Sales is one of the few professions where performance is measured in numbers that leave no room for ambiguity. You hit quota or you don't. The pipeline is real or it isn't. Which makes it an interesting environment for evaluating AI honestly: the question of whether it helps isn't a soft one. It shows up in conversion rates, deal size, and time to close.
The pattern that's emerging is straightforward. Top performers are using AI to compress the preparation and administrative work that surrounds selling, which frees up more time for the actual conversations that move deals. Average performers are using AI to generate outreach at scale and wondering why response rates haven't improved. The difference is in how they're applying it.
Pre-Call Account Research with Perplexity
The standard pre-call research workflow has always been slow: navigate to the company's website, read an About page, hunt through LinkedIn for the buyer's background, check the news tab in Google for recent press, look for an earnings call transcript or recent announcement. Putting together a coherent picture of a target account before a discovery call takes 20 to 30 minutes if you're doing it properly.
Perplexity compresses this substantially. Unlike ChatGPT, which works from its training data, Perplexity retrieves live information and cites its sources. For account research, you can ask it:
"Give me a summary of [Company Name] relevant to a sales conversation. Include: their primary business model and revenue streams, any recent news or announcements in the last 90 days, known technology stack or initiatives if public, leadership changes, and any stated strategic priorities from their website or press releases."
You get a sourced summary in 60 seconds. You read it, click through on anything that looks particularly relevant, and you're oriented before you open the LinkedIn profile of your specific buyer. Total research time drops from 25 minutes to under 10 for a typical account.
Use Perplexity for the account; use LinkedIn directly for the individual. LinkedIn data doesn't surface as cleanly through Perplexity, and the buyer's specific background, their path to this role, what they've posted about recently, their visible professional interests, is still best assembled manually. Five minutes on LinkedIn for the buyer plus Perplexity for the account is a more efficient research stack than doing both manually.
Personalized Outreach That Doesn't Sound Like AI
The failure mode of AI-generated outreach is obvious to anyone receiving email in 2026. The compliment that's slightly too broad, the "I came across your profile and was impressed by your work in [industry]," the pivot to value proposition that happens exactly three sentences in. People recognize it and delete it.
The technique that works is specificity injection. AI writes the structure; you inject the specific observation that proves you actually looked. The prompt:
"Write a cold outreach email to a VP of Operations at a mid-market manufacturing company. The reason I'm reaching out is [specific business reason connected to something real about their situation]. The outcome I help with is [specific, measurable outcome]. Keep it to four sentences. No jargon. Don't start with 'I'. Don't mention their name in the first sentence. End with a single low-commitment ask."
Then you edit the first sentence to include a specific, observed detail: a recent product launch, a quote from a LinkedIn post, a specific challenge mentioned in a conference talk, something from the Perplexity research that's actually specific to their situation. This is the sentence AI cannot write for you, because it requires you to have actually looked. The rest of the email AI can draft. That one sentence is yours.
Reps who do this consistently report meaningfully higher response rates than fully AI-generated sequences. The reason is simple: the specific observation signals genuine research, and genuine research signals that the rep believes there's a real fit. Recipients can tell the difference between "I researched you" and "I ran your name through a template."
Objection Response Frameworks
Every seasoned rep has a mental library of objection responses. AI can help you build that library systematically and quickly, which is particularly valuable for reps earlier in their career or when entering a new market where you haven't heard every objection yet.
The prompt structure:
"I sell [product/service] to [buyer type]. A prospect just said: '[exact objection in their words].' Give me three response options. Option 1: acknowledge and reframe without dismissing the concern. Option 2: ask a clarifying question that surfaces the real objection beneath this one. Option 3: share a specific customer example where this same concern came up and what happened. Write each as natural spoken language, not bullet points."
Read all three, pick the one that fits the conversation, and adapt the language to match how you actually talk. Do this with five or ten of your most common objections and you have a reference library built in an afternoon that would have taken months to assemble through trial and error. Then practice them until they're natural.
CRM Note-Taking After Calls
Salesforce, HubSpot, and similar CRM platforms get poor data not because reps don't have information but because entering notes after a call is tedious, it happens when energy is lowest, and there's no immediate revenue consequence for doing it badly. The consequence shows up later, when you're three months into an enterprise deal and need to remember what the CFO's specific concern was in month one.
The AI workflow here: immediately after a call, spend 90 seconds voice-noting or typing the key points while they're fresh. What they said about their current situation, the specific pain they described, what they said they care about in a vendor, any named stakeholders, the agreed next step. Then drop those notes into Claude with this prompt:
"Turn these call notes into a structured CRM note. Sections: Situation (what they described about their current state), Pain (the specific problem they articulated), Criteria (what they said they're looking for), Stakeholders (anyone named and their role), Next Step (exactly what was agreed). Use their words where I've quoted them. Keep it under 150 words total."
Copy it into your CRM. The 90-second voice note plus 60 seconds of AI formatting beats staring at a blank field and trying to reconstruct a 45-minute call from memory two hours later.
Proposal First Drafts from Discovery Notes
A proposal that wins doesn't look like a product brochure. It looks like you were listening. The most effective proposals mirror the buyer's language back to them, reference the specific problems they described, and connect the solution to the outcomes they said they care about. The challenge is that building this from scratch after a long discovery process is time-consuming, and the temptation is to reach for a prior proposal template and do a find-and-replace.
AI handles this well when you give it good inputs. Feed it your CRM notes from every discovery call with the account, organized by stakeholder. Ask it to draft an executive summary section that reflects the situation and pain in the buyer's language, then a section that connects each solution element to a specific stated concern. Review it against your notes to verify every claim maps to something the buyer actually said. Add in pricing, timeline, and specifics manually.
The output needs editing for accuracy and persuasive judgment. But you're editing a draft that already uses the right language and structure, not writing from scratch.
QBR Agenda Preparation
Quarterly business reviews are where enterprise relationships get strengthened or where they start deteriorating. A QBR that feels like a summary of the past quarter is a missed opportunity. One that positions the next quarter's priorities in terms of the customer's goals, shows them what's working and why, and advances a conversation about expansion actually moves the account forward.
AI can build the agenda structure and the prep questions if you give it the account history. Paste in the relevant CRM notes from the past quarter, the agreed goals from the prior QBR, any usage data or success metrics you have, and ask for an agenda that: opens with what they care about (not a review of your company's quarter), covers the metrics that matter to their business, surfaces any issues proactively before they raise them, and ends with a clear next conversation. That structure doesn't come from AI's general knowledge. It comes from the specific account context you've fed it.
What Top Performers Do Differently
The reps who get the most from AI are using it to multiply their preparation, not to replace it. They do the Perplexity research and then add their own insight on top of it. They inject the specific observation into the AI-drafted outreach. They take 90 seconds of notes after every call so they have something worth feeding into AI later. They use AI to draft the proposal and then verify every claim against their actual discovery.
The reps who don't get much from AI are using it to send more emails with less preparation. Volume without relevance doesn't convert. The tools that help you prepare faster and think more clearly about specific accounts are the ones that improve outcomes. The tools that help you contact more people without being more relevant to any of them don't move the number that matters.