---
title: "Tokenomics for Bankers: Rethinking the Business Case for AI Agents"
description: Explore how banks can rethink AI cost structures as generative agents reshape operational expenses and budget strategies. Discover strategies for cost management.
image: https://ked-consulting.com/hubfs/AI-Generated%20Media/Images/Brass%20Utility%20Meter%20In%20Bank%20Boardroom%20At%20Dusk.png
---

[KEDConsulting](https://ked-consulting.com/?hsLang=en)

[Home](https://ked-consulting.com/?hsLang=en)[Advisory](https://ked-consulting.com/advisory?hsLang=en)[Solutions](https://ked-consulting.com/solutions?hsLang=en)[About](https://ked-consulting.com/about?hsLang=en)[KEDInsights](https://ked-consulting.com/insights?hsLang=en)[Contact](https://ked-consulting.com/contact?hsLang=en)

[Book a consultation](https://ked-consulting.com/book-a-consultation?hsLang=en)

[HOME](https://ked-consulting.com/?hsLang=en) / [INSIGHTS](https://ked-consulting.com/insights) / [ARTIFICIAL INTELLIGENCE](https://ked-consulting.com/insights/tag/artificial-intelligence)

KEDInsights

# Tokenomics for Bankers: Rethinking the Business Case for AI Agents

By **KED Consulting** November 17, 2026 4 min read

There is something curious about the way banks talk about the cost of AI. The conversation almost always turns to the price of the licence, the size of the pilot and the payback date in the business case. Far less is said about what it costs each time an agent thinks.

With agentic AI, that second number is quietly becoming the one that matters.

### The Budget Was Built for a Different Machine

Most bank technology budgets were designed for software that costs roughly the same whether it is used once or a million times. You buy it, implement it and depreciate it. Generative AI bent that pattern. Agents break it.

An agent that reviews a credit file does not run a single query. It plans, retrieves documents, calls tools, checks its own work and sometimes asks another agent for help. Every step consumes tokens, and every token is billed.

The obvious interpretation is that this is a vendor pricing issue, to be settled at the negotiating table. In practice, the situation is more complicated. **Most of the bill is written by the design of the workflow, long before procurement sees it.**

### A Cost Line That Moves on Its Own

McKinsey's latest Global Banking Annual Review puts a number on it. Spending on tokens, compute and AI operations can reach as much as 20 percent of a bank's labour costs, which is considered "a material and volatile cost line". Early controls, such as routing simple queries to cheaper models, stop working once agents talk to each other, because tokens quickly add up.

None of this is an argument against agents. The same report found an average payback of about ten months across 52 agentic use cases. What has changed is that value and cost now move together, and cost moves faster than finance processes were built to notice.

Research outside banking echoes this. In one vendor survey, 72 percent of firms reported a surprise AI bill in the past year.

### The Roaming Bill Problem

Many executives remember the early days of international mobile data. The phone worked perfectly abroad. Maps loaded, emails synced, nothing felt unusual. The shock came a month later, when the bill arrived.

The phone was never faulty. Consumption was simply invisible at the moment it happened, priced by a meter nobody was watching.

Consider a hypothetical bank whose credit-document agent cost very little during a three-month pilot. In production it handles ten times the volume, retries whenever a document is unclear and calls a second agent to check compliance. Each design decision is reasonable on its own. Together they can multiply the cost per file, and the first signal is an invoice.

### Why Indonesian Banks Should Care Now

According to a PERBANAS and IBM survey, AI adoption in Indonesian banking reached 57.9 percent in Q1 2026, up from about 30 percent in 2024. Many banks are now crossing from pilot to production, which is precisely when architecture choices harden and become expensive to reverse.

There is also a local twist. Most frontier AI services are priced in US dollars while bank budgets are set in rupiah, so a cost line that is already volatile carries currency exposure as well.

Gartner expects inference, the cost of running AI rather than building it, to account for 55 percent of AI-optimised cloud spending in 2026. The money is moving to the run phase. **Governance has to move with it.**

### Where the Real Leverage Lies

1. **Price the task, not the tool.** Set a target cost per completed task, such as a reviewed credit file or a resolved complaint, as a design requirement. A practical first step is to measure tokens and tool calls per task for the three most mature use cases.
2. **Put cost on the architecture review agenda.** Ask how an agent knows when to stop, how much context it carries and when it is allowed to call another agent. McKinsey identified more than 20 architectural patterns that lower spend. Choose the few that fit the bank's platform.
3. **Give every production agent an owner and a visible bill.** A monthly showback to the business owner changes behaviour faster than any policy. The capability often already sits in technology, since 78 percent of FinOps teams now report to the CIO or CTO.
4. **Treat volatility as a risk rather than a surprise.** Add currency exposure and vendor dependence to the technology risk register, and negotiate price predictability into contracts.
5. **Re-run the business case after 90 days in production.** Compare the actual cost per task with the pilot estimate before scaling further.

### The Meter Was Always Running

The question worth asking in every bank is whether anyone can see the meter while it is running. **In the agentic era, cost discipline will either be designed into systems or discovered on invoices.** Banks can still choose which, though perhaps not for much longer.

