Why the next phase of enterprise AI will be multiplayer
Enterprise AI has a duplication problem.
We have spent the past few years celebrating how much faster one person can work with AI. Less attention has been paid to what happens when hundreds of people use AI independently, repeatedly researching the same questions, processing the same information, and rebuilding context that already exists somewhere else in the organization.
Individually, everyone looks more productive. Collectively, the economics become much less convincing.
When AI is used primarily as an individual tool, organizations may make each interaction more efficient while still duplicating enormous amounts of work across the enterprise. But performing essentially the same work comes with a cost too.
That is the hidden economics of what we might call single-player AI. And as organizations move toward broader adoption, it is worth asking whether making individual employees faster is really the best measure of AI value.
The duplication problem
Consider how knowledge work typically happens today. An employee starts a project, gathers background information, searches through documents and emails, asks an AI assistant to summarize what matters, and begins developing an analysis.
A colleague joining the project later may repeat much of that process. A third person might ask another AI system similar questions. An AI agent assigned to a related task may retrieve and process some of the same information again.
Each interaction may be productive in isolation. Collectively, however, the organization is repeatedly paying to reconstruct context it already created.
The cost is not limited to model calls or tokens. Employees spend time finding information, rebuilding context and validating outputs, then reconciling independently generated answers across the team.
This is not a new organizational problem. Knowledge has always become fragmented across inboxes, meetings, documents, and individual employees. AI can simply accelerate the fragmentation because machines can now generate research, analysis, and content at a scale humans never could.
As organizations deploy more specialized AI agents, that challenge is likely to grow. Instead of one employee working with one AI assistant, a project may involve multiple people working alongside research agents, analytical agents, workflow agents, and specialized models. If each operates independently, more AI can mean more duplicated intelligence rather than more collective intelligence.
What changes when context persists?
The alternative is a more “multiplayer” approach to AI. Multiplayer AI shifts the unit of productivity from the individual interaction to the shared body of work. Rather than every person or agent starting with a blank prompt window, people and AI systems can work from persistent context that develops over the life of a project.
Imagine a market analysis involving multiple people and AI agents, each contributing research, data and expertise. In a single-player environment, those contributions remain scattered across separate sessions and applications. Someone eventually has to assemble it.
In a multiplayer environment, each contribution becomes part of the context available to the next participant, whether that participant is human or AI. New people and agents do not need to reconstruct the entire history before they can contribute. They can see what has already been explored, what evidence supports a conclusion, where questions remain and how the team’s thinking has evolved.
That changes more than the user experience. It changes the economics. Research becomes reusable. AI-generated analysis can become an input to subsequent work rather than a disposable output. Institutional knowledge accumulates instead of disappearing when a chat session ends, an employee changes roles or an agent completes its task.
The goal is not simply to get an answer faster. It is to avoid paying repeatedly to rediscover what the organization already knows.
From individual productivity to collective outcomes
The shift also suggests that enterprises may need different ways to evaluate AI investments.
Many of today’s AI metrics are naturally tied to consumption: number of users, licenses, prompts, model calls, tokens, or agents deployed. Productivity measures often follow the same individual orientation. How much faster can an employee draft a report? How many hours can an AI assistant save? Those measurements matter, but they tell only part of the story.
Despite widespread adoption, translating individual AI use into enterprise-level value remains difficult. McKinsey’s State of AI research found that more than 80% of respondents were not yet seeing tangible enterprise-level EBIT impact from generative AI. Its research also identified workflow redesign as one of the most important factors associated with capturing value from AI.
That shift is already beginning. One federal organization, for example, is developing an AI environment that can draw on multiple models and applications rather than simply giving every user access to the most powerful model available. The reasoning is partly economic since frontier models are expensive, and the incremental value can diminish for many everyday tasks. Instead, the focus is shifting toward applying the right AI capabilities where teams are working, sharing context and making decisions.
An employee who completes a task 30 percent faster has created measurable value. But if that work is being duplicated elsewhere, individual productivity gains may overstate the value the organization is actually capturing.
A multiplayer model creates an opportunity to measure something more consequential—collective outcomes. How much duplicated work was avoided? How quickly could new people or agents contribute because context already existed? How much institutional knowledge was preserved? And did teams make better, faster decisions because evidence and analysis could be evaluated together?
These measures are harder to fit neatly into a dashboard, but they get closer to what enterprises ultimately expect from AI, chiefly not just more output, but better organizational performance.
Multiplayer is the next phase of enterprise AI
The first wave of generative AI understandably focused on giving individuals powerful new capabilities. That was the easiest way for millions of people to experience the technology and discover where it could improve their work.
The next phase will be different. The new shape of work will increasingly involve teams of humans and AI agents contributing different capabilities, entering and leaving projects, and building on one another’s work.
The question is no longer simply whether every employee has access to AI. It is about whether all that intelligence adds up to something greater than the sum of thousands of individual interactions.
The organizations that answer that question well may discover that the biggest economic opportunity in AI is not making every person and every agent work faster on their own. It is creating an environment where they no longer have to keep starting over.
Author:
John Greenstein is CEO of Bluescape, where he leads the company’s strategy and growth in co-Active AI and secure AI-powered operational workspaces. With more than 20 years of experience in enterprise software, he has held leadership roles spanning sales, marketing, and business development at both emerging technology companies and global enterprises.
