Technology Report

How AI Is Changing Data Monetization

As AI agents reshape enterprise software, companies must take a fresh look at what data to expose, how to provide access, and how to price it.

By Eric Maltiel, Josh Sandberg, Justin Murphy, and Mukarram Bhaiji

Published on September 29, 2026
min read


At a Glance Model Context Protocol (MCP) is emerging as a new standard that allows AI agents to access enterprise software and data, prompting companies to make some critical decisions. Deciding what data to expose means weighing the risk of opening up valuable data against the risk that customers and AI agents bypass your platform in favor of more accessible alternatives. Deciding whether to use MCP means balancing easier integration into agent workflows against the potential for customers to accomplish more without ever engaging with your interface. Deciding how to monetize comes last, with pricing designed to capture the value of agent access without cannibalizing higher-value revenue elsewhere in the portfolio. This article is part of Bain’s Technology Report 2026 Explore the report Model Context Protocol (MCP), an emerging standard that lets AI agents securely connect to enterprise software and data, has gone from a niche protocol to a de facto standard at remarkable speed, growing from its late 2024 launch to more than 20,000 servers less than two years later (see Figure 1). Figure 1 Model Context Protocol server growth signals its rise as the new default for agentic access to data Sources: MCP.so (a server registry); Glama.ai; PulseMCP Vendors are understandably asking how to charge for AI agents that access their data and workflows through MCP. But pricing should be the third question, not the first. Before deciding how to monetize, software companies need to first answer two more fundamental questions: What data should we expose programmatically? Should that access be provided through MCP? Only then should they decide how to monetize it. Those first two decisions determine a company's strategic position in the agent era. Pricing should rest on this foundation. Decide what data to expose This is the hardest decision because exposing data is often a one-way street. Once customers and partners build their systems around this data, removing it becomes very difficult. Companies should weigh the benefits against the risks for each category of data rather than treating openness as all-or-nothing. Benefits: meet growing customer expectations for access to the provider’s data; become embedded inside customer workflows, increasing stickiness; and create new monetization opportunities where customers are willing to pay. Risks: disintermediation as customers and competitors bypass the user interface; greater data vulnerability as agents gain read/write access; higher infrastructure and telemetry costs; and regulatory issues around data sharing. Not all data should be treated equally. Customer records, proprietary analytics, benchmark data, schemas, credentials, and workflow metadata each have different strategic value and risk profiles. A key strategic decision is determining where your competitive advantage lies. If it’s proprietary data, then protecting access may strengthen your position. If your advantage is workflow orchestration and governance, broader access may actually reinforce your platform. The contrast between LinkedIn and ServiceNow illustrates this distinction. LinkedIn treats its professional graph as the core asset and tightly controls programmatic access; ServiceNow treats workflow execution as the asset, exposing governed workflows broadly through MCP while monetizing usage and retaining operational data. Decide whether that access needs an MCP Once you've decided what to expose, the next question is how to expose it. Traditional APIs are designed for developers who write custom integrations against specific end points. MCP serves a different purpose: It provides a standard way for AI agents to discover and use an application's data and actions without requiring a bespoke integration for each vendor. As MCP adoption grows, it lowers the cost of connecting enterprise software into agentic workflows. That creates a strategic trade-off. Offering MCP access makes it easier for customers' agents to work with your platform, boosting adoption but also making it easier for users to accomplish work without ever opening your interface. Declining to offer MCP reduces that risk but raises another: Agents may simply route around your platform in favor of vendors that are easier to integrate. The question, then, isn't whether MCP is risky; it's which risk is greater for your business: enabling disintermediation or being excluded from emerging agent ecosystems. One way to mitigate this trade-off