Topics [Artificial Intelligence](https://ked-consulting.com/insights/tag/artificial-intelligence)

Share [LINKEDIN](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fked-consulting.com%2Finsights%2Ftokenomics-for-bankers-rethinking-the-business-case-for-ai-agents) [X](https://x.com/intent/post?url=https%3A%2F%2Fked-consulting.com%2Finsights%2Ftokenomics-for-bankers-rethinking-the-business-case-for-ai-agents&text=%3Cspan+id%3D%22hs_cos_wrapper_name%22+class%3D%22hs_cos_wrapper+hs_cos_wrapper_meta_field+hs_cos_wrapper_type_text%22+style%3D%22%22+data-hs-cos-general-type%3D%22meta_field%22+data-hs-cos-type%3D%22text%22+%3ETokenomics+for+Bankers%3A+Rethinking+the+Business+Case+for+AI+Agents%3C%2Fspan%3E) [FACEBOOK](https://www.facebook.com/sharer/sharer.php?u=https%3A%2F%2Fked-consulting.com%2Finsights%2Ftokenomics-for-bankers-rethinking-the-business-case-for-ai-agents) [EMAIL](mailto:?subject=%3Cspan+id%3D%22hs_cos_wrapper_name%22+class%3D%22hs_cos_wrapper+hs_cos_wrapper_meta_field+hs_cos_wrapper_type_text%22+style%3D%22%22+data-hs-cos-general-type%3D%22meta_field%22+data-hs-cos-type%3D%22text%22+%3ETokenomics+for+Bankers%3A+Rethinking+the+Business+Case+for+AI+Agents%3C%2Fspan%3E&body=https%3A%2F%2Fked-consulting.com%2Finsights%2Ftokenomics-for-bankers-rethinking-the-business-case-for-ai-agents)

Keep reading

## Related insights.

### [Artificial Intelligence The Expensive Comfort of a Human in the Loop Nov 7, 2026READ →](https://ked-consulting.com/insights/the-expensive-comfort-of-a-human-in-the-loop?hsLang=en)

### [Artificial Intelligence AI in Banking: From Hype to Realization Oct 18, 2025READ →](https://ked-consulting.com/insights/ai-in-banking-from-hype-to-realization-practical-steps-for-cios-and-cdos?hsLang=en)

### [Artificial Intelligence Opportunities and Challenges of AI Adoption in the Telecom Industry Jan 3, 2025READ →](https://ked-consulting.com/insights/opportunities-and-challenges-of-ai-adoption-in-the-telecom-industry?hsLang=en)

Next step

## Turn insight into action.

Talk through your technology challenges with our advisors — free, and with no obligation.

[Book a free consultation →](https://ked-consulting.com/meetings/ked-consulting/meeting-with-ked?hsLang=en) [Contact us](https://ked-consulting.com/contact?hsLang=en)

[KEDConsulting](https://ked-consulting.com/?hsLang=en)

Your trusted partner in IT strategy, enterprise architecture and governance, turning complexity into clarity.

 Indonesia Stock Exchange Tower 1 Level 3  
Jl. Jend. Sudirman Kav 52-53, Jakarta Selatan  
Indonesia 12190  
[+62 (0) 21 2960 7267](tel:+622129607267)  
[connect@ked-consulting.com](mailto:connect@ked-consulting.com)

#### Advisory

[Enterprise Architecture](https://ked-consulting.com/enterprise-architecture-advisory?hsLang=en)[Cloud Advisory & FinOps](https://ked-consulting.com/cloud-advisory?hsLang=en)[Architecture Review](https://ked-consulting.com/architecture-review?hsLang=en)[IT Governance & Risk](https://ked-consulting.com/itgrc-advisory?hsLang=en)[AI Advisory](https://ked-consulting.com/ai-advisory-ked-consulting?hsLang=en)

#### Solutions

[EA Management](https://ked-consulting.com/ea-management-platform?hsLang=en)[IT GRC Platform](https://ked-consulting.com/itgrc-platform?hsLang=en)[Technical Debt](https://ked-consulting.com/technical-debt-management?hsLang=en)

#### Company

[About us](https://ked-consulting.com/about?hsLang=en)[KEDInsights](https://ked-consulting.com/insights?hsLang=en)[Contact](https://ked-consulting.com/contact?hsLang=en)[Privacy Policy](https://ked-consulting.com/privacy-policy?hsLang=en)[Terms & Conditions](https://ked-consulting.com/tnc?hsLang=en)

© 2026 KED CONSULTING

[LINKEDIN](https://www.linkedin.com/company/karya-ekosistem-digital/)[FACEBOOK](https://www.facebook.com/profile.php?id=61560968952272&mibextid=LQQJ4d)[X](https://x.com/kedconsulting)[INSTAGRAM](https://www.instagram.com/ked.consulting)

```json
{
  "@context" : "https://schema.org",
  "@type" : "BlogPosting",
  "author" : {
    "@type" : "Person",
    "name" : "KED Consulting",
    "url" : "https://ked-consulting.com/insights/author/ked-consulting"
  },
  "dateModified" : "2026-11-17T09:33:07.367Z",
  "datePublished" : "2026-11-17T09:33:07.000Z",
  "headline" : "Tokenomics for Bankers: Rethinking the Business Case for AI Agents",
  "image" : [ "https://ked-consulting.com/hubfs/AI-Generated%20Media/Images/Brass%20Utility%20Meter%20In%20Bank%20Boardroom%20At%20Dusk.png" ],
  "mainEntityOfPage" : {
    "@id" : "https://ked-consulting.com/insights/tokenomics-for-bankers-rethinking-the-business-case-for-ai-agents",
    "@type" : "WebPage"
  },
  "publisher" : {
    "@type" : "Organization",
    "logo" : {
      "@type" : "ImageObject",
      "url" : "https://ked-consulting.com/hubfs/KEDConsulting-round-1.png"
    },
    "name" : "KED Consulting"
  }
}
```