How to Solve Any B2B Sales Challenge with AI

When most sales leaders think about doubling revenue, they think bigger. More reps. More leads. More spend. But there's a simpler path: better execution at the moments that actually decide deals.

A 10% improvement at each of seven key conversion moments in the sales process — prospecting, qualification, discovery, value confirmation, proposals, negotiation, close — nearly doubles revenue. $1.7M becomes $3.3M on the same 1,000 leads and $50K ACV. No new headcount. No new pipeline. Just better execution at the moments that matter.

This is Part 2 of a two-part series. Part 1 introduced a five-step approach for solving any business problem with AI. This article applies that same approach to the challenge I've spent my career on: how to systematically improve sales results without adding headcount or pipeline. The core argument: best practices exist for every key moment in the sales process, and AI now lets you embed them directly into execution.

The underlying logic isn't new. It's the same discipline behind Six Sigma and DMAIC in manufacturing. In sales, the production line is your sales process: a series of interactions between your org and customers. Those interactions can be defined, measured, analyzed, redesigned, and reinforced. The discipline is identical. The application is different.

Sales leaders already know this intuitively. They invest in training. They build playbooks. They roll out methodologies. And then they hope reps actually follow them.

But 84% of training is forgotten in 90 days. And forgetting is only part of the problem. Time passes. Reps who understood the method in training never practiced it enough to make it automatic. Others lack the underlying skills; they can recite the methodology but can't execute it under pressure in a live conversation. Playbooks live in PowerPoints, not in practice. Managers who should be reinforcing are stretched too thin to coach consistently. The gap between what the playbook says and what the human does is the most expensive problem in B2B sales.

Here's what's changed. Best practices exist for every key moment in the sales process, backed by decades of research, proven across thousands of sales calls, documented in methods that work. AI now lets you embed those practices directly into execution. Not as training people forget. As systems that run at every key moment, with full effort, every time.

Here's the five-step approach, briefly:

In over a decade of building sales playbooks and analyzing pipelines for B2B clients, I've become convinced that most enterprise sales processes follow the same general conversion points. The labels vary. The CRM stages differ. But the moments where deals actually advance or stall are remarkably consistent.

Companies with formal, enforced sales processes win 48% more, cycle 37% shorter, and generate 2x revenue per head. The mechanism is consistency — the same theme from Part 1.

The problem isn't that teams don't have processes. It's that the processes don't execute themselves.

This is the part most AI implementations skip. It's also the reason most of them produce mediocre results.

The gap between average and excellent is methodology.

In every field with measurable performance, the difference between average and top performers traces back to specific, documented behaviors. Not talent. Not effort. Behaviors, executed consistently, at the moments that matter. That's what methodology is: the codified practices of top performers, validated across enough data to trust.

A surgeon's hands follow a sequence refined across thousands of procedures. A concert pianist's technique reflects years of deliberate practice on fundamentals that most listeners never notice. A free throw shooter bends the knees, aligns the elbow, follows through the same way every time. Not because it's natural, but because the research says it works, and the repetitions made it automatic. Top performers in every discipline have internalized the best practice so deeply that it looks like instinct. It isn't. It's methodology, practiced until it disappears.

Sales is no different, except we rarely treat it that way.

Rackham's SPIN Selling study — 35,000 sales calls, 10,000 salespeople, 23 countries, 12 years — is one of the largest field studies of sales effectiveness ever conducted. The core finding: the top 10% of sellers outsell the bottom 10% by 3:1. The difference wasn't charisma or closing technique. It was what they did differently in live conversations. Specifically, how they asked questions that moved buyers from recognizing problems to articulating the value of solving them. And Rackham himself has noted that the measured 17% productivity gains came not from the question sequence alone, but from embedding it into a systematic coaching process. The questions were the what. The system was the how.

That's one moment. One body of research. Every other key moment has an equivalent: how to test champion strength, how to structure proposals around quantified impact, how to calculate cost of delay to maintain urgency.

The problem: AI without methodology produces average output.

Ask an LLM to write you a discovery call script. What you get back is a synthesis of everything it's been trained on: the median of every blog post, sales book summary, and generic advice article on the internet. It's competent. It's also average. And average is exactly what the bottom 90% of sellers already do.

This is where deep research changes the equation. Tools like Gemini Deep Research, Perplexity, and Claude's extended thinking don't just search. They synthesize across peer-reviewed studies, published field research, and practitioner-validated methods. They surface the specific findings that separate top-quartile performance from the mean. Not "ask open-ended questions." The specific question types, in the specific sequence, that decades of research say predict close rates.

But even research-backed best practices are still generic until you apply them to your sale.

SPIN says to ask implication questions. But implication questions for a CFO evaluating an analytics platform are different from implication questions for a superintendent buying EdTech. The research tells you what type of question works. Your business context tells you what to ask about.

