---
title: "Wharton and Accenture put a five-year price tag on enterprise AI agents: up to $520 million"
description: "The Wharton-Accenture report prices the enterprise AI agent for the first time, and the derived figure runs to $2.97 million per agent per year. Tokens are a fifth to a third of that bill; the rest is integration work Accenture sells."
publisher: "The Inference"
section: "Applications"
published: 2026-11-17T17:10:49.495Z
modified: 2026-11-17T17:10:49.495Z
canonical: https://theinference.org/article/wharton-and-accenture-put-a-five-year-price-tag-on-enterprise-ai-agents-up-to-520-million
language: en
keywords: "AI Agents, Accenture, Enterprise AI, Consulting"
---

# Wharton and Accenture put a five-year price tag on enterprise AI agents: up to $520 million

> The Wharton-Accenture report prices the enterprise AI agent for the first time, and the derived figure runs to $2.97 million per agent per year. Tokens are a fifth to a third of that bill; the rest is integration work Accenture sells.

When companies put AI agents to work, they reach for the newest, most powerful model and let it handle everything. A joint report from the Wharton School and [Accenture](https://theinference.org/markets/companies/accenture), "[What Does Agentic AI Really Cost?](https://finance.yahoo.com/technology/ai/articles/most-doesn-t-ferrari-ai-140000997.html)", published November 16, counted more than 18,000 tasks across 923 occupations in 19 industries and found that more than 90% of the work agents could take over is routine enough for cheaper, less powerful models [1][2].

In every industry studied, only 6% to 11% of agent-capable work needs a frontier model, meaning one of the top-tier systems at the edge of current AI capability, which carry premium pricing [1][2].

"I think they're taking possibly their most sophisticated model and applying it to all of the workflow," said Eric Bradlow, Wharton's vice dean of AI and analytics and a coauthor [2].

## 35 agents, $520 million, five years

The report also prices what enterprises rarely count. For a company with 100,000 employees running 35 agents, estimated five-year costs run from about $155 million in health care to about $520 million in insurance [2]. Divided out, that is roughly $886,000 to $2.97 million per agent per year, arithmetic the authors leave to the reader [2].

## The token is the tip

Tokens, the text units providers bill by, are only 20% to 33% of the total cost of owning an agent system; orchestration, human oversight and integration take the rest [1]. Integration is the business Accenture sells. Coauthor James Crowley's diagnosis, that "people tend to use the model that they already have access to rather than what's best suited to the task," describes a lock-in that dynamic routing would break, and routing that kind of work would be [billable consulting for a firm like Accenture](https://theinference.org/article/pwc-deloitte-and-bdo-india-put-model-ml-s-ai-on-live-client-work-inside-a-10-billion-consultancy-ai-bet) [2].

The report's own warning sits beside it: fewer than one-third of organizations report sustained, board-level value from agentic AI [1], and Deloitte counts just 15% of organizations at scaled, orchestrated multi-agent deployments [4].

## Not an argument against frontier models, Crowley says

"It's not an argument against frontier models," Crowley says; they remain essential for complex, high-stakes work [2]. Bradlow concedes the per-task math: "I'm asking it to do 100 different things I wasn't asking it to do a year ago [2]. So my total cost is going up," even as per-task cost falls [2].

## Accenture beat the quarter, then guided to 3% to 6%

Accenture's [fiscal fourth quarter, reported October 1](https://www.insidermonkey.com/news/accenture-acn-soars-15-8-as-q4-earnings-ai-bookings-crush-disruption-fears-1847513/), beat on both lines: revenue of $18.68 billion, up 6.25%, new bookings of $22.2 billion, up 4%, and net income up 40.8% to $1.99 billion; the stock jumped 15.78% that day and [guidance for fiscal 2027 is 3% to 6% revenue growth](https://theinference.org/article/ai-is-cutting-accenture-s-prices-acquisitions-carry-2-to-2-5-points-of-its-2027-growth) [5].

*Chart: **Accenture's revenue keeps compounding at a slowing single-digit pace** Line chart of Accenture's quarterly revenue across twelve calendar quarters, rising from $16.2 billion in the quarter ended November 2023 to $18.7 billion in the quarter ended August 2026.*

*Quarterly revenue in USD billions, Accenture plc (ACN), by calendar quarter; Accenture's fiscal year ends August 31. Source: Accenture quarterly SEC filings [6]. [6]*

*Chart: **Accenture's 2027 guidance low end sits below every quarter since late 2024** Line chart of Accenture's year-over-year quarterly revenue growth from the quarter ended November 2023 to the quarter ended August 2026, ranging from minus 0.6% to 9%, with the 3% to 6% fiscal 2027 guidance marked as two horizontal lines.*

*Year-over-year change in quarterly revenue, in percent, Accenture plc (ACN), by calendar quarter; the fiscal year ends August 31. The two lines mark the 3% to 6% revenue growth guidance Accenture gave with its fourth-quarter results [5]; it is guidance, not a reported figure. Sources: Accenture quarterly SEC filings [6]; fourth-quarter results [5]. [6][5]*

The next reading is the fiscal first quarter ended November 30, reported in December [5]. As the data-center trade sheds value, roughly $4.2 trillion since June by [a Financial Times count](https://finance.yahoo.com/markets/stocks/articles/ai-infrastructure-stocks-shed-4-172156505.html) [3], the integration dollar looks like the durable one: model prices fall and contracts reprice, but the wiring work already paid for stays spent, which is why the bill concentrates where it does.

But size the position against the value statistic. When fewer than a third of buyers can show the board a return, budgets get scrutinized, and the billable hours go with them [1].

## Takeaway

The Wharton-Accenture report puts a five-year bill of $155 million to $520 million on 35 enterprise agents, and says tokens are only a fifth to a third of it: the integration work Accenture sells is most of the rest.

## Sources

1. [Crypto Briefing, Wharton and Accenture report finds most AI agent tasks don't need frontier models, 16 November 2026](https://cryptobriefing.com/wharton-accenture-ai-agents-frontier-models/)
2. [Fortune via Yahoo Finance, Most work doesn't need the Ferrari of AI, Wharton and Accenture find, but companies may be paying for it anyway, 16 November 2026](https://finance.yahoo.com/technology/ai/articles/most-doesn-t-ferrari-ai-140000997.html)
3. [Yahoo Finance, AI Infrastructure Stocks Shed $4.2 Trillion Since June, but Building Continues, 16 November 2026](https://finance.yahoo.com/markets/stocks/articles/ai-infrastructure-stocks-shed-4-172156505.html)
4. [ZDNET, 3 surveys deliver the same uncomfortable truth about adopting agentic AI, 28 August 2026](https://www.zdnet.com/article/3-surveys-uncomfortable-adopting-agentic-ai/)
5. [Insider Monkey, Accenture (ACN) Soars 15.8% as Q4 Earnings, AI Bookings Crush Disruption Fears, 1 October 2026](https://www.insidermonkey.com/news/accenture-acn-soars-15-8-as-q4-earnings-ai-bookings-crush-disruption-fears-1847513/)
6. [Accenture, quarterly figures from its SEC filings, retrieved 17 November 2026](https://www.sec.gov/cgi-bin/browse-edgar?action=getcompany&CIK=0001467373)


---

Wharton and Accenture put a five-year price tag on enterprise AI agents: up to $520 million — The Inference. Canonical: https://theinference.org/article/wharton-and-accenture-put-a-five-year-price-tag-on-enterprise-ai-agents-up-to-520-million
