Speed is the advantage: Five signals from leaders at Dreamforce 2026
AI is embedded in how nearly every commercial organization operates. At Dreamforce, Salesforce’s annual technology conference, the question animating conversation after conversation was whether organizations are capturing any real value from AI. One answer came through clearly: Putting AI on top of existing ways of working isn’t enough.
“Avoid the trap of infusing AI on top of the way things are done today,” says Steve Reis, a McKinsey senior partner. “Pick an area of the business that’s particularly important and rethink it—that’s where you find the most value.”
Across three days of roundtables, conversations with commercial leaders, and interviews with senior executives at McKinsey’s Growth Launch Pad—our Dreamforce hub exploring the shortest path from AI to ROI—five themes stood out.
The biggest opportunity is closing the AI impact gap
“This year is the age of action,” says Lisa Harkness, a McKinsey partner. “We know we have to do something differently to get the ROI. Treating AI like a widget doesn’t work. Treating it as an enabler of human and strategic choices does.”
One question surfaced in almost every conversation: Where is the profit-and-loss impact? Nearly 90 percent of companies use AI in at least one business function, but only 37 percent attribute any EBIT impact to it. In commercial functions specifically, fewer than 10 percent of organizations have scaled AI in any given function.
The distinction that kept coming up was between surface-level adoption and genuine transformation. Many companies are still in the chat window era: prompting tools to summarize, draft, or answer questions. Leaders are moving beyond that, using AI to run experiments simultaneously across sales, marketing, and customer success, compressing the cycle time from idea to experiment to production decision.
The emerging competitive advantage is not model sophistication but how quickly companies can test, learn, and adapt. Early AI leaders are achieving EBITDA uplifts of 20 percent or more by rewiring their commercial engines around AI. What separates the companies pulling ahead is leadership willing to rethink how growth happens in the first place—not just leadership willing to adopt more tools.
The operating model is the unlock
The bottleneck in AI transformation is not the technology. What works is breaking down functional boundaries and bringing data scientists, sellers, prospecting leaders, and analytics leads into integrated teams with clear goals and decision rights. Go-to-market teams of the future will bring together a broader mix of skills, with greater fluidity across functions.
Tjark Freundt, a McKinsey senior partner, pointed to a shift in how leaders are approaching this: “There is a new spirit in discussing how to deploy AI for growth rather than productivity—but with a good measurement system in place, so we not only are doing exciting stuff but also actually know what we’re yielding.”
AI cannot be layered onto old business processes and expected to produce transformation. The reimagination has to reach the level of what decisions get made, by whom, and how work flows. Our research indicates that for every $1 spent on deploying AI, companies may need to spend $3 on change management. Yet most organizations invert that ratio.
“It can’t just be the tech,” says Brian Gregg, a McKinsey senior partner. “It’s humans, lots of technology, and reworking and rewiring the organization. That’s what’s breaking through.”
AI is shifting commercial organizations from asking to doing
Designing sales territories—deciding which sellers should cover which accounts—once required weeks of work across revenue operations, strategy, and sales management. Today, AI tools can draw on customer-relationship-management (CRM) data, internal communications, and external signals to recommend a plan within minutes.
The deeper implication is organizational. The commercial function is shifting from a sales organization to a learning organization—one that captures every signal from every customer interaction and routes it in real time to product, engineering, and customer support.
At the agent level, leading companies are moving through a maturity progression: from assistive agents that augment human judgment to augmentative agents that do the heavy lifting on specific workflow segments to more autonomous agents that execute end-to-end workflows. Win-back programs, personalized outreach, and dynamic prospect engagement are among the use cases already moving into autonomous territory. Companies redesigning customer journeys around agents are seeing faster sales cycles, higher conversion rates, and lower cost-to-serve ratios.
Everybody is building agents; almost nobody has figured out how to orchestrate them
Agentic AI hasn’t eliminated organizational silos; in many cases, it has replicated them. Marketing agents handle the upper funnel. Sales agents handle the middle of the funnel. Too often, the agents aren’t talking to each other any more than the teams behind them were. The orchestration problem is real and largely unsolved. Individual agents may work well within their lanes, but companies are still figuring out how to connect them across the customer journey. And because nearly every vendor has become an AI vendor, organizations are navigating a proliferation of tools that can be useful in isolation but do not necessarily fit into a coherent architecture.
That’s the challenge Jan-Christoph Köstring, a McKinsey senior partner, sees companies confronting.
“Too many companies are thinking about individual use cases rather than thinking end to end,” he says. “You end up with isolated use cases that aren’t integrated into the business—and that doesn’t drive value.”
What leading companies are building toward instead is a more connected model: persistent agents that proactively surface signals to human experts, bringing forward the right information at the right moment, a recommended next-best action, and the relevant customer context.
Most companies are still some distance from this model. But the answer isn’t to wait for the perfect architecture. Start with individual use cases, but design them with the end-to-end journey—and eventual orchestration—in mind. Then iterate.
Trusted AI starts with trusted data
Data was the thread running through every other conversation. The real unlock is building a data foundation that gives persistent agents a complete, current view of each customer across the life cycle, combined with external signals about prospects not yet in the funnel. Agents cannot generate trusted recommendations if customer, product, pricing, and interaction data remain fragmented or poorly governed. That is also the premise behind the move toward more open architectures: Any AI interface is only as useful as the data beneath it. That data needs to be structured, trusted, and complete. The same principle applies to people. Leaders need to make clear what’s in it for employees and align incentives with new ways of working. Companies capturing value from AI are building better data foundations, redesigning workflows, and changing how their organizations learn. That’s what turns AI investment into compounding advantage—and why speed is the advantage.