That's why Steps 2 and 3 of the approach work together. Deep research gives you the evidence-backed principles. The AI interview personalizes those principles to your buyers, your product, your competitive position, and the specific gaps in your team's execution. Research without personalization gives you a generic playbook. Personalization without research gives you someone's best guess. You need both.

Discovery quality determines everything downstream: deal velocity, win rates, contract value, forecast accuracy. It's also where execution risk is highest. Rackham's research found that in complex B2B sales, aggressive closing backfires and traditional techniques fail. What works is structured questioning that moves buyers from implied needs to explicit needs. Top performers ask 11-14 questions per discovery call versus 6-8 for average performers.

Step 1: Start with the problem. Reps run discovery calls that check boxes but don't uncover real pain. They accept vague answers. They pitch before they diagnose. Conversion from discovery to value confirmation is below benchmark.

I saw this firsthand in a client engagement last year. We analyzed 78 recorded discovery calls and found the answer to a conversion problem the team couldn't explain: four completely different market segments were being treated identically: same talk track, same questions, same qualification criteria. The insight was embedded in the calls, invisible to every structured report in the CRM.

Step 2: Research best practices. Run deep research on SPIN methodology, optimal question sequencing (Situation → Problem → Implication → Need-Payoff), pain funnel technique (surface → business → emotional), and impact quantification. Feed the research into an LLM: "What are the proven best practices for running a discovery call in complex B2B sales? What question types, sequences, and behaviors does the research say separate top performers from average ones?"

Step 3: Let AI interview you. Give Claude your discovery method, your buyer personas, your product context, and transcripts from your best reps' calls. Let it ask every question it needs. It surfaces the gap between what your best reps actually do and what your playbook says.

Step 4: Let AI build it. A call analysis system that scores every discovery transcript against the research-backed method. Scores each dimension. Identifies coaching moments. Provides specific next-call questions: "You confirmed the problem but never explored implications. Here are three implication questions tailored to this buyer's role."

Step 5: Iterate. First version scored too rigidly. Second version weighted implication and need-payoff questions more heavily, the two categories most predictive of close in Rackham's data. Each iteration: hours, not weeks.

Most deals don't die in negotiation. They die because the proposal doesn't connect to the pain and impact established in discovery. The average B2B win rate sits between 20-21%. Fully customized proposals achieve 50-65%. That's a 40-percentage-point spread from a single variable: personalization. And the research shows that 70% of deals sales teams attribute to price were actually lost because value wasn't adequately demonstrated. The proposal is where everything upstream either compounds or collapses.

Step 1: Start with the problem. Proposals are generic. They lead with features and company history instead of quantified business impact. They're written for the champion, not the 6-10 stakeholders on the buying committee. Reps don't calculate cost of delay. They don't design the document to enable internal selling. Urgency evaporates between verbal commitment and signature.

Step 2: Research best practices. The evidence says: a proposal is a commitment document, not a feature list: financial commitment, effort commitment, decision commitment. The executive summary is the most critical section and should be written last. Optimal length is 10-11 pages, 7 sections. Case studies increase win probability by 73%. Pricing tables increase conversion by 54% over narrative pricing. Proposals delivered within 24 hours increase win probability by 25%. Every one of these is a documented, researchable best practice.

Step 3: Let AI interview you. Feed Claude your proposal templates, your ROI model, your pricing structure, and the discovery data from the deal. It asks about the specific value streams, the stakeholders who need to see themselves in the document, the procurement timeline.

Step 4: Let AI build it. A proposal generation system that pulls quantified impact from discovery, structures the document around the buyer's stated priorities (in the buyer's language, not yours), calculates cost of delay, designs stakeholder-specific content layers (CFO sees ROI, CTO sees integration path, champion gets internal selling points), and generates a mutual action plan with owners and dates. Not a template. A system that builds a deal-specific commitment document grounded in everything the rep uncovered.

Step 5: Iterate. First version was too long. Research says win rates decline above 11 pages. Second version added role-specific executive summaries and pricing tables. Third version linked to the discovery scoring system so proposals can't be generated until discovery scores meet a minimum threshold. No more generic proposals on unqualified deals.

You don't need to improve everything. You need to pick the two or three moments where your conversion data shows the biggest gaps, then apply the five steps:

The traditional version of this (train the team, hope the behaviors stick) takes quarters and produces temporary results. The AI-native version takes weeks and the system keeps running. I've watched this pattern across dozens of client engagements: the teams that improve fastest aren't the ones with the best talent. They're the ones that systematize their best practices and reinforce them consistently.

The data supports that observation. BCG found that below-average performers improved 43% with AI, compared to 17% for above-average. The 40% of your team that's underperforming is the highest-ROI target. AI doesn't replace your best reps. It brings everyone else closer to their standard.

The sales teams that win in the next two years won't be the ones that bought the most tools. They'll be the ones that identified their key moments, embedded best practices into systems, and made excellence the default — not the exception.