Vendors Start to Productize Agentic AI at Scale, but Monetization Remains a Significant Challenge

The Agentic Artificial Intelligence (AI) market remains in an early stage with no clear leadership. But the market is being flooded with Agentic AI solutions from the usual suspects—hyperscalers, System Integrators (SIs), data and AI analytics platforms, Independent Software Vendors (ISVs), and many others.

But each of these vendors faces a trade-off between commercial scalability and value creation. On one end of the market, new entrants are providing low/no-code Agentic AI platforms that are enabling low AI-skilled workers to quickly build simple Agentic AI tools or workflows for specific processes or applications. These platforms have quickly taken off commercially, but over the long term, they will provide limited value to end users.

On the other end of the market, SIs are developing custom Agentic AI systems tightly integrated into enterprise systems. These are hugely valuable and provide a foundation for enterprises to scale Agentic AI solutions across various business processes, but they face huge talent bottlenecks, which limits their Serviceable Addressable Market (SAM). This dilemma creates a challenge for vendors entering the market space, with many opting for mixed strategies that combine “horizontal platforms” with either verticalized channel partners, SIs, or even business process experts:

• AWS/Capita: Capita is a Business Process Outsourcing (BPO) company with extensive U.K. government contracts. It has partnered with Amazon Web Services (AWS) to build CapitaContact, which supports the implementation of Agentic AI contact center solutions for customer experience use cases. Capita’s expertise in BPO provides a deep process-driven understanding, while reducing development costs by leveraging a horizontal platform.
• NVIDIA/Accenture: Accenture has launched a range of Agentic AI frameworks and tools based on NVIDIA AI Enterprise. This includes the Distiller Agentic AI Framework, which can be deployed across multiple industry verticals. In addition, Accenture has launched and expanded the AI Refinery platform built on top of the NVIDIA AI software stack.
• Cognizant/Lineage: Built a partnership to enhance cold chain logistics applications. The collaboration aims to deliver enhanced resources, and reliable service models for Agentic AI solutions to support customer care organizations and service Lineage’s customers. Cognizant brings the Agentic AI capabilities to support Lineage’s customer engagement.
• Wipro/Google Cloud: The partnership aims to co-develop 200 production-ready Agentic AI solutions across different industries. These co-developed agentic solutions are available through Google Cloud Marketplace as they aim to drive broader consumption and solution adoption.

A partner-led product strategy makes the most sense for a variety of reasons. First, SI customer teams are much closer to the end problem, which shifts the “sales process” from technology development to solving a business problem. Second, the horizontal technology platform provides an extensible tool that can be effectively turned into a real solution without end-to-end development for each customer. Third, for Agentic AI depth of integration and implementation has a direct impact on outcomes and Return on Investment (ROI), which ultimately is the biggest determinant of value. SIs need to deploy and integrate Agentic AI systems with multiple “process-specific” models, access to applications, proprietary datasets, and running according to specific Service-Level Agreements (SLAs). Fourth, Agentic AI implementation is not just about technology, but about business systems that include humans, regulation, governance, controls, and other core business processes. This operational intensity requires skilled co-development, co-Go-to-Market (GTM), and co-implementation partners, making SIs the obvious choice.

Even as these vendors build out their “partner-led” product strategy, monetization remains an open-ended question for many. There are multiple challenges facing vendors:

• Customers need predictable pricing, but running costs vary massively depending on usage, system complexity, SLA requirements and other factors. This inhibits a fixed cost model, unless vendors are just building solutions and handing off running costs to the customer.
• ROI is not clear with enterprise usage and other factors impacting the real value being created. This has led to a risk-averse mindset for many Chief Information Officers (CIOs).
• Open-source development continues to lower barriers to entry, increasing the number of substitute products available. Most vendors do not have an effective Agentic AI product moat that can be monetized; most currently rely on existing products, platforms, or internal capabilities (e.g., vertical data) to build differentiation.
• Multi-vendor, multi-agent ecosystems create commercial confusion. Most expect requirements for profit sharing or other arrangements within Agentic AI systems, but this remains limited.
• An effective GTM strategy requires developing a partner ecosystem with SIs, vertical or process experts, and others deeply embedded within product development and commercialization. This creates further profit drain implications given the scope of these partner ecosystems, especially for hyperscalers that control cloud resources.

These new challenges add further complexity to a market that has struggled to build an effective monetization strategy for Generative Artificial Intelligence (Gen AI). Chatbots and copilots, powered by Large Language Models (LLMs), have fallen into the “value add” trap. The market quickly saturated, creating commoditization and competitive pressures forcing vendors to provide Gen AI at best in a “freemium” model, or worse as a “value-add” tool. Given the much higher running cost implications of Agentic AI systems, this “value-add” commercial strategy cannot continue, and vendors must build better monetization strategies.

The challenges highlighted above continue to push vendors toward “easy” consumption-based pricing models with customers consuming Agentic AI solutions through developer platforms, e.g., Dataiku or marketplaces, e.g., AWS or Google Cloud Marketplace. But this model has huge disadvantages:

1. Consumption-based models will reduce enterprise usage. As more employees use Agentic AI tools, usage will grow, pushing up costs; however, this increased usage will not lead to immediate value creation that could contribute to customer attrition. Users will be increasingly afraid to utilize tools without clear value being created.
2. Consumption-based pricing models are often opaque with bills determined by Application Programming Interfaces (API) calls, data fetching, etc. This will cause confusion within enterprises without complete transparency.
3. Agentic AI systems are often inefficient with customers negatively impacted by failures. For example, Agentic AI solutions may make multiple API calls, which increase token consumption driving up price and reducing ROI.
4. Risk-adverse customers drawn to consumption-based pricing models will not commit Capital Expenditure (CAPEX) upfront to support deployment of tightly integrated solutions, which limits the value created from Agentic AI.
5. Agentic AI works best in systems with tight integration between different agents, datasets, etc. Consumption-based agents deployed through marketplaces or from data and analytics platforms lack tight integration (although open-source protocols have made strides to enable connection). This impacts governance, orchestration, transparency, and operations—all of which influences value creation.

Instead, vendors should look to develop outcome-based pricing that ties cost to a specific and measurable outcome. For Agentic AI, this pricing model could be tied to a variety of metrics: ticket resolutions, SLAs, security, reliability, or process accuracy. A handful of vendors are in an early stage of implementing this pricing model. Zendesk charges customers for issues fully resolved by AI agents, but if a human steps in, then there is no charge; Intercom charges on a pay-per-success model, which means customers only pay when AI successfully resolves a support conversation.

The benefits of this pricing model for customers are obvious—they only pay for quantifiable outcomes, which means that they can directly link cost to value, making the ROI conversation more transparent and incentivizing deployment and usage. It also makes Total Cost of Ownership (TCO) more predictable and de-risks investment, given the lower CAPEX requirements. For vendors, it builds differentiation within a saturated marketplace, while also ensuring deeper alignment with customers as they need to engage to determine desired outcomes. These tighter “advisory” relationships will likely support organic contract growth. Of course, this pricing model comes with risks—especially financial risks due to delayed, lacking, or even disputed outcomes. But overall, this pricing model will offer longer-term benefits for both customers and vendors.