The Deal Starts Before the CRM Does. Your AI GTM Motion Should Too.

A new opportunity appears in the CRM. Sales sees a fresh deal. Marketing sees a conversion. RevOps starts the clock. But the account may already have spent months forming an opinion, and the people who shaped it may not be the person who filled out the form.

That gap matters more than the speed of any individual campaign or prospecting sequence. If every team starts working from a different version of the account, the buyer has to carry the context between them. The resulting experience can be perfectly automated and still feel strangely uninformed.

The opportunity record is a late starting line

In research released this September, Factors.ai reports analysing 50,000 closed deals across more than 850 B2B companies. Its finding: buyer research starts, on average, 124 days before a deal appears in the CRM. That is not a universal timetable for every SaaS purchase, nor proof that every early interaction is visible. It is a useful warning about what an opportunity creation date cannot tell you: when the account began to make up its mind.

The same study reports that deals involving six or more engaged contacts before opportunity creation had win rates 17 percentage points higher. Treat that as an association, not a promise that adding contacts will produce the same lift. Stronger vendors may both engage more people and win more deals for other reasons. Still, it challenges a familiar GTM habit: treating one responsive lead as an adequate representation of an account.

The business consequence is not simply missed attribution. It is lost continuity. Marketing may know which issue drew attention. The SDR may know who replied. Sales may know the objection that stopped a meeting. Customer success may know that the same issue threatens expansion at a related account. If those observations do not inform the next action, each team keeps rediscovering what the organisation already knows.

Functional AI can make the handoff worse

There is a tempting response: give each function an agent. One drafts campaign copy. Another researches prospects. Another summarises calls. Another forecasts deals. Each may save time. But if their goals and context remain separate, they can accelerate conflicting decisions. Marketing keeps promoting a claim sales has learned is difficult to defend. An outbound agent contacts someone who already raised a concern. A deal-summary agent accurately condenses a conversation without getting the underlying objection back to content, positioning, or product marketing.

This is the difference between automating tasks and improving a GTM system. The unit of work cannot always be an email, a lead, or a call. Often it is an account-level decision that stretches across months, stakeholders, channels, and teams. When the unit of work is too small, even excellent local execution can produce a disjointed buyer experience.

The risk is not theoretical in its broader form. In a July account of its internal agent pilots, Salesforce said its agent count climbed into the hundreds almost overnight; adoption became inconsistent and quality suffered. Its lesson was to build fewer agents better and scale deliberately. Salesforce was describing its own internal deployments, not measuring GTM performance across the market. But the operational warning applies: the number of agents is a poor proxy for the quality of the work they coordinate.

An AI GTM motion follows the decision, not the org chart

Here is the more useful design brief: give an agent, or a coordinated set of agents, responsibility for maintaining and advancing a defined commercial decision across functions. For example, an account-readiness motion might track whether a target account has the right stakeholders engaged, what each appears to need, which objections are substantiated, and what the next appropriate action should be. It would not own the relationship. It would make the relevant evidence available to the people who do.

Before an opportunity opens, the motion could connect permitted account engagement, known contacts, campaign responses, and existing customer intelligence to identify gaps in understanding. Marketing might receive a brief for a technical evaluator rather than another generic nurture assignment. An SDR might receive a reason to wait, not a prompt to send. Strategy and brand teams might see that a recurring buyer question calls for a clearer proof point, not merely more impressions.

During an active deal, the same motion could compare call notes and approved deal data with earlier account signals. If security becomes the blocking issue, it could ask for the current approved documentation, flag the claim that needs review, and prepare a role-specific follow-up for the account executive to approve. After a loss or renewal discussion, it could route the learned objection back to content, RevOps taxonomy, and customer success where relevant. These are proposed operating choices, not claims about results already achieved.

The distinction matters. A handoff says, “My task is done; yours begins.” A motion asks, “What does this account need next, who is authorised to act, and what did we learn?” The first passes a record. The second preserves decision context.

Shared context is necessary. Shared judgment is the harder part

Vendors are beginning to make the underlying systems more accessible to agents. In a September 9 platform update, HubSpot described agents accessing contacts, deals, conversations, tickets, and activity, while distinguishing raw records from the context needed to interpret them. That is a vendor’s product direction, not evidence that any CRM automatically creates a coherent GTM motion. The implication for leaders is more practical: system access is the prerequisite, while definitions, permissions, and decision rules determine whether access is useful.

OpenAI’s August enterprise analysis offers another useful distinction. Among weekly active users in its enterprise customer base, 21% at its highest-usage firms used Plugins, compared with 9% at typical firms. OpenAI describes these as ways to combine instructions with company tools and data. This is usage evidence, not proof that plugins improve win rates. It does suggest why a standalone assistant that cannot work with current account context is the wrong endpoint for serious GTM design.

The hard work remains managerial. Someone has to define what counts as a meaningful buying signal, which account data is reliable, when a message requires human approval, and what an agent must never infer from sparse evidence. Someone also has to resolve disagreements: marketing’s engagement score is not the same as sales’ conviction that a deal is real, and neither substitutes for customer success’ view of adoption risk.

Start with one motion and make those choices explicit:

• Choose a consequential cross-functional decision, such as whether an account is ready for outreach or whether an objection requires a change in messaging.
• Name its human owner and the teams that must receive its output. Specify the records it may use, their freshness, and the evidence it must show.
• Set approval and escalation rules for external communication, uncertain identity matches, sensitive customer data, and changes to positioning.
• Measure the decision, not just the agent: stakeholder coverage, time to resolve an objection, unnecessary outreach avoided, and whether the resulting activity helps advance or preserve the account.

Do not try to make the agent omniscient. An account-level view will have blind spots; private buyer conversations will stay private, and a signal is not consent to contact. A useful motion acknowledges uncertainty, asks for missing context, and hands off to a person when the cost of a wrong assumption is high.

The competitive question is whether your GTM organisation remembers

The September research gives leaders a reason to question when their pipeline story begins. It does not prescribe an AI architecture. The choice of architecture follows from the operating problem: buyers develop preferences before the CRM calls them an opportunity, and their concerns do not respect your team boundaries.

So the next GTM investment should not be judged by how many agents it puts to work or how much content they produce. Judge it by whether the organisation gets better at recognising what an account knows, what it still needs, and who should act next. A collection of tools can make every department faster. An AI GTM motion has a harder job: helping the whole company respond as if it has been paying attention.

This blog was created & posted by Jan’s AI content agent. Get in touch if you want to know how it works.