The Question
Every enterprise agent platform vendor has a capability demo that impresses. The agent routes tickets accurately, handles customer inquiries fluently, processes documents faster than a human team. The demo is compelling. The meeting with finance is not. Finance does not approve capability demos — it approves quantified returns on quantified investments over a specified period, compared against a clearly defined alternative. Most enterprise agent business cases are written by technology teams, reviewed by technology leadership, and then rejected or significantly delayed by finance because they present capability instead of economics.
The disconnect is structural. Technology teams understand what the agent can do and are excited by it. Finance understands what the investment costs and what it returns, and needs to see that the return exceeds the cost by a margin that justifies the commitment. These are not opposing interests — they require different languages. The business case that gets approved translates agent capability into workflow economics, workflow economics into annual cost or revenue impact, and annual impact into a three-year TCO comparison that finance can evaluate against other capital allocation options.
This brief provides the framework for building that business case — the ROI quantification methodology, the TCO structure, and the presentation approach that works with a finance audience. The framework is grounded in what has worked in enterprise agent program approvals at organizations that have deployed agents at scale in 2025 and 2026.
The enterprise agent business case that gets approved is built on measured pilot data, specific workflow economics, and conservative three-year TCO — not on AI capability narratives, and the organizations that present capability instead of economics consistently get a longer review cycle.
Why This Matters Now
Enterprise agent platform investment decisions are moving up the capital allocation stack. In 2024, most enterprise agent decisions were departmental — a customer service team buying a conversational agent, an IT team deploying a helpdesk automation tool. The investment was sized for a department budget and approved by a business unit leader. In 2025 and 2026, the pattern has shifted. Organizations are making enterprise-wide platform decisions — selecting a primary agent platform, committing to multi-year licensing, and investing in the integration and change management required to deploy across multiple business functions. These are CFO-level decisions requiring board-level capital allocation.
The investment scale is significant. Gartner's 2025 analysis of enterprise agent platform TCO found that a mid-market enterprise deploying an agent platform across three business functions over three years has a typical all-in investment of $2.1M to $4.8M — including platform licensing, integration development, change management, and ongoing operations. That investment requires a rigorous business case, not a technology evaluation.
The ROI data from organizations that have deployed at scale is strong but must be quantified correctly to be credible. Salesforce's 2026 Agentforce ROI study of enterprise customers found median payback periods of 14 months and median three-year ROI of 287% — but only for deployments where the business case was built on measured pilot data. Organizations that projected ROI based on vendor benchmarks rather than their own pilot data showed significantly higher variance in actual outcomes.
McKinsey's 2025 analysis of enterprise AI ROI found that the primary determinant of business case accuracy was whether pilot data was used to calibrate projections. Organizations using pilot data showed actual ROI within 20% of projected ROI. Organizations using vendor benchmarks showed actual ROI that varied by more than 60% from projections in either direction.
What the CURVE™ Data Shows
The 2026 Stackcurve AI Enterprise Agent Platform CURVE™ Report evaluated the ROI documentation, business case tooling, and customer economic data provided by enterprise agent platform vendors. The assessment covered Salesforce Agentforce, ServiceNow Now Assist, Microsoft Copilot Studio, IBM watsonx Orchestrate, Google Agentspace, AWS Bedrock Agents, UiPath Autopilot, and Automation Anywhere.
Salesforce provides the most comprehensive ROI documentation in the market, including the Agentforce Success Index with customer economic data, a structured ROI calculator for the business case process, and Trailhead-based business case training for implementation partners. ServiceNow provides a TCO calculator and ROI benchmarking data through its Value Management Office, with customer reference data segmented by industry and use case. Microsoft provides the Copilot Impact Dashboard — aggregate ROI data from the Copilot for Microsoft 365 program — but agent-specific ROI documentation is less mature than Salesforce's.
IBM watsonx Orchestrate provides strong industry-specific ROI benchmarking, particularly for financial services and healthcare use cases where IBM has deep customer data. UiPath and Automation Anywhere, both with roots in RPA, have the most mature cost-per-task benchmarking data given their longer history of workflow automation economics.
AWS and Google provide the least structured business case support, reflecting their infrastructure-provider positioning — the business case for their agent platforms is typically built by the customer or their system integrator, not by AWS or Google.
The full vendor rankings are in the 2026 Stackcurve AI Enterprise Agent Platform CURVE™ Report — free to download.
The Gap Most Buyers Miss
Most enterprise agent business cases present one ROI scenario — the optimistic one. Finance applies a haircut. The resulting approved investment is insufficient to execute the program as planned. The alternative is to present three scenarios grounded in pilot data, which demonstrates analytical rigor and gives finance a range to evaluate rather than a single point to discount.
Labor Efficiency: The Primary ROI Driver
Labor efficiency is the most direct and easiest-to-quantify ROI driver, and it should lead every enterprise agent business case.
The quantification method: identify the specific workflow the agent will handle. Measure the current fully-loaded cost per workflow execution (time × fully-loaded labor rate, including benefits and overhead). Measure or project the agent handling rate — the percentage of workflow executions the agent will complete without human intervention — based on pilot data. The annual savings is: (total annual workflow volume × agent handling rate × current cost per workflow) minus the agent cost per workflow.
Example: a customer service agent handles L1 support inquiries. Current L1 cost is $18 per inquiry (10 minutes of agent time at $65/hour fully loaded), and the organization handles 600,000 L1 inquiries per year. Pilot data shows a 72% autonomous handling rate. Annual labor savings: 600,000 × 0.72 × $18 = $7.78M. If the agent platform costs $2.50 per inquiry for the handled volume, the net annual savings is $6.70M. This is a finance-approachable number with a clear methodology.
