---
title: "Should a founder build Machine Relations in-house or hire an agency?"
description: "Keep the judgment in the company. Hire the placements and the sampling a small team cannot staff, and let the tool count what those people produce."
canonical: "https://jaxonparrott.com/answers/agency-or-in-house-for-machine-relations"
last-updated: "2026-11-14"
---

# Should a founder build Machine Relations in-house or hire an agency?
Keep the judgment in the company. Someone inside has to open the cited sentence and say whether it is the one a buyer should see. Hire the placements and the sampling you cannot staff, and let the tool count what those people produce.
The person inside the company opens the answer, reads the cited sentence, and says whether that sentence is the one a buyer should see. That reading is the job you keep. An agency that places earned coverage shaped for retrieval is doing Machine Relations work. A tool that shows the cited URL is doing measurement. A retainer that sends clips with the URLs attached is showing you the work. AuthorityTech's page on in-house, agency, and tool is the execution map. You decide what done looks like.
Hire for a named question shape, a named source class, and a way to see whether those sources appeared. The pitch I take can say which buyer questions it is aiming at. The price I take is tied to those sources. The deliverable I want is a page the company does not control, which an answer used, plus the release that showed it.
In-house carries the judgment and needs outside capacity when the relationships or the sampling volume are beyond the team. The combination that works is an internal owner of done, plus the outside capacity required to earn and to sample. The Index is the scoreboard both sides read.
A GEO title earns its salary when the person can say which sources count. An agency retainer earns its fee when the prompt list is the buyer's real questions. Both change what an engine retrieves when the sources show up in the next release.

## The Index number

This release validated 144 of 151 measurable source segments across 36,189 answer runs. The work of earning a place in those segments is placements and sources, not a job title.

Published November 5, 2026, updated November 14, 2026. Counts from Machine Relations Index release November 5, 2026 (`mri_score_v2.1+2026-11-05+66c560c07279`), runs collected May 10 to November 5, 2026. Dataset: [doi:10.5281/zenodo.23652423](https://doi.org/10.5281/zenodo.23652423).

- Definition: [Machine Relations](https://machinerelations.ai/glossary/machine-relations)
- In practice: [AuthorityTech on who owns what across in-house, agency, and tool](https://authoritytech.io/blog/in-house-vs-agency-vs-tool-ai-visibility-who-owns-what-2026)

## Related answers

- [What does it mean for a brand to be cited by an AI answer engine?](https://jaxonparrott.com/answers/what-being-cited-actually-means)
- [What does a share of voice spike tell a founder?](https://jaxonparrott.com/answers/when-to-ignore-a-share-of-voice-spike)

## FAQ

### Can a tool replace an agency for Machine Relations?

A tool measures. An agency can earn the pages. Someone inside still owns what counts as done.

### What stays inside the company?

The ability to open an answer and say whether the cited sentence is the one a buyer should see.

[All answers](https://jaxonparrott.com/answers)