As Accenture and Nvidia seek to accelerate enterprise adoption of AI, their first testing ground has come closer to home: the consulting giant’s own marketing department.
Last month, the companies announced the creation of a new Accenture division called Accenture Nvidia Business Group, which plans to train more than 30,000 people using Nvidia’s full technology stack: Nvidia AI Foundry for creating custom AI models, AI Enterprise for deploying AI solutions, and Omniverse for 3D Simulation and collaboration. These efforts, part of an expanded partnership, also include Nvidia’s “NIM Agent Blueprint” to help companies more easily develop enterprise-grade AI agents to learn, adapt and automate complex tasks.
To show how the AI platform works in practice, Accenture’s marketing team wanted to reinvent how the department works using AI agents, said Jill Kramer, Accenture’s chief marketing officer. The first step was to determine which parts of the process could benefit agents the most, including which tasks required the most brainpower or were most repetitive.
In a recent interview with Digiday, Kramer explained the end result: More than a dozen agents helped reduce the total number of steps in a marketing project from an average of 135 steps to 85. Some agents help marketing research, data analysis, and social media planning. Others help research internal documents, write strategy, and identify how to fund projects based on other parts of a budget.
“Someone described it to me as thinking you’re racing at this track, but you constantly have to make a pit stop,” Kramer said. “Research team, extract this data. Analytics team, pull these reports… Depending on the activity of these other teams, you would get something in a day, a week, a month, and then you had to integrate it. Now (agents) bring it all back to you and help you integrate it.
Accenture and Nvidia are just two of many companies heavily invested in the development, development and scaling of AI agents and co-pilots. Others in recent months include Snowflake, Salesforce and Microsoft. Last week, Microsoft made his debut a new Azure AI Foundry – previously known as Azure AI Studio – which came as Microsoft prepares to launch its highly anticipated AI Copilot Studio.
The development of enterprise-grade agents is part of the ongoing race toward the adoption of generative AI, but widespread adoption has been slower than some would like. According to a recently published report by Deloittethe number of companies launching AI agents will increase to 25% in 2025 and 50% by 2027. Meanwhile, only 30% of AI pilots will reach full production. However, analysts expect AI agents to stick around.
“Generative AI and ChatGPT have shown that there is consumer appetite, and big tech companies are responding in kind,” said Forrester principal analyst Stephanie Liu. “That said, as with all emerging technologies, there is a lot of noise and hype around AI agents, and companies need to do their due diligence to identify their use case and the capabilities they have really needed.”
Initially, Accenture tried to deliver AI tools to as many people at once, but Kramer said that created anxiety among those who weren’t using them, while actual use cases were limited . Even after deploying AI agents, Accenture had to adopt a new mindset to help employees rethink marketing rather than readapting tools to old operating models.
“It’s any group, anywhere: If you ask them to work in a fundamentally different way, there’s an initial organ rejection,” Kramer said. “The most important part is this capability: we have changed our entire operating model. We have changed the way our processes work to avoid any organ rejection.
Rather than building tools for simpler tasks like translation, Lan Guan, Accenture’s chief AI officer, said the company is tackling more difficult marketing problems. This meant creating AI agents for casual inference, strategic planning, and pricing optimization. It also requires using a powerful computer provided by Nvidia as well as the latest major language models like Meta’s Llama 3.
“They want that holistic view, what I call connected data, so they can really understand the root causes,” Guan said. “Almost like human detectives when we put these puzzle pieces together. With 1,500 puzzle pieces, you don’t want to miss a single one. »
Although Accenture is starting with marketing, the platform can be operationalized in other areas of the business, said Justin Boitano, vice president of enterprise AI at Nvidia. Indeed, it was designed to take into account the way teams work together and to embody different roles and characters. Although Accenture’s early adopters were quickly impressed with the LLMs’ initial tasks, he said enterprise applications took longer to build and improve.
“People are realizing that we need to treat them not just as AI models that magically do everything, but also as employees that we bring into our business,” Boitano said. “Basically, we incentivize them to do the right thing for a certain business function, we monitor them to make sure they’re doing a specific job, we evaluate them in doing that job. There is a process there.
Adapting general-purpose AI models to understand a company’s specific language, vocabulary, and jargon still remains a challenge. Others include skills gaps, performance gaps, rising costs and regulatory risks. However, Gartner analyst Nicole Greene said AI has proven capable of optimizing workflows and creating more personalized content. It will simply take time to develop AI models powerful enough to act autonomously.
“Marketers should be wary of ‘AI Agent Washing,’ where vendors advertise AI agents, but few will live up to the name,” Greene said. “Because AI agents are designed to act autonomously and proactively in an environment, and because they often learn and adapt as they operate in their target environment, they pose risks considerable. Organizations that do not systematically manage AI risks are much more likely to experience adverse consequences.
Prompts and Products – AI News and Announcements
- Cognitiv and Index Exchange are the last companies to partner on another way to use deep learning algorithms and large language models for programmatic media.
- With companies like Perplexity and Amazon adding AI tools for e-commerce, there are growing interest in improving the personalization and transparency of search engines powered by LLMs.
- Perplexity announcement a new Perplexity Shopping platform to help people find and buy products using AI generative search.
- Pratik Thakar, Vice President of Coca-Cola and Global Head of Generative AI, spoke with Digiday about the brand’s AI-generated holiday ad.
- French AI startup Mistral has added a new way for its rival ChatGPT “Le Chat” to search the internet and create high-quality images.
- The Content Authenticity Initiative said it now has 4,000 members worldwide, from media companies, technology companies, camera and smartphone manufacturers.
- Snowflake announcement plans to acquire startup Datavolo to offer multi-modal data pipelines. It also struck a new deal with Anthropic to integrate the latter’s AI models into Snowflake’s data cloud.
- Microsoft has announced more ways for its Copilot users to perform repetitive tasks for Microsoft 365 apps.
- Amazon extended use for its AWS AP studio which uses natural language to create and modify applications.
- The parent company of Pokemon Go is would have use user data to train a real-world AI model.
- The Atlantic has built a new database which shows all the ways AI models use Hollywood content.
Thoughts of humans
People often say that generative AI is now easy and accessible, just plug and play, but Lan Guan, Accenture’s director of AI, said this is unrealistic for businesses. She mentioned that a client wanted to use generative AI in their contact center by powering standard operating procedures (SOPs) to their chatbot. However, they found that they had 37 different versions of SOPs, which challenges the idea of a single source of truth.
“What we see here in enterprise organizations is largely what we call enterprise disorder,” Guan said. “Data is messy and existing integration technologies are ubiquitous. »