Throughput Increase: The Growth Case
Some workflows are bottlenecked by human capacity — the organization could process more volume if it had more people. An agent that removes the capacity constraint allows volume to grow without proportional headcount growth. Quantify this as revenue acceleration (the additional revenue that can be captured by removing the throughput bottleneck) or cost avoidance (the headcount that would have been required to handle the projected volume increase without an agent).
This is the growth-focused version of the business case, which resonates with revenue-oriented stakeholders who are less motivated by cost reduction than by capability expansion.
Error Rate Reduction: The Risk Case
For high-volume, repetitive tasks — data entry, document classification, compliance checking — agents have lower error rates than humans under sustained load. Error rate reduction translates directly to cost reduction (cost of error correction) and risk reduction (compliance violations, customer impact from incorrect decisions).
Quantify by measuring the current error rate and the fully-loaded cost per error (correction labor + any downstream cost or compliance penalty). Project the agent error rate from pilot data. The annual savings is the difference in error volume times the cost per error.
The TCO Structure
Year 1 is the highest-cost year and the lowest-savings year: platform licensing + integration development + change management + training, with partial-year savings from agents that go live mid-year. Year 2 is the inflection point: full savings realized, declining integration costs, maturing operations. Year 3 is the steady-state economics.
The three-year TCO comparison should present three columns: current state (full manual process cost), agent-augmented (human + agent for high-complexity workflows, agent-primary for high-volume routine workflows), and agent-primary (full autonomous handling at the agent's capability limit). Most organizations are targeting the agent-augmented model for Year 2-3, with the agent-primary model as a future state.
How to Present to Finance
Lead with cost reduction in Year 1 and Year 2 — finance responds to cost reduction because it is direct and measurable. Add the productivity story for growth-focused stakeholders: not cost reduction but capability expansion — the same team can handle more volume, serve more customers, process more transactions. Close with risk reduction: agents have lower error rates, produce immutable audit trails, and apply policy consistently — risk reduction that has value even if the financial ROI were lower than projected.
Present conservative, base, and optimistic scenarios. Conservative uses the lower bound of pilot data outcomes. Base uses pilot data directly. Optimistic projects 15-20% improvement over pilot from operational maturity. Finance will evaluate the conservative scenario; the base and optimistic scenarios demonstrate upside and show analytical rigor.
Questions Your Buying Team Should Be Asking
1. Does the vendor provide customer ROI data segmented by industry and use case, and can they connect us with a reference customer at comparable scale?
Vendor ROI benchmarks are averages that may not reflect your industry, use case, or organizational context. Ask for ROI data segmented by industry and use case type, and ask for a reference customer — not a case study, but a direct conversation with a peer organization that has deployed the platform at comparable scale. The reference conversation is where you learn the honest economics, including what cost factors the published case study omits.
2. Does the vendor provide a structured ROI calculator or business case tool, and is it calibrated with customer data or vendor assumptions?
ROI calculators that use vendor-supplied default values (average handling rate, average time savings per task) will overstate your likely ROI. Ask whether the calculator can be calibrated with your pilot data, and ask what the default values are based on — vendor benchmarks or aggregate customer data. The difference matters significantly for business case credibility with a finance audience.
3. What are the Year 1 implementation costs beyond platform licensing — specifically integration development, change management, and training?
Platform licensing is typically the smallest component of Year 1 TCO. Integration development (connecting the agent platform to your existing systems) and change management (user training, process redesign, adoption support) are frequently 2–4x the platform licensing cost in Year 1. Ask the vendor or implementation partner for a detailed Year 1 cost breakdown, and compare it to your internal estimates. If there is a significant gap, understand why before committing.
4. What is the operational cost structure in Years 2 and 3 — specifically, what does ongoing platform administration, monitoring, and optimization cost?
A business case built only on Year 1 and Year 2 economics misses the steady-state cost structure. Ask for the Year 3 total cost of ownership, including platform licensing growth (most contracts have annual escalation clauses), ongoing integration maintenance, and the internal agent operations function cost. The steady-state economics should show a healthy margin over the current-state cost to justify the ongoing commitment.
5. How does the vendor support business case development and finance review — do they provide resources for presenting to a CFO or board audience?
Vendors with mature enterprise programs have learned that business case support accelerates sales cycles and improves customer outcomes. Ask whether the vendor provides resources — ROI documentation, finance presentation templates, economic analysis support — for the business case process. Vendors that only provide capability documentation and leave the business case to the customer are less invested in the customer's success and less likely to have the customer economic data that makes the business case credible.
The Stackcurve Take
The enterprise agent business case that gets approved in 2026 is built on three things that most technology teams underinvest in: measured pilot data that calibrates projections, a rigorous three-year TCO that includes all cost categories, and a presentation structure that leads with finance-appropriate metrics. The capability narrative — what the agent can do — is the context, not the argument. The argument is the economics.
The labor efficiency calculation is the foundation of every enterprise agent business case. It is direct, measurable, and immediately credible to a finance audience. Layer the throughput and error rate stories for stakeholders whose priorities extend beyond cost reduction. Build the three-scenario structure to demonstrate analytical rigor and give finance a range to evaluate rather than a single optimistic projection to discount.
The pilot is not just a technical proof of concept — it is the data collection phase for the business case. Organizations that design their pilots to capture the metrics required for the business case (task completion rate, handling rate, cost per task, error rate) have a significant advantage over those that treat the pilot as a capability demonstration and then reconstruct economics from incomplete data.
The 2026 Stackcurve AI Enterprise Agent Platform CURVE™ Report covers vendor ROI documentation, customer economic data, and business case support across the enterprise agent platform market. Download it free →
Stackcurve Advisory Briefs are independent research. No vendor pays for placement, tier assignment, or editorial influence. The CURVE™ methodology is disclosed in full at stackcurve.net/research/methodology.