---
title: "Bring It In-House"
canonical: https://withicarus.com/thesis
site: Icarus
description: "Every industry is bringing in-house the work it used to send out, and AI agents make it simple. Icarus's thesis, backed by ANA, Gartner and MIT data."
---

# Bring It In-House

Every industry is bringing in-house the work it used to send out, and AI agents make it simple. Icarus's thesis, backed by ANA, Gartner and MIT data.

By Tanmay and Vibhor. Published 2026-09-29.

For decades, most companies sent their specialist work out. Ads went to an agency, photos to a studio, video to a production house, and the same pattern ran through law and customer service. Doing it yourself meant building a team: hiring a photographer, a designer, an editor and a producer, then keeping them busy all year. For most companies that never added up, so the work went out.

## Companies were already leaving the agency

Big brands saw it first, because they could afford the team. Among the advertisers in the ANA, the share with their own in-house team went from 42% in 2008 to **82% in 2023**.[1] That was before generative AI. They did it for cost and for control of their own brand and data.

**Share of big advertisers with their own in-house team, 2008 to 2023**

| Year | Share |
| --- | --- |
| 2008 | 42% |
| 2013 | 58% |
| 2018 | 78% |
| 2023 | 82% |

Share of ANA member companies with an in-house team (the ANA calls it an in-house agency). Surveys run before generative AI; the 2023 edition was fielded February to March 2023. Source: ANA surveys, 2008 to 2023. [1]

## Now AI does the work that team was hired for

Klarna is the clearest public example. Using AI, it cut what it spent on outside marketing suppliers by 25%, took $6 million off its image production costs, and **cut the time to make its images from six weeks to seven days**.[2] At Kimberly-Clark's India hub, content that took 24 days two years ago now takes two hours, on an AI platform the company built in-house.[3]

**Time to make the same work, before and after bringing it in-house with AI**

| Work | Before | With AI in-house | Faster | Source |
| --- | --- | --- | --- | --- |
| Klarna, marketing images | 6 weeks | 7 days | 6x | [2] |
| Kimberly-Clark India, content | 24 days | 2 hours | 288x | [3] |

The budgets show it too, in Gartner's 2025 survey of 402 marketing leaders.[4] MIT's 2025 study of how companies use AI found the same thing from the other side: the real gains came from replacing outsourced work and external agencies, not from cutting their own staff, with top performers cutting outside creative and content costs by 30%.[5]

- **39%** of CMOs plan to cut agency budgets [4]
- **22%** say generative AI already cut their reliance on agencies [4]
- **64%** of in-house legal teams expect to rely less on outside firms [6]

> AI brings the work in-house, in every industry. **The companies that win will be the ones that make that simple.**

## It is not only marketing

Legal teams are moving the same way. In-house lawyers expect to bring back drafting, contract work and research, the work they used to send to outside firms.[6] Wherever a company paid someone outside because the skill was too costly to keep on staff, AI is changing the sum.

## Agents make it simple

Most AI tools still expect you to know which model to pick and how to word a prompt. An agent takes the goal instead, and does the job. The person using it does not need to learn the craft behind it; they need to know what they want. Gartner expects 40% of enterprise apps to include agents that do specific jobs by the end of 2026, up from under 5% in 2025.[7]

1. It takes the goal, not a prompt.
2. It picks the right model for each step.
3. It checks its own work.
4. It asks you only when it matters.

## We are starting with content

Content is where the outsourcing is easiest to see. A brand that wants product photos, a campaign, a creator-style video or an ad usually has to brief someone outside, wait, pay and go round again. Icarus is an agent for that work. You tell it what you need in plain words. It plans the shots, uses the product and the model you saved so they stay the same in every frame, and makes the photos and videos. You can use it in Icarus, or from inside Claude and ChatGPT.

## Quality is the whole point

The fair doubt is quality. As one media analyst told Reuters, "if mediocre is all you need, then absolutely you can do it yourself."[3] We agree that mediocre is not enough. That is why Icarus keeps your real product and the same face in every shot, and why the taste stays with you: the agent takes the production off your plate, and the people who care most about the brand make the calls.

Content is only the first place we are proving this. The same pattern shows up anywhere a company sends work out because doing it itself was too hard, and we think agents will make it easy to keep that work at home.

## Sources

1. [ANA, The Continued Rise of the In-House Agency: 2023 Edition (members with an in-house team: 82% in 2023, 78% in 2018, 58% in 2013, 42% in 2008)](https://www.ana.net/content/show/id/79185)
2. [Klarna press release: AI helps Klarna cut marketing agency spend by 25% (May 2024)](https://www.klarna.com/international/press/ai-helps-klarna-cut-marketing-agency-spend-by-25-and-run-more-campaigns/)
3. [Reuters: Global firms use AI at Indian hubs to bring ad work in-house (27 May 2026)](https://www.aol.com/articles/global-firms-ai-indian-hubs-143309000.html)
4. [Gartner 2025 CMO Spend Survey, 402 marketing leaders (as reported by Chief Marketer and Brand Innovators)](https://brand-innovators.com/cmos-turn-to-genai-cut-back-on-agencies-as-budgets-remain-flat-gartner/)
5. [MIT NANDA, The GenAI Divide: State of AI in Business 2025 (July 2025)](https://cloudelligent.com/wp-content/uploads/2026/02/v0.1_State_of_AI_in_Business_2025_Report.pdf)
6. [ACC and Everlaw, GenAI survey of 657 in-house legal professionals (October 2025)](https://www.everlaw.com/press/release/acc-report-2025/)
7. [Gartner forecast: 40% of enterprise apps will feature task-specific AI agents by 2026 (August 2025)](https://www.uctoday.com/unified-communications/gartner-predicts-40-of-enterprise-apps-will-feature-ai-agents-by-2026/)


---

Site: Icarus (https://withicarus.com)
Canonical: https://withicarus.com/thesis
Contact: tanmay@hooked.design