Hello! I'm your financial analysis assistant. This article by Wang Zhiyuan discusses a very profound and ongoing shift in business logic: AI employees (agents) are already in use within companies, but how to manage them and how to bill them has become a new battleground.
In the past, when we bought software, we were purchasing tools; now, we are buying results and control over those tools. Let me break down this news into five key points to help you fully understand the business dynamics surrounding these "AI stewards."
1. Current Situation: AI Employees Abound, but Management is Chaotic
First, let's understand the extent of the chaos. There's a striking statistic in the article: on average, each company is running more than 12 AI agents. It's like a company that suddenly hires a bunch of interns, half of whom work independently, without any coordination, and the boss doesn't even know their names or what they're doing.
In response, established software giants like SAP and Salesforce have divided into two camps:
- The Traditional Camp (Gatekeepers): Represented by SAP, their approach is "I control my territory." SAP has changed the rules, requiring external AI to use their designated channels (Joule) to connect to their systems. They even charge a fee for each API call. SAP has this power because they hold 50 years of financial and supply chain data. Trying to replace SAP would be extremely costly and practically impossible.
- The Open Camp (Innovators): Represented by Salesforce, their strategy is to make their systems more accessible. They have made all their functions available as APIs, allowing external AI to use them freely. They even suggest that users can manage data directly through chat apps without logging into Salesforce. They can afford to be open because the cost of switching from CRM (Customer Relationship Management) to ERP (Enterprise Resource Planning) systems like SAP is much lower, and customers are less dependent on these systems.
Core Logic: If customers can't leave, they charge a "toll"; if they can, they retain them through openness.
2. The New Business Model: From Selling Software to Providing Management Services
Although the two camps have different approaches, both are ultimately aiming to control AI within companies. Why? Even though 72% of companies think they manage AI well, when asked about its costs or effectiveness, most can't answer (only 13% claim effective management). Many companies have set up AI systems without involving IT departments, leading to a mess.
As a result, a new business model has emerged: selling management services rather than just software. Giants like Microsoft (Agent 365), ServiceNow, Workday, Google, and Atlassian have launched services to help companies manage their AI:
- Microsoft (Agent 365): Issues "IDs" and "work logs" for each AI agent. Customers pay not to buy software but to track the AI's activities.
- ServiceNow: Provides a "control tower" to monitor all external AI actions, charging per action performed.
- Workday: Quantifies AI activities, deducting fees based on actions like generating reports or screening resumes.
- Google & Atlassian: Have started pricing services for semantic governance and context-based usage.
In simple terms: Instead of buying a hammer (software), you're hiring a supervisor (management service) that not only monitors the AI but also keeps track of its activities. Whoever controls the access (interfaces/gateways) controls the revenue.
3. Billing Traps: From Monthly Subscriptions to Pay-As-You-Go, with Frequent Unexpected Bills
The biggest issue for companies is billing. Traditional software came with monthly subscriptions, but AI software now uses a pay-as-you-go model, similar to utilities. Here are some common pitfalls:
- Hidden Price Increases: Many vendors only specify the monthly fee in the contract, but the usage limits and overage charges are hidden in additional documents. A change in the documentation can significantly increase the price without notice.
- Unexpected Bills: 78% of IT managers have encountered unexpected bills. For example, a GitHub model's price soared from $7.5 to $27 within two months, discovered only when the bill was received.
- More Power, More Cost: AI services like Intercom charge based on the number of issues resolved. The smarter the AI, the more it costs. Instead of saving labor, companies end up paying for the number of tasks completed.
Core Logic: Bills have become unpredictable. What were once fixed costs have turned into variable costs, and since AI runs automatically, it's easy for expenses to get out of control. Some companies spend up to $500 million on AI in a month, often due to the accumulation of small charges.
4. Buyers' Counteraction: Seeking Outcome-Based Pricing and Transparency
Companies are reacting to this opaque billing. They realize that simply switching suppliers isn't enough; they need to change how they pay:
- Outcome-Based Pricing: Only charge if the goal is achieved.
- HubSpot: Charges based on results.
- Intercom: Was acquired by Salesforce for $3.6 billion because of its outcome-based pricing model.
- Sierra: Increased revenue by charging based on results.
- Zendesk: Charges only after verifying the effectiveness of solutions.
- Transparency: Companies demand clear billing. Nearly 30% can't account for their AI expenses, so some set limits or allocate costs to departments. Netflix's engineers even released open-source tools to monitor AI usage.
- Changing Suppliers: More than half of companies plan to switch or add new AI providers within a year if the terms aren't clear.
In simple terms: Buyers want to see where their money goes and are willing to pay only for results.
5. Future Trends: Profit Shifts from Smart Technology to Effective Management
Wang Zhiyuan predicts that by 2030, 40% of software revenue will come from usage, actions, and results. This means the source of profit is shifting:
- Previously: Profit came from advanced software features and algorithms.
- Now and in the Future: Profit will come from the ability to effectively manage AI.
The battle continues: vendors are developing new billing systems, and buyers are using monitoring tools. Whoever can clearly demonstrate the value of their services will set the prices for the next round.
In summary: AI has entered the workplace, but the power to manage and bill it is shifting from software sellers to those who control the interfaces, data, and bills. For companies, understanding bills and gaining control over AI management is more important than just focusing on the intelligence of the technology. The next two years will be critical for shaping the AI management services market.