What Happened When We Gave Our Sales Team an AI Assistant

I want to describe what actually happened, not the version we put in the quarterly update. We gave our sales team access to an AI assistant connected to our CRM, our product documentation, our past proposals, and our win/loss analysis. The pitch was straightforward: faster proposal drafts, better call prep, smarter objection handling based on what had worked historically. Eighteen months later I have a clear picture of what changed, what did not, and what I would do differently. The thing that changed the most was not what I expected. I had anticipated productivity gains around proposal drafting, which happened, and around call preparation, which happened partially. What I did not anticipate was how much the AI changed the internal dynamics of the sales team itself. Before the AI, institutional knowledge about what worked in our sales process lived primarily with the senior reps. They had been through hundreds of deals. They knew which objections were deal-killers versus which ones were negotiating postures. They knew which product features actually mattered to which buyer personas. They knew how deals typically progressed at each stage. New reps learned this by sitting with senior reps and by losing deals and doing retrospectives. After the AI had been running for six months and had access to our win/loss data and proposal history, it could surface much of that institutional knowledge on demand. A new rep could ask "what objections do we typically hear from compliance-focused buyers at the proposal stage and how have we addressed them successfully?" and get an answer grounded in our actual historical deals. This was good for the new reps. It was complicated for the senior reps. A few of the senior reps felt their expertise had been commoditized in a way they had not consented to. Their years of accumulated knowledge had become a training dataset. Some of them became less willing to document their reasoning because they had connected the documentation to their own job security. The dynamic between senior and junior reps shifted in ways that were not immediately visible in the metrics but were very visible if you paid attention to how the teams were talking to each other. I do not think we handled this well. We treated the AI deployment as a technology project and not enough as an organizational change. We did not have explicit conversations with the senior reps about how their expertise would continue to be valued in a world where some of it was now accessible through a tool. We did not redesign their roles to reflect the changed information landscape. We expected the productivity gains to speak for themselves. What I would do differently: before giving any AI system access to knowledge that lived in specific people's heads, have explicit conversations with those people about what is happening, why it serves the organization, and how their role evolves. The resentment that built up in a few of our senior reps took longer to address than the AI deployment itself. The proposal drafting gains were real. Average proposal turnaround went from four days to one and a half days. The quality held up, which I had been skeptical of. The AI was better at retrieving relevant past language and client-specific context than I had expected. The call preparation gains were more mixed. The reps who used the AI most effectively for call prep were the ones who had enough context about the deal to know what to ask for. The ones with less experience sometimes got the AI prep and mistook it for understanding. Knowing what objections had historically come up is not the same as knowing how to respond to them in a live conversation. A few junior reps went into calls more confident than they should have been and came out having missed signals that an experienced rep would have caught. The lesson there is that AI-assisted preparation is not a substitute for experience, it is a multiplier on experience. Reps who had the underlying skills got more out of the tool. Reps who did not have the skills got a surface-level confidence that occasionally made things worse. What I am still working out: how to use the AI to accelerate the development of genuine sales judgment, not just to provide shortcuts around it. The tools exist. The workflow design to use them well for skill development rather than skill substitution is harder and we are not there yet.