Make or buy AI? Evidence on AI sourcing decisions and workforce outcomes

We examine how firms acquire artificial intelligence (AI) capabilities and how their AI sourcing decisions relate to workforce outcomes. Using Swedish enterprise surveys from 2019 and 2021 linked to financial statements and employer–employee data, we directly observe whether firms adopt AI and whether they outsource AI-related activities. AI adoption is concentrated among larger firms. Among adopters, pre-existing cost pressure is associated with outsourcing. Relative to non-adopters, both in-house adopters and…

By the Truvace deskOctober 15, 20262 min read1 sourcePeer-reviewedTRV-2026-1508· cite & verify

Make or buy AI? Evidence on AI sourcing decisions and workforce outcomes — "Minimum Wage Event with US Secretary Perez at Boloco" by MDGovpics is licensed under CC BY 2.0. To view a copy of this license, visit https://creativecommons.org/licenses/by/2.0/.

In brief

The study uses Swedish enterprise surveys from 2019 and 2021 linked to financial statements and employer–employee records. This lets the authors observe directly which firms adopt AI and which outsource AI-related activities, then compare their later workforce outcomes with non-adopters.

The findings suggest that how a firm sources AI, by building it or buying it, matters little for who it employs afterward. Adopters did better on employment and skill composition than non-adopters. The abstract does not establish whether adoption caused these differences, and it reports only a short window after adoption.

Main points

AI adoption is concentrated among larger firms in the Swedish data.
Among adopters, pre-existing cost pressure is associated with a choice to outsource AI activities.
Both in-house adopters and outsourcers have higher subsequent employment than non-adopters.
Adopters keep persistently higher shares of university-educated and CS-educated workers.
Workforce outcomes are statistically indistinguishable between in-house adopters and outsourcers.

The gain

Swedish firms that adopted AI, whether building it in-house or outsourcing it, had higher subsequent employment and persistently higher shares of university-educated and CS-educated workers than non-adopters.

The rundown

The data combine Swedish enterprise surveys from 2019 and 2021 with financial statements and employer–employee data. Because the surveys directly record AI adoption and outsourcing, the authors do not need to infer AI use indirectly.

Adoption is concentrated among larger firms. Among those that adopt, firms already under cost pressure are more likely to outsource AI-related activities instead of developing them in-house.

Relative to non-adopters, both groups of adopters show higher subsequent employment and persistently higher shares of university-educated and CS-educated workers. In the year after adoption there is no further change in workforce composition, and in-house adopters and outsourcers cannot be statistically distinguished.