At least half of organisations large and small plan to take a blended approach to GenAI, using a combination of publicly available tools and building their own LLMs (large language models). While it's not surprising that larger organisations are leaning more towards building their own models, the widespread hybrid approach suggests organisations of all sizes should consider whether developing their own models is valuable

Which approach is your organisation taking or likely to take when adopting generative AI/ LLMs? (by company revenue)

BUILD: Develop our own models, based on open-source models

BUY: Use proprietary models developed by vendors, supplemented with our data

BOTH: A combination of the buy and build approaches

36%

50%

$500m to $1bn

14%

65%

26%

$1bn to $10bn

9%

18%

60%

22%

More than $10bn

BUILD: Develop our own models, based on open-source models

BUY: Use proprietary models developed by vendors, supplemented with our data

BOTH: A combination of the buy and build approaches

36%

50%

$500m to $1bn

14%

65%

26%

$1bn to $10bn

9%

18%

60%

22%

More than $10bn

BUILD: Develop our own models, based on open-source models

BUY: Use proprietary models developed by vendors, supplemented with our data

BOTH: A combination of the buy and build approaches

36%

18%

65%

60%

50%

26%

$500m to $1bn

$1bn to $10bn

More than $10bn

22%

14%

9%

BUILD: Develop our own models, based on open-source models

BUY: Use proprietary models developed by vendors, supplemented with our data

BOTH: A combination of the buy and build approaches

36%

18%

65%

60%

50%

26%

$500m to $1bn

$1bn to $10bn

More than $10bn

22%

14%

9%

MIT Technology Review Insights, 2023