Providing robust security services to our financial services clients is one of the most important elements of our work. By combining the strengths of our people, technology and configurations, we work continuously to protect our clients against cyber attacks to keep them safe. A key part of this effort is providing regular, insightful reporting to our clients on the state of their security posture and how Lanware and our technology have performed.

Critically, this is not just about the stats; our value lies in the expert analysis and context we provide.  Our security team interpret emerging trends, assess changes in risk and highlight evolving threats that our clients may face. We provide tailored information and outlooks, specific to our clients, so they can confidently respond to questions from their board, stakeholders, or own clients.

Given the scale and frequency of our security reporting, which involves in-depth analysis and tailored report writing for each client, we’ve explored the use of AI to help speed up the process to generate these reports, while preserving the personalised, business-centric and risk-focused analysis our clients value. We looked at two technologies to help:

Microsoft CoPilot Studio – Allows you to build AI agents in Microsoft 365, using a no-code interface. These agents can be assigned specific tasks and given specific knowledge (such as documents), enabling them to act as specialists.

Microsoft Power Automate – A low-code automation platform in Microsoft 365 that enables users to create workflows through a series of steps to perform automated activities within the Microsoft 365 suite of applications.

Using CoPilot Studio, we built two agents that replicate our existing report generation process:

1. Security Engineer – A security engineer typically reviews raw data and statistics to perform an initial interpretation. For our agent, we provided the Security Engineer role access to all previous security reports as a foundational knowledge base. We also gave it a job description (as a contextual prompt), detailing what was expected of the agent, namely that it would be provided with statistics from various security products that it would need to analyse and summarise.
2. Security Leader – When the Security Engineer agent provides its initial response, it produces a technically detailed analysis based on the data. To translate that into our existing language for security reporting, which focuses on easy-to-read, simplified, business and risk-focused analysis, we created the Security Leader agent. This agent was provided with proposal documents as its knowledge base, an existing set of documents that Lanware has, which are always aimed at a high level. Additionally, we provided the Security Leader access to the internet to further enhance understanding and output.

This two-step approach, aligning agent development with our existing human processes, ensured that the outputs we generated closely resembled those produced through our existing reporting methodology.

To tie this all together, we used Power Automate to build an automation pipeline. This pipeline takes in the raw data, passes it through to our Security Engineer agent for initial analysis, and then passes the results to the Security Leader agent. The final output is used to pre-populate a structured Lanware Microsoft Word template, completing the security report. We also leveraged structured Microsoft Excel templates along with Power Automate’s Excel scripting capabilities to automatically populate the Word template with graph visuals to enhance the analysis performed by the agent.

We are still very much in the initial stages of this process. We view the text generated by the agents as the first draft, and our security team continues to validate and update the text as needed. However, this approach has already significantly reduced the overall time required for us to generate our security reports and potentially enables us to increase their frequency in the future to keep financial services businesses even more secure.

This has been our first step into a true agentic AI workflow to tackle a real business problem faced by Lanware. This particular workflow was ideal for AI as it involved data summarisation, text generation and a fixed, structured output, which are the typical challenges where AI and automation excel.

If you would like to see this in action, specifically the Word document generation and Excel graph generation as part of the automation we used in our security reporting, you should try our FCA Security Checklist here.