is to pair MCP access with investment in your own first-party agents—that is, agents you build on your own platform, such as Atlassian has done with Rovo. This allows customers to work through your agent experience while still participating in broader agent ecosystems. In effect, you're placing two bets: You're making it easy for customers to use MCP as that ecosystem grows while also building your own agent experience to compete for those same users. Decide how to monetize Only after deciding what to expose and whether to offer MCP access should vendors turn to pricing. Here, the market is still experimenting rather than converging. Roughly half of vendors monetize programmatic access through usage- or outcome-based pricing, about one-third charge for additional capacity or higher rate limits, and roughly one-fifth monetize access itself. Most apply similar pricing principles to MCP and traditional APIs, although many, particularly AI-native companies, differentiate pricing by the type of action performed (see Figure 2). Figure 2 Five API pricing models account for how most software vendors charge for access Source: Bain analysis The right pricing model depends less on the prevailing market practice than on how agent access fits within the broader commercial model. Low-cost MCP access can inadvertently cannibalize higher-value seat licenses if customers shift work from users to agents, making it essential to evaluate pricing across the entire product portfolio rather than in isolation. Vendors must also decide who should pay. Charging end customers keeps the commercial relationship direct; charging developers or ecosystem partners can reduce customer friction but may also weaken ownership of the customer relationship over time. Finally, pricing changes should be introduced deliberately. Once developers have built against an interface, tightening access or introducing new charges has repeatedly triggered customer backlash and, in some cases, regulatory scrutiny. Where changes are necessary, a phased transition is generally more effective than an abrupt shift. Durability of decisions Pricing will continue to evolve as the market matures, but the strategic choices made around data access are far more durable. The data that gets exposed, the interfaces through which it is accessed, and the role your organization plays in agent workflows will shape competitive positioning even as pricing models change. Choosing the pricing method is the easy part. The harder decision is what you let out the door and whether doing so strengthens your moat or erodes it. Acknowledgments David Crawford, chairman of Bain & Company’s Global Technology, Media, and Telecommunications (TMT) practice; Anne Hoecker, leader of the Global TMT practice; and Dana Aulanier, vice president of the TMT practice, prepared this report. Bain Partners Cory Boles, Padraic Brick, Mark Brinda, Alessandro Cannarsi, Thibaud Chabrelie, Aaron Denman, Greg Fiore, Frank Ford, Jonathan Frick, Arun Ganti, Peter Hanbury, Jonny Holliday, Alec Koh, David Lipman, Stanley Liu, Eric Maltiel, Justin Murphy, Sandeep Nayak, Christopher Perry, Bill Radzevych, Josh Sandberg, Michael Schallehn, and Jue Wang; Associate Partners Tatum Quinn, Kristie Tagawa, and Vito Trinchera; Expert Partners Syed Ali, Mukarram Bhaiji, Purna Doddapaneni, and Chris McLaughlin; and Practice Manager Sushant Bansal wrote its chapters. The authors also wish to thank Partners Peter Bowen, Ann Bosche, Arjun Dutt, Tamara Lewis; Associate Partner Garima Nijhawan; Bain Capability Network (BCN) Director Eva Gupta; BCN Manager Vaishali Sharma; BCN Senior Manager Arpit Jain; BCN Project Leaders Anjali Mishra and Venkatesh Mishra; BCN Associates Sahil Garg, Aryan Gupta, and Tejal Janbandhu; BCN Analysts Praapti Bansal, Kankshit Kumar, Manya Singh, and Tejas vardhan T; Practice Directors Tarun Gupta and Alexander Smyth; Operations Manager Raghav Sharma; and Practice Editors Jeff Bauter Engel, Adam Jones, and David Sims for their editorial support. More from the report New Innovation Is Required to Fund AI’s $6 Trillion Buildout Software Investing in the Age of AI and Slower Growth AI Data Center Boom: Can We Build It If They Come? The $100-Billion SaaS Opportunity Hiding in Cross-System Labor Hardware Strikes Back in the AI Era Tech Services: Cracking the Code on AI-led Growth Cybersecurity’s New AI Imperative: Attacking the Backlog of Vulnerability Alerts The AI-Native Enterprise: Absorption Is the New Advantage The Half-Finished Redesign: How AI Reshapes Software Organizations Managing Token Spending without Choking Off Opportunity The Missing Architecture for Agentic Software Development How AI Is Changing Data Monetization Read our Technology Report 2026 EXPLORE THE FULL REPORT DOWNLOAD THE PDF

