Rise of AI Agents

AI Agents. If you’re in tech—or really, if you’re just breathing these days—you’ve probably heard a lot about them. Let’s be clear: the capabilities of AI models today are already mind-blowing. If predictions are true, the AI models of tomorrow will have PhD-level domain knowledge, making them experts across various industries. Imagine an AI that can not only understand your industry’s intricacies but can enable solutions tailored specifically to your business needs. That’s where we’re headed, and it’s a future worth getting excited about.

Real world impact

We are already seeing enterprises derive real benefits. Klarna and BNY Mellon have achieved significant productivity gains by building AI agents and democratizing AI across their organizations. Klarna's internal AI assistant, Kiki, has enhanced efficiency by handling customer service inquiries and streamlining employee workflows, leading to a 25% reduction in repeat inquiries. BNY Mellon's AI platform, Eliza, has empowered employees with advanced AI tools, improving operations across various functions. McKinsey also announced their Generative AI platform Lilli and have seen significant gains.

It's still early days for enterprise adoption of AI Agents or Agentic workflows but there is a large and growing number of use cases where the technology can help improve efficiency, in fact BNY cited that they have 600 use cases and growing. Addressing this number of use cases needs a thoughtful approach.

Addressing broader use-cases

What we see is that every team in the enterprise wants to build with the available open source tools (of which there are many) and end up building their own bespoke solution that works for their use case but can’t really scale to others.  This may work for a really important core business value driving use case but in most cases the engineering effort to build and maintain a solution for a particular use-case may not deliver the necessary ROI. 

Take, for example, a recent conversation I had with an AI leader. Their finance team decided to build a Large Language Model (LLM)-based app to address invoice amount mismatches. It took about 2-3 engineers to build the app and one to maintain it, resulting in annual savings of around $150,000 to $200,000. But here’s the thing—when you factor in the total cost of ownership, there wasn’t a net gain because there is a cost to building and maintaining. There are several similar examples. 

The DIY Approach: High Customization, High Costs

The DIY method requires integrating different open source projects like Langchain, Llama Index, Eval tools and guardrails, not to mention infrastructure to run the applications and its workflows. The advantage with this is that there is a high degree of customization since you build it yourself but it requires engineering talent and can take time to build and operationalize and hence is suited more for the top priority highest value use cases. 

The Point Solution Approach: Ready-Made But Limited

The other approach is to buy a point solution. This could be a tool that is designed to do invoice processing and can detect mismatches. The advantage here is that it's a solution designed to address the pain point but the con is that there is a high overhead of vendor onboarding, security reviews, risk assessment and this cannot scale where you have 100’s of use cases since it will create silos of 100s of point solutions. 

The Platform Approach: Democratizes AI

At BREVIAN we believe a platform approach is best suited for a large range of use cases. A centralized AI Agent platform that can deliver a strong return on investment by making it easy for different teams to quickly build and deliver on multiple use cases. This approach allows teams to focus on what they do best, without needing to worry about the underlying infrastructure. The key insight here is that most use cases have a common set of building blocks, e.g. search and retrieval, knowledge extraction, iterative query planning, tool use, guardrails, workflow engine. 

The benefits of such an approach is that it can allow for quick experimentation and deployment without having to rebuild all the common modules. It is self-serve for the different business functions so they can operate without strong dependencies on a development team. A platform like this could also give enough control to IT teams to securely integrate enterprise data and also govern its use while enabling business users to drive efficiencies in their own domain. 

BREVIAN AI Agent Platform

BREVIAN provides a no-code AI Agent Platform that scales automation securely across the enterprise. We empower users to leverage their domain expertise to automate workflows without the need for writing code or building infrastructure. By democratizing automation, we help organizations reduce costs, increase productivity, and reduce time to market. 

No-code interface for workflow automation

BREVIAN's no-code interface enables users to create AI agents without needing technical skills. This democratizes AI applications, enabling users to construct agents using only natural language instructions that can perform tasks like analyzing data from enterprise systems (such as ticketing systems, CRMs), automating workflows that involve multiple different applications, and streamlining triage, reviews and decision-making processes.

Integration with enterprise data and applications

BREVIAN seamlessly integrates with enterprise data sources and applications. It automates and manages the data pipelines that ingest, clean, normalize and prepare the data from these systems for use with LLMs. This enables business teams to operate in a more self-service fashion instead of depending on other teams to set up and query the data. 

Knowledge extraction and query understanding

Integrating knowledge extraction with reasoning significantly enhances the quality of answers provided by LLMs. By extracting structured insights from unstructured data, these techniques allow LLMs to access a broad knowledge base that is relevant and context-specific, leading to more accurate and detailed responses. Additionally, reasoning capabilities enable LLMs to make logical connections between different pieces of information, improving the coherence and relevance of their outputs in complex query scenarios. 

Enterprise security

BREVIAN embeds multiple layers of protection to safeguard enterprise data. The platform features built-in mechanisms such as safety filters to prevent unsafe responses and purpose-built models to protect PII and PHI. Additionally, every agent built on BREVIAN supports robust logging and role-based access control, ensuring that only authorized personnel can access sensitive functionalities and data, thereby maintaining a high standard of security and compliance within the enterprise environment.

Use cases

People are hired to do jobs, jobs are composed of tasks and AI automates tasks. There are a lot of tasks in the enterprise that involve understanding some unstructured data (document, email, images, websites) - applying some analysis or reasoning and then taking some action (send email, draft a report, send slack message) can be easily automated using a platform like ours. Following are some examples. 

Compliance

Compliance is a critical area where automation can deliver immediate value. With our platform, you can easily create agents that understand your security and compliance policies. These agents can answer security questionnaires, speed up vendor risk assessments by analyzing responses and compliance reports, and even field security and compliance questions from partner teams on Slack or Teams. 

Trust and Safety

In the realm of trust and safety, our platform empowers organizations to automate the enforcement of community guidelines and moderation policies on submitted content. By deploying AI agents, you can reduce the reliance on manual reviewers, saving both time and money. These agents can detect fraudulent or irrelevant URLs in user-submitted content and even identify fake profiles or posts, automating the review process to maintain the integrity of your platform.

Security Operations

Security operations is another domain where automation can make a substantial impact. Our platform enables the creation of agents that can analyze and summarize threat feeds in the context of your enterprise, parse and analyze logs for signs of malicious activity, and automate the initial triaging of security tickets or incidents. By automating these critical tasks, you can enhance your security posture while allowing your security teams to focus on more complex and high-priority issues.

Conclusion

As we navigate the evolving landscape of AI in the enterprise, it's clear that AI Agents are not just a buzzword—they are a transformative force. The success stories from Klarna and BNY Mellon and others illustrate the tangible benefits of integrating AI across business operations. However, the road to widespread adoption requires a balanced approach. While bespoke solutions offer customization, they can be resource-intensive and challenging to scale. On the other hand, point solutions, while addressing specific needs, often lead to silos and inefficiencies.

At BREVIAN, we advocate for a platform-based approach, where a centralized AI Agent platform can deliver scalable, secure, and cost-effective solutions across multiple use cases. By democratizing AI, we empower teams to build and deploy agentic workflows quickly, without the overhead of complex infrastructure.

With BREVIAN’s AI Agent Platform, the future of enterprise AI is not just within reach; it's already here.