CONTACT
Connie Yang Connie Yang

Insights from the Enterprise AI Deployment Summit

Insights from the Enterprise AI Deployment Summit
3:58

AI Realized 2024 Enterprise SummitI’d like to share some key insights and learnings from the recent Enterprise AI Deployment Summit in San Francisco, which may help shape our approach to ongoing projects and our strategy for ML/AI deployments for DesignMind clients. Below are the top trends and takeaways that I believe will be particularly valuable as we move forward in our work:

1. RAG (Retrieval-Augmented Generation) Models are Being Widely Adopted

RAG models are proving to be highly effective across industries, from media to HR tech:

  • Tubi’s VP of Product, Blake Bassett, shared how RAG systems are improving user experience by providing personalized content recommendations, helping users find what they want faster.
  • Betterworks’ VP of Engineering, Maher Hanafi, highlighted RAG's role in HR tech, enabling more efficient candidate matching and employee management through contextual recommendations.
  • Bank of America’s Head of Data & AI, Awais Bajwa, discussed how AI and RAG models are boosting developer productivity in the financial sector.bullseye and accuracy

However, one of the key challenges remains the evaluation of accuracy. As these models generate probabilistic results, the question of what constitutes ground truth is still being debated, especially when evaluating AI-generated content at scale. Most companies are relying heavily on human-in-the-loop systems, but scalability is still a concern.

We’ve faced similar challenges in our client projects here at DesignMind. In cases where ground truth was unclear, we worked directly with clients and subject matter experts to help them define what constitutes ground truth for their specific use cases. This hands-on collaboration ensured that the models were generating relevant and accurate results, and it also provided our clients with more trust in the outputs of the AI systems. However, as with other companies, we’ve found that scaling this process beyond individual projects remains a key focus area for improvement.

2. Small Language Models for Enterprise Use Cases

A key trend is the shift toward smaller, fine-tuned language models designed for niche enterprise use cases.

A key trend is the shift towards smaller, fine-tuned language models designed for niche enterprise use cases. Large, generalized models are being replaced by domain-specific models optimized for business needs, providing more relevant and efficient results. This is an opportunity for us to explore tailored language models to better meet our clients' unique challenges.

Microsoft-365-copilot-600

3. The Growing Need for AI Resource Management Teams

  • Compliance: Companies like NBCUniversal and Vera AI spoke about navigating the legal and IP challenges related to AI-generated content and image personalization.
  • Security: Leaders such as Jake Martens, SVP & CISO at Aristocrat, emphasized the importance of securing AI models, particularly as they scale across global enterprises.
  • Sustainability: Ongoing assessments of model accuracy and operational efficiency will be vital for ensuring long-term success.

4. The Future of Work with AI Agents

A dominant theme in the conference was the increasing push toward a future where AI agents play a central role in our work. 

A dominant theme in the conference was the increasing push toward a future where AI agents play a central role in our work. Jeremiah Owyang, General Partner at Blitzscaling Ventures  (advised by Reid Hoffman), gave a compelling talk on the path to an agentic enterprise, where companies will work alongside AI assistants, copilots, and agents to enhance decision-making and operational efficiency.

5. Our Early Exploration of Knowledge Graph Integration

Since the open-source release of GraphRAG in July 2023, we've been strategically exploring its integration as part of our GenAI solution offering for our clients.

I'm excited to share that we've begun exploring knowledge graph integration to enhance our RAG models. This is an exciting opportunity to work with some of the latest technology, like Microsoft’s GraphRAG. 

Since the open-source release of GraphRAG in July 2023, we've been strategically exploring its integration as part of our GenAI solution offering for our clients.  This integration has the potential to significantly improve data retrieval and insight generation, and it's great to be working together on something so innovative.

Create a Custom Copilot Using Microsoft Creator ToolsThese insights present opportunities for us to stay ahead by responsibly adopting AI and preparing for future integration of AI into all facets of our workforce. As we continue to explore these trends, there’s potential to apply them to both current and future projects and clients at DesignMind. By proactively considering how RAG models, small language models, and AI agents can enhance our offerings, there’s a lot of opportunities to position DesignMind as a leader in the evolving AI space. 

Connie Yang is VP, Data Science and ML at DesignMind. Previously she was Data Scientist II on Microsoft’s Core Data Science team.

Please check out our AI and Data Science Solutions, including Machine Learning, LLMs, and AI Strategy. You can view additional AI and Data Science articles by our experts.