New Data on Generative AI Budgets, Challenges, Use Cases, and Impact on Innovation

By Alex Slawsby |  June 13, 2024

Two sessions at this week’s Front End of Innovation Conference in Boston explored the ways that generative AI is being tested and deployed inside large organizations. One session was led by Kate Carruthers, Chief Data and Insights Officer at the University of New South Wales in Australia, and another by Bradley Shimmin and Natalia Nygren Modjeska of Omdia, a research firm.

Seven takeaways:

• GenAI is already being implemented in a wide range of use cases; Omdia pointed to a list of 15 use cases where they are seeing more than 30 percent of their data set of 300+ “early adopter” leaders having implemented GenAI.

Some of the top use cases where those organizations are deploying generative AI: improving employee productivity; external customer support; monitoring control of logistics or devices; content analytics; general purpose, externally-facing virtual assistants; and product or service personalization. (In all six of those use cases, 46 percent or more of the survey respondents said generative AI had already been implemented. Of the survey respondents, 28 percent were based in Europe, 26 percent in Asia-Pacific, 20 percent in North America, 17 percent in Latin America.)

• The top five challenges associated with GenAI deployment:

  1. Security issues
  2. Data privacy issues
  3. Integration challenges
  4. Technical complexity of GenAI
  5. Budget constraints

• Enterprise budgets for GenAI are set to increase in 2024, according to the Omdia survey. Forty-one percent of respondents said their organizations will spend $1 million or more. (Thirteen percent of those respondents were in the $5 million or more segment.)

You don’t need to build GenAI solutions from scratch, Shimmin and Modjeska pointed out: “Lean on partners, open source, and managed solution providers. There are many options, with new ones entering the market all the time.”

• In the context of Gartner’s Hype Cycle, GenAI is somewhere around the “Peak of Inflated Expectations,” with the “Trough of Disillusionment” yet to come. No one knows how deep that trough will be for GenAI — which raises the interesting question of whether or not Gartner should have different peaks and troughs for different technologies. Consider how deep the metaverse went (and it’s still probably there.) Meanwhile, because it democratized so quickly in an “everyone can touch and feel it” way, GenAI is likely to not go very deep for consumers and, despite all the challenges related to data / privacy / power consumption / etc., the same may be true for enterprises as well. It’s going to become table stakes for companies, and so if you’re not experimenting broadly — hoping the trough will save you — you’re already way behind.

• Carruthers of the University of New South Wales predicted that hallucinations are not going away. To mitigate them, she said, GenAI tools need fact-checking so that they do not operate unsupervised. Another thorny issue: questions about data ownership: “When you use AI to create, who has rights to that?” And what about ownership issues with data that large language models are trained on?

• One near-term opportunity for AI, Carruthers said, is automation of routine and repetitive tasks. That will result in the elimination of some existing jobs — but new jobs will emerge, she predicted.

• Adoption of GenAI has big implications for the innovation team. You can apply gen AI to at least somewhat increase the efficiency and effectiveness of every innovation team task — and in some cases, to completely take over the task. That means we’ll need fewer specialized humans to “do innovation,” as more and more core business unit and functional unit professionals will be able to perform some or all of those tasks. But more complex, critical things will remain human-centric for a long time, like figuring out growth strategy; the right balanced innovation portfolio; build / buy / partner decisions; selecting the right tools; considering new opportunities and how they may or may not integrate into the enterprise; change management and building coalitions of willing. Although humans can already use GenAI to assist in all of those areas.

(Featured image by Sean on Unsplash.)