apollo is worth understanding before committing to a stack, since the quality of the enrichment data determines the quality of the AI-generated output.
For more on how AI sales automation tools compare in production, see our analysis of the best AI sales automation tools based on review data.