QuotaforgeAI Productivity in Sales: What Studies Really Show
Almost every software solution promises that a team will be able to get twice as much done starting tomorrow. Reliable research paints a more nuanced picture: AI makes people faster at what they already do—both good habits and bad ones—and it isn’t equally effective in every area.
The Short Answer
AI is not an additive but a multiplier. It amplifies what’s already there. A clear, shared language of qualification, multiplied by AI, yields scaled quality. An unclear language, multiplied by AI, yields—above all—faster chaos. Four studies show how varied the effect is in practice.
Who Benefits the Most
The largest field study on the topic to date comes from Erik Brynjolfsson, Danielle Li, and Lindsey Raymond. They analyzed how the introduction of an AI assistant changed things for 5,179 customer support employees. On average, employees resolved 14 percent more inquiries per hour. More intriguing is the distribution: For new hires, the increase was 34 percent, while for the most experienced employees, there was barely any measurable effect. The explanation: The AI passes on the best practices of the top performers to everyone else. For sales, this is a powerful insight, because this is precisely the most costly problem in onboarding—it takes months for new hires to ask questions, listen, and qualify leads as well as the best performers on the team.
The Jagged Technological Frontier
Researchers at Harvard Business School, in collaboration with the Boston Consulting Group, had 758 consultants work on realistic tasks, some of them using GPT-4. For tasks that were well-suited to the AI’s capabilities, it completed 12.2 percent more tasks, was 25.1 percent faster, and delivered results of more than 40 percent higher quality. For a task outside these capabilities, the picture reversed: those who used AI were 19 percentage points less likely to arrive at the correct solution. The authors speak of a jagged technological frontier; AI is not uniformly good, but rather brilliant at some tasks and surprisingly unreliable at others that seem similar.
AI as a Translator Between Ways of Thinking
A research team led by Fabrizio Dell’Acqua studied how AI is changing collaboration at Procter & Gamble with 776 professionals. Individuals using AI achieved the same performance as teams without AI. Without AI, participants remained stuck in their professional thought patterns: people from research and development tended to propose technical solutions, while those from the business side tended to suggest commercial ones. With AI, this difference largely disappeared; both groups developed balanced solutions that combined both perspectives. This is the same issue as with the “thought worlds” of marketing and sales, only viewed from the opposite direction.
Perceived and Measured Efficiency
The research organization METR observed sixteen experienced software developers as they worked on 246 real-world tasks from their own projects. Using AI tools, they took 19 percent longer. They had previously expected a 24 percent time savings, and even afterward, they estimated that AI had sped them up by 20 percent. The sample size is small, but the lesson still applies to every department: perceived efficiency is not the same as measured efficiency. Anyone introducing AI should define in advance how success will be measured, rather than asking the team afterward if it feels faster.
What this means
Criteria before tools: The common language for evaluation must be established before anything is automated. If you’ve never defined what a good discovery conversation looks like, you’re not giving the AI anything to work with. And if you automate lead scoring before marketing and sales have agreed on a definition, you’re mainly accelerating the conflict between the two departments.
Practical tip: We align deal scoring and conversation analysis directly with MEDDIC and MEDDPICC; lead qualification in marketing uses the same terminology as in sales; and partner scoring uses the same criteria as direct sales. Otherwise, a second silo emerges—this time an automated one.
Questions about AI productivity in sales.
Does AI really make sales teams more productive?
Yes, but the impact is unevenly distributed. In a field study involving 5,179 support staff, the increase for new hires was 34 percent, while for the most experienced employees, there was barely any measurable effect because the AI primarily replicates the best performers’ approaches and applies them to everyone else.
Is AI equally effective for every sales task?
No. A study by Harvard Business School in collaboration with the Boston Consulting Group reveals a distinct technological limit: For tasks within the scope of AI’s capabilities, quality improved by more than 40 percent; for tasks outside that scope, the team using AI was 19 percentage points less likely to arrive at the correct solution.
Can AI bridge the different ways of thinking between marketing and sales?
A study at Procter & Gamble involving 776 professionals suggests so: Without AI, participants remained stuck in their functional thought patterns; with AI, this difference largely disappeared because both groups developed more balanced solutions.
Does using AI always feel faster than it actually is?
Yes, as shown by a METR study involving experienced software developers: Using AI tools, they actually took 19 percent longer, but afterward estimated they had been 20 percent faster. Perceived efficiency is not the same as measured efficiency.
Do you know where your AI is already going wrong today?
15–30 min · Free · No sales pressure