Agentic Workflows & AI Automation
Senior Cloud and AI Engineer · 4most
July 2025 – April 2026
AI-Enabled Operations Optimisation
I was asked to review the existing Operations ways of working and identify where AI could be used to reduce manual effort, improve decision-making, and introduce scalable efficiencies across core workflows. Through this review, two high-impact opportunities were identified: resource optimisation and RFP management.
Challenge
A key challenge for Operations teams is ensuring the right people are assigned to the right projects. This requires balancing both capability and personal interest. While skills and qualifications are essential, long-term engagement also depends on ensuring people are placed on projects they find meaningful and fulfilling.
Historically, this process relied heavily on one-to-one conversations and the Operations team’s informal knowledge of individuals’ preferences, experience, and aspirations. This approach worked at a smaller scale, but became increasingly difficult to maintain as the organisation grew, teams changed, and employee turnover increased.
A similar challenge existed within the RFP process. Teams were spending significant time searching for previous RFPs, identifying relevant examples, and manually copying content to create starting points for new proposals. Although templates existed, they were often too generic, which meant teams frequently reused past RFPs as the basis for new submissions. This created inefficiency, duplication, and inconsistency across the process.
In both cases, the core challenge was the same: the underlying data needed to be structured and organised in a way that would allow agentic workflows and automation to be applied effectively.
Resource Optimisation
Databricks was selected as the technology stack for the resource optimisation solution due to its strong LLM capabilities, including AI_QUERY, and its ability to integrate with Genie for natural language querying.
As part of the quarterly review process, employees were asked to provide free-text input on their interests, career goals, and the types of projects they wanted to work on. This information was ingested into Databricks, where AI_QUERY was used to extract structured, deterministic attributes from the unstructured responses.
This enabled the business to move away from relying solely on informal knowledge and manual conversations. Instead, employee preferences could be captured, updated, and used systematically over time to support project allocation decisions.
The solution was deployed through a Genie space, allowing Operations teams to query the data using natural language. Genie handled the SQL logic behind the scenes, supported by custom tooling for more specific operational requirements.
Timesheet data and RFP descriptions were then ingested to create a broader view of both employee experience and the upcoming pipeline of work. This enabled Operations teams to make more considered allocation decisions, matching people to the most suitable opportunities across the pipeline rather than simply assigning them to the next available project.
RFP Management
For RFP management, the goal was to reduce the time spent searching for previous proposals, manually assembling first drafts, and reworking content to match internal standards. The first step was to centralise the key proposal assets, including historic RFPs, approved templates, reusable content, and examples of 4most styling and tone of voice.
This created a single knowledge base that could be used for retrieval-augmented generation, allowing teams to retrieve relevant historic content and combine it with standardised templates and company-specific styling guidance. Rather than relying on generic templates or manually searching through previous submissions, users could generate stronger starting points based on client requirements, sector, service line, proposal context, and established 4most ways of writing.
While Databricks could have supported this workflow technically, Microsoft Copilot was the more appropriate choice for this use case. The solution was fundamentally a straightforward RAG implementation, and because it needed to be used by a wider group of people beyond Operations, accessibility and adoption were critical. Copilot’s integration with Microsoft Teams made it the obvious option, allowing users to access the capability directly within the tools they were already using day to day.
By centralising RFP knowledge and making it accessible through Copilot, the solution reduced repetitive manual work, improved consistency across proposal outputs, and gave teams more time to focus on tailoring responses to each client rather than searching for and reformatting existing material.
Key Achievements
The work created a scalable foundation for AI-enabled Operations. Resource allocation became more data-informed, balancing project suitability with employee satisfaction, while the RFP process became faster and easier to manage through improved search, retrieval, and reuse of existing content.
Most importantly, both solutions focused on organising operational data in a way that made it usable for automation, natural language querying, and future agentic workflows.
Key Achievements
Resource Optimisation
- Improved employee satisfaction and retention by aligning project assignments with both capability and personal interest
- Increased operational agility by enabling faster, more informed resourcing decisions across the upcoming project pipeline
- Reduced key-person dependency by creating a scalable, data-led approach to resource planning
RFP Management
- Reduced proposal turnaround time by making relevant content easier to find, reuse, and tailor
- Improved proposal quality and consistency by standardising access to approved content, templates, and company tone of voice
- Increased cross-team productivity by making RFP support accessible to a wider group of users through existing collaboration tools
Technologies & Tools
Impact & Results
Reduction in operational processing time
Weeks MVP to production