Skip to content

Future Tech/business-models

An Agent Metric Lands in Software IPO Key Metrics by 2029

Software IPO prospectuses list the key business metrics management uses to run the company. The prediction: by 2029, at least three of them add a metric that counts AI agent work, such as tasks completed or cost per task.

// By 2029 · medium confidence · disruption 6/10

Prediction

// 2029

By the end of 2029, at least three US software IPO final prospectuses will list an AI agent metric (cost per task, tasks or resolutions completed by agents, autonomy rate or agent hours) among their key business metrics.

Confidence██████████medium
Disruption██████████6/10

What dies

  • → per seat saas pricing

Who wins

  • → Intercom
  • → Zendesk
  • → Salesforce

filed: 2026-11-17 · guptadeepak.com

The hook

A software prospectus typically sets out the numbers management says it uses to evaluate the business. If agents do the work, a company selling them will eventually have to count the work.

Thesis. Key business metrics follow the unit a company bills in. Once enough companies going public bill per task, per resolution or per action, an agent metric moves from the pricing page into the key metrics table of the prospectus.

The story

The setup

A key business metric is a number the company chooses to present as central to how it measures performance, alongside the audited financials. The choice says what management thinks drives revenue.

Agent pricing creates new candidates. Intercom counts an outcome when a customer confirms resolution, does not ask for more help, or Fin completes a workflow. Zendesk calls automated resolutions "the unit of measurement used for calculating and billing your account for AI agent usage."

The baseline

A count of 750 US tech IPO final prospectuses filed 2004 to 2026 by guptadeepak.com's Rewired Metrics research found outcome-based pricing or the company's own AI agents in 1 of 12 US software prospectuses in 2025 and 2 of 16 in 2026 so far, and none earlier. The count is preliminary.

Describing agents is not the same as reporting a metric about them. The research has not yet found an agent metric in a key metrics table.

The threshold

The prediction is right if at least three US software IPO final prospectuses filed by December 31, 2029 each list an agent metric among the key business or operating metrics management presents. Qualifying metrics are cost per task, tasks or resolutions completed by AI agents, autonomy or automation rate, and agent hours.

A metric mentioned only in the business description or risk factors does not count. It has to appear where the company lists the metrics it uses to evaluate the business.

Why it could fail

Companies may fold agent revenue into ARR and report nothing new, since ARR is already familiar to investors. Lawyers may also resist a metric whose definition is still unsettled, because a key metric invites questions about how it is calculated.

First signals (verify today)

Vendors already bill in agent units: Intercom per outcome, Zendesk per automated resolution, Salesforce per conversation and per action. A company that bills in those units has the number. The Rewired Metrics research has not yet found a US software prospectus that lists one as a key business metric.

Key data points

  • Intercom counts an outcome when a customer confirms resolution, asks for no more help, or Fin completes a workflow (Intercom pricing page, fetched 2026-11-17).
  • Zendesk: automated resolutions are "the unit of measurement used for calculating and billing your account for AI agent usage" (Zendesk pricing page, fetched 2026-11-17).
  • Outcome pricing or own AI agents in US software IPO prospectuses: 1 of 12 in 2025, 2 of 16 in 2026 to date (Rewired Metrics count, preliminary).
  • No agent metric found in a US software prospectus key metrics table so far (Rewired Metrics research, preliminary).

Contrarian angle

Most commentary expects ARR to absorb agent revenue and stay the headline. The bet here is narrower and checkable: three companies is a small number, but each has to choose to define and defend a new metric in a document with legal liability attached. That step is a stronger signal than any amount of AI wording.

The flip side

What this kills

The paired obituary in Tech Graveyard.

Read the obituary

FAQ

Why three prospectuses and not one?

One filing can be an outlier. Three separate companies choosing the same kind of metric suggests a practice rather than a single company's choice.

Does ARR from agent products count?

No. The metric has to count agent work or its cost: tasks, resolutions, autonomy rate, agent hours or cost per task.

How is this checked?

By reading the key business metrics section of each US software IPO final prospectus filed through 2029. It is a manual check, because the metric names are not yet standard.

More from guptadeepak.com

Want the technical deep-dive behind this prediction?

Read the companion article

More from the business-models desk.