Ten years have passed since artificial intelligence (AI) first appeared in sales technology, and the results are mixed. Early tools applied rudimentary machine learning (ML) models to customer relationship management (CRM) exports, assigning win probability scores or advising on the “ideal” time to call. The mathematics was sound, the demos impressive, yet adoption faltered because little thought was given as to how sellers should use this information.
Subsequent products tried to be prescriptive rather than predictive. They looked at the deal progression, compared it with past successful opportunities and issued “next best actions”: update the legal clause, add a stakeholder, send a pricing sheet.
Today the marketing phrase and technological direction is “agentic AI.” Software providers promise autonomous digital processes, or agents, that can run pre-meeting research, draft follow-up notes and perhaps even advance a deal while you sleep. Some of that future will materialize. Repetitive tasks such as compiling account reviews or scheduling demos are obvious targets for automation. Yet the success of any agent, no matter how sophisticated, depends on the depth and accuracy of the information it ingests. If we do not capture the essence of each buyer–seller interaction, the most elegant agent will repeat the failures of previous incarnations—delivering impressive demos but failing to improve sales and revenue win rates.
Selling is ultimately an information alignment exercise. A buyer’s needs must intersect with a seller’s solution, and any misalignment is rarely the fault of the product. More often, the two parties never reach a shared definition of value. That definition resides in the full record of their conversation: objections voiced, targeted outcomes quoted, success criteria refined, issues with price, competitors’ positioning. Unfortunately, relying on the manual entry of this type of data is a fool's errand. And even when sales engagement does capture part of the story, incomplete information is actually worse than no data as it causes models to offer up misleading insights and recommendations, leading to eventual distrust from the sales team.
One corrective is to instrument the conversation itself. Modern speech to text engines, natural language processing (NLP), and large language models (LLMs) can transcribe calls, tag speakers, extract intent and map each utterance to the milestones that genuinely move a deal forward. Once that data is captured, an AI system can push context-aware guidance directly into the tools a salesperson already uses—calendar invites, email composers, messaging apps—rather than burying it in another analytics dashboard. At that point the machine stops being a spectator and becomes an on-field coach.
Consider a discovery meeting in which the buyer articulates cost reduction goals and cites a two-month deployment window. Minutes after the call, AI can draft a recap, quantify the stated impact, suggest the next step aligned to the company’s own methodology and attach the case study that addresses that exact pain. Or picture a model trained on thousands of closed won opportunities that can recognize, midcycle, when a live deal drifts away from the historic success pattern and recommends corrective action while there is still time to act. These capabilities already exist, but they are only as effective as the underlying data.
Internal knowledge retrieval is another frontier. Sales reps routinely spend considerable time locating the slide that explains encryption or the customer story from a related vertical. An LLM can transform that haystack into a curated feed: “Provide a two-page ROI brief for a midmarket fintech operations lead” and have it assembled in seconds. Multiply that time-saving across a quarter and the reduction in sales cycle duration becomes material. In addition, it enables a salesperson to handle more opportunities at the same time, more “at bats.”
Customer success and product teams benefit as well. The same models that analyze prospect calls can examine onboarding sessions, support tickets and quarterly business review (QBR) transcripts, surfacing early churn indicators or expansion triggers and routing them to the right owner. The feedback loop between what was sold and how it is consumed tightens; roadmap decisions become data-driven.
These AI layers can sit on top of the CRM rather than inside it. That is a strength, because they are free to pull signals from Zoom, Microsoft Teams, Slack, email, ticketing systems and learning platforms. But it is also a risk, because a brilliant recommendation stranded in a dashboard or report is doomed. The guidance must surface exactly where a salesperson or manager already works or it will not be used.
When evaluating any AI software provider, ignore sales and marketing hype and question the measurable business impact. Phrases like “better decisions” are meaningless—after all, who wants to make worse decisions? Does the tool increase average deal size by sharpening value articulation? Does it shorten cycle time by eliminating friction? Does it raise the opportunity load a salesperson can manage without quality loss? Does it curb churn or accelerate expansion? If the provider cannot demonstrate measurable movement on at least one of those levers, keep your budget closed.
Rumors of CRM’s demise are premature. We will almost certainly see interfaces evolve—static database fields yielding to dynamic timelines, native experiences melting into messaging tools—but the shared ledger of who engaged whom, about what, and with what result remains indispensable. What changes is the cognitive layer. A passive record system becomes an active advisor, summarizing, predicting, nudging. The terminology may shift, yet the foundational need endures.
While traditional CRMs will undoubtedly evolve, maybe even radically, there are technologies and AI-driven capabilities that can be implemented today. To ensure success, implementation discipline matters as much as model sophistication. First, audit data plumbing: are calls recorded and transcribed, emails archived, messages retained and contacts tagged consistently? Second, confirm integration: will insights appear in the systems of action—Gmail, Outlook, Salesforce, Dynamics, HubSpot—or will they demand another browser tab? Third, insist on a realistic change management plan to encourage adoption. Ensure that the capability on offer works with natural workflows and cadences. Applications or solutions that presume a wholesale cultural overhaul rarely make it to the next quarter.
In sum, AI will not rescue disorganized processes or compensate for missing data, but paired with comprehensive instrumentation and thoughtful workflow design, it can increase key drivers of success. It can help lead to Increased deal size through better value demonstration and bundling of products and services. And less need for manual activities such as data entry and searching for relevant information across the many complementary data stores improves the probability and time to close win. Bringing it all together will lead to everyone being able to have more at bats. All these will contribute to better attainment rates on the way to hitting revenue targets. Clean the inputs, embed the results, and the machines will earn their keep. Leave either step undone and you will have purchased nothing more than a new, expensive reminder that technology cannot fix what leadership chooses not to measure.
Regards,
Stephen Hurrell