Concentration, Not Coverage: Where AI and Your People Actually Pay Off

AI is saving sellers close to five hours a week. Almost nobody is doing anything with the time: 72% of organizations never redeploy it to higher-value work, so the productivity gain quietly disappears, according to a report from Gartner. The tool is not the problem. The problem is that most companies invest the way they always have, spreading AI thinly across every task and spreading coaching evenly across the funnel. Both feel responsible. Both dilute the return.

The organizations pulling ahead do the opposite. They concentrate AI on the few repeatable tasks where it compounds, and they concentrate their people on the few moments where a human changes the outcome. Coverage is the enemy. Concentration is the strategy. The question was never AI versus your people. It is where each one earns its keep.

Where we still bet on a person

Some moments turn on judgment, not information, and no model has cracked them.

Eight years ago we worked with a VP of sales at a medical diagnostics company who knew his pipeline report cold and knew almost nothing about his people. His team’s forecast came in about 33% higher than what they actually closed, and the company built capacity for demand that never materialized. The CRM only knew what the reps typed into it. Telling fact from fiction, and coaching reps to forecast accurately, is human work. What that leader needed was not a better dashboard. It was the skill to read his own team.

On one deal, our champion had her boss’s backing and set the decision meeting with the executive sponsor. The day before, the sponsor canceled it and asked a newly hired change-management lead to review everything first, a stakeholder we had not known existed and who was in no hurry. On paper the deal had stalled. Because we had spent months earning our champion’s trust, she did what no dashboard would predict: she took what we had built, worked around the new reviewer, and carried it to another member of the C-suite who could move it. It closed. No model flags the person quietly blocking a deal, or the relationship that rescues it.

Long pursuits work the same way; a year of iteration can turn on reading indecision that AI analyzes but cannot resolve. Buyers want the human in it too: in 2026, 69% said they would validate whatever AI told them with a person before acting, and those who skipped that step reported more purchase regret, according to a Gartner report. That read, and the trust to act on it, is learnable, and it is exactly what your coaching should build.

Where we give AI its due

The repeatable work is a different story, and here the gains are real. Sellers spend well under half their week actually selling; the rest disappears into research, admin, and prep, according to a report from Salesforce. That is the pool AI can drain. It pulls account research, drafts the recap and the next-step note in a seller’s own voice, and surfaces the account going quiet that a rep would have missed. Done well, it hands hours back across the team. The catch is the one from the opening: speed only pays off if the freed time goes back into the moments above.

The highest-return use is reinforcement. A few years ago we built a moments-based program for a global chemical manufacturer whose sellers were still selling transactionally while the market shifted under them. Instead of a full-funnel course, each phase dropped sellers into a simulation of the exact moments that decided their deals, forcing a real decision against the capability we were building, then went deeper in manager and frontline reinforcement. Repetition where it mattered, not coverage everywhere. In the period right after, the business saw a 10% increase in EBITDA.

A simple test

For any activity, ask three questions.

• Is the work repeatable and rule-bound? The more it is, the stronger the case for AI.
• Is the cost of getting it wrong high and hard to reverse? The higher it is, the more it belongs to a person.
• Will the buyer attribute the trust to a human? If so, keep one in it, even when AI does the mechanical part.

Where to spend the next dollar

Start with forecasting, and resist the obvious move. Do not automate a forecast on top of reps who were never taught to build one; you only make a bad number look credible and arrive faster. Fix the skill and the accountability first, then add the AI.

Everywhere else, treat it as a budget decision. Stop funding full-funnel training and thinly spread AI. Concentrate coaching on negotiation, buying-committee navigation, and the judgment that turns ambiguity into a number. Concentrate AI on research, drafting, and reinforcement. And before any of it, answer the boring question, “What happens to your data once it enters a vendor’s tool?” If no one can answer that, you have exposure, not a strategy.

Chasing AI because everyone else is, without building the skills of the people who use it, buys the appearance of progress. Bet on your people first, point AI at what is repeatable, and stop pretending both deserve the same size investment. That is not a neutral take. It is one experience has taught us.

Patrick Kammerer is Associate Director for the Sales and Marketing Practice at BTS, a global professional services firm.