At a Glance

Model Context Protocol (MCP) is emerging as a new standard that allows AI agents to access enterprise software and data, prompting companies to make some critical decisions.
Deciding what data to expose means weighing the risk of opening up valuable data against the risk that customers and AI agents bypass your platform in favor of more accessible alternatives.
Deciding whether to use MCP means balancing easier integration into agent workflows against the potential for customers to accomplish more without ever engaging with your interface.
Deciding how to monetize comes last, with pricing designed to capture the value of agent access without cannibalizing higher-value revenue elsewhere in the portfolio.

Model Context Protocol (MCP), an emerging standard that lets AI agents securely connect to enterprise software and data, has gone from a niche protocol to a de facto standard at remarkable speed, growing from its late 2024 launch to more than 20,000 servers less than two years later (see Figure 1).

Bar chart tracking the cumulative number of published Model Context Protocol (MCP) servers from its November 2024 introduction through July 2026. The count grows from about 1,000 in February 2025 to about 7,000 by July 2025 and about 20,000 by July 2026. — Model Context Protocol server growth signals its rise as the new default for agentic access to data

Vendors are understandably asking how to charge for AI agents that access their data and workflows through MCP. But pricing should be the third question, not the first. Before deciding how to monetize, software companies need to first answer two more fundamental questions:

What data should we expose programmatically?
Should that access be provided through MCP?

Only then should they decide how to monetize it.

Those first two decisions determine a company's strategic position in the agent era. Pricing should rest on this foundation.

Decide what data to expose

This is the hardest decision because exposing data is often a one-way street. Once customers and partners build their systems around this data, removing it becomes very difficult. Companies should weigh the benefits against the risks for each category of data rather than treating openness as all-or-nothing.

Benefits:

meet growing customer expectations for access to the provider’s data;
become embedded inside customer workflows, increasing stickiness; and
create new monetization opportunities where customers are willing to pay.

Risks:

disintermediation as customers and competitors bypass the user interface;
greater data vulnerability as agents gain read/write access;
higher infrastructure and telemetry costs; and
regulatory issues around data sharing.

Not all data should be treated equally. Customer records, proprietary analytics, benchmark data, schemas, credentials, and workflow metadata each have different strategic value and risk profiles.

A key strategic decision is determining where your competitive advantage lies. If it’s proprietary data, then protecting access may strengthen your position. If your advantage is workflow orchestration and governance, broader access may actually reinforce your platform. The contrast between LinkedIn and ServiceNow illustrates this distinction. LinkedIn treats its professional graph as the core asset and tightly controls programmatic access; ServiceNow treats workflow execution as the asset, exposing governed workflows broadly through MCP while monetizing usage and retaining operational data.

Decide whether that access needs an MCP

Once you've decided what to expose, the next question is how to expose it.

Traditional APIs are designed for developers who write custom integrations against specific end points. MCP serves a different purpose: It provides a standard way for AI agents to discover and use an application's data and actions without requiring a bespoke integration for each vendor. As MCP adoption grows, it lowers the cost of connecting enterprise software into agentic workflows.

That creates a strategic trade-off. Offering MCP access makes it easier for customers' agents to work with your platform, boosting adoption but also making it easier for users to accomplish work without ever opening your interface. Declining to offer MCP reduces that risk but raises another: Agents may simply route around your platform in favor of vendors that are easier to integrate.

The question, then, isn't whether MCP is risky; it's which risk is greater for your business: enabling disintermediation or being excluded from emerging agent ecosystems.

One way to mitigate this trade-off is to pair MCP access with investment in your own first-party agents—that is, agents you build on your own platform, such as Atlassian has done with Rovo. This allows customers to work through your agent experience while still participating in broader agent ecosystems. In effect, you're placing two bets: You're making it easy for customers to use MCP as that ecosystem grows while also building your own agent experience to compete for those same users.

Decide how to monetize

Only after deciding what to expose and whether to offer MCP access should vendors turn to pricing. Here, the market is still experimenting rather than converging. Roughly half of vendors monetize programmatic access through usage- or outcome-based pricing, about one-third charge for additional capacity or higher rate limits, and roughly one-fifth monetize access itself. Most apply similar pricing principles to MCP and traditional APIs, although many, particularly AI-native companies, differentiate pricing by the type of action performed (see Figure 2).

Stacked bar chart (0%–100%) showing the share of software vendors by API pricing model, grouped into three labeled brackets covering five specific models. About 50% price mostly by usage or outcomes, split between pay-per-call-or-transaction and pay-per-credit-or-compute; about 30% require a tier upgrade for more capacity, split between tiered-by-throughput-speed and tiered-by-capacity; and about 20% require a tier upgrade just to access the data, via premium-tier access with no metering. — Five API pricing models account for how most software vendors charge for access

The right pricing model depends less on the prevailing market practice than on how agent access fits within the broader commercial model. Low-cost MCP access can inadvertently cannibalize higher-value seat licenses if customers shift work from users to agents, making it essential to evaluate pricing across the entire product portfolio rather than in isolation.

Vendors must also decide who should pay. Charging end customers keeps the commercial relationship direct; charging developers or ecosystem partners can reduce customer friction but may also weaken ownership of the customer relationship over time.

Finally, pricing changes should be introduced deliberately. Once developers have built against an interface, tightening access or introducing new charges has repeatedly triggered customer backlash and, in some cases, regulatory scrutiny. Where changes are necessary, a phased transition is generally more effective than an abrupt shift.

Durability of decisions

Pricing will continue to evolve as the market matures, but the strategic choices made around data access are far more durable. The data that gets exposed, the interfaces through which it is accessed, and the role your organization plays in agent workflows will shape competitive positioning even as pricing models change.

Choosing the pricing method is the easy part. The harder decision is what you let out the door and whether doing so strengthens your moat or erodes it.

Acknowledgments

David Crawford, chairman of Bain & Company’s Global Technology, Media, and Telecommunications (TMT) practice; Anne Hoecker, leader of the Global TMT practice; and Dana Aulanier, vice president of the TMT practice, prepared this report.

Bain Partners Cory Boles, Padraic Brick, Mark Brinda, Alessandro Cannarsi, Thibaud Chabrelie, Aaron Denman, Greg Fiore, Frank Ford, Jonathan Frick, Arun Ganti, Peter Hanbury, Jonny Holliday, Alec Koh, David Lipman, Stanley Liu, Eric Maltiel, Justin Murphy, Sandeep Nayak, Christopher Perry, Bill Radzevych, Josh Sandberg, Michael Schallehn, and Jue Wang; Associate Partners Tatum Quinn, Kristie Tagawa, and Vito Trinchera; Expert Partners Syed Ali, Mukarram Bhaiji, Purna Doddapaneni, and Chris McLaughlin; and Practice Manager Sushant Bansal wrote its chapters.

The authors also wish to thank Partners Peter Bowen, Ann Bosche, Arjun Dutt, Tamara Lewis; Associate Partner Garima Nijhawan; Bain Capability Network (BCN) Director Eva Gupta; BCN Manager Vaishali Sharma; BCN Senior Manager Arpit Jain; BCN Project Leaders Anjali Mishra and Venkatesh Mishra; BCN Associates Sahil Garg, Aryan Gupta, and Tejal Janbandhu; BCN Analysts Praapti Bansal, Kankshit Kumar, Manya Singh, and Tejas vardhan T; Practice Directors Tarun Gupta and Alexander Smyth; Operations Manager Raghav Sharma; and Practice Editors Jeff Bauter Engel, Adam Jones, and David Sims for their editorial support.