More than 30% of the increase in demand for application programming interfaces (APIs) will come from AI and tools using large language models (LLMs) by 2026, says Gartner.

“With technology service providers (TSPs) leading the charge in GenAI adoption, the fallout will be widespread,” says Adrian Lee, vice-president analyst at Gartner. “This includes increased demand on APIs for LLM- and GenAI-enabled solutions due to TSPs helping enterprise customers further along in their journey. This means that TSPs will have to move quicker than ever before to meet the demand.”

A Gartner survey of 459 TSPs conducted from October to December 2023 found that 83% of respondents reported they either have already deployed or are currently piloting generative AI (GenAI) within their organisations.

“Enterprise customers must determine the optimal ways GenAI can be added to offerings such as by using third-party APIs or open-source model options,” Lee adds. “With TSPs leading the charge, they provide a natural connection between these enterprise customers and their needs for GenAI-enabled solutions.”

The survey found that half of TSPs will make strategic changes to extend their core product/service offerings to realise a whole product or end-to-end services solution.

With this in mind, Gartner predicts that by 2026 more than 80% of independent software vendors will have embedded GenAI capabilities in their enterprise applications – up from less than 5% today.

“Enterprise customers are at different levels of readiness and maturity in their adoption of GenAI,” says Lee. “And TSPs have a transformational opportunity to provide the software and infrastructure capabilities, as well as the talent and expertise, to accelerate the journey.”

Throughout the product life cycle, TSPs need to understand the limitations, risks, and overhead before embedding GenAI capabilities into products and services. To achieve this, they should:

* Document the use case and clearly define the value that users will experience by having GenAI as part of the product.

* Determine the optimal ways GenAI can be added to offerings (such as by using third-party APIs or open-source model options) and consider how the costs of new features may affect pricing decisions.

* Address users’ prompting experience by building optimisations to avoid user friction with steep learning curves.

* Review the different use-case-specific risks – such as inaccurate results, data privacy, secure conversations, and IP infringement – by adding guardrails specific to each risk